My Kenwood TL-922A RF PA,  ham linear Power AMP. (and mods)  

I will cover both inspections and modification's. (and myths) and more (this page is UNDER )
Stay out of the Insides, and the LETHAL 3000 volt DC if not a trained tech, for HV certifications (see this  first)  Please be safe 1st. (btw, Lethal means zero 2nd chances!)
Rule #1 Do not run any PA with out a real and proper case ground. (if you float the case the case may go to HV and kill you) ok? is this clear enough? (I use double grounds, line cord and bench)
To use a PA make sure all grounds to the shack bench and the 3rd round pin are good all the way to your AC power service entrance. (a shack local ground stake is a good idea)

Photos first: (click to zoom them) RF PA , Power Amplifier.  (some call it just a linear)  Spec.: 2 KW P.E.P.-SSB ! 1KW CW.   wired strapped on the rear for 120VAC.  Circa 1978 manufactured. (no QSK mod.'s)
Basically a HIGH Qual., Heathkit, SB220.  "A" version does 10M band
If you must go tube's (no  cash $3000 -$6000 for MOSFET modern low voltage PA?), this one is PRIME CUT ! Make sure the fan works first, just after making sure the AC power strapping is set right for your SHACK AC  line voltages.(actual)

Check the AC strapping first , DAY 1
Line voltages are best between 118 to 122 volts AC. (means no 15amp service wiring in the wall use  20a more more) I run 30a.(RV wired from shack to main service AC panel and I get no AC sag when keyed up.
Do not block fan rear.


Near mint.?  mine. Look for damage or even dirt. (all here is Lethal.)
Never operate with TX with covers off as seen below or flood the room with RF energy, you will and is not at all safe for eyeballs. (ever) do not cook your corneas. or touch 3000vdc ever or death happens.

Photos by a  licensed trained and skilled technicians never by granny lacking a Teflon suit /shoes/gloves and face mask.  (HV PPE)
Ther is no RF here , unkeyed but 3000vdc sure is. (never work here tired, drunk ,drugged, or confused)
(telephoto Camera  lens zoomed safely) TUBES 3-500z ($240 each at RFPARTS.com)   {  those china tubes are like most things from china IFFY}
Last ones  made real were in USA or France    ,  and ended in 1995 (or near)
Photo below (not mine) ; DO NO KEY UP LIKE THIS.
Below is just  normal Filament heates glowing.  This photo is best done (if at all) using  telephoto lens, with a real camera (mine is Canon EOS)
DO NOT TOUCH 3000 volts it is 100% lethal every time!

The list problems of age and time and evolution, now. (real or not)  My list of worse worries. (there are more)
1978 to 2020+;(now)
  • China made cloned tubes fail... cheap materials and constructed, and now gassy or GAIN (Mu or GM ) is way too high.  (tubes new/newer not to Amperex / EiMac spec) {rumor's  of gain 2 times normal}
  • Some cheap tubes have broken  spot welded leads down on the underside and if on th grid and you lose 2 of 3 grid pins the tube can go into 100MHZ parasitic osciations. (I pass 1amp current on pins 2,3,4 and make sure  all 3 are 0volts  ,bench PS)
  • Bad Cap.'s, happen on all old gear, but did not on mine, (blessed ne and crossing fingers) but Elecrolytic caps can fail (even dry out inside) and those old server mica; caps I sure think we do no have here, DOMINO caps mica not here.
  • Fear of gas or flashover is real on all old gear.
  • The fear of grids not having HUGE copper solid bus bars to the chassis ground tabs. (mostly just wild fear or real bad tubes)? Run the rig with 1 tube yet? see if only 1 tube fails?
  • L7/8 bad? this is DC ground here. You can see that best below, L7/8 chokes but is high Z at 100MHz possible parasitic Ocillations if present at all.  My Amperex tubes never fail. (as so most others with good tubes)
  • C28 to C33 in most cases are not bad and are low Xc at 100Hhz. how low  is  2.4 ohms and at 21Mhz is 11.5 ohms (math)
  • Kenwood did not design the PA  to self protect from gas (flashover) Most if not all parts in the cathode path ARE NOT 2-3KV safe at all,  that means the best protection of all is  GDT (modern today) to the 4 filament pins (p1/5)
  • The second fear and fully justified is that huge flashover landing on the ALC jack and blowing your $1000 transceiver to hell (doomed).( hopelessly dead for ever)  I can promise you solid state will not live at 2KV.
If you lost 2 of 3 grid pins the grid inductance is now 66% higher, and not good that ever. Do not use tubes lacking 3 good grid pints.
Some cheap or overheated tubes have bent cathode  elements, and lost grid to cathode concentric status.  Now wants to arc over. oops. tubes can fail in many ways, not just  dead. (gas too)

The manual states no neutriaztion needed ever. (why is the question?) but you see clear as day see the 2 suppressors below, 
   C28 to C33 if   bad  ?  from
silver migration)  I'd say we do not have Domino mica caps. seen back in 1936 to 50s. (Mine are CDE marron modern caps and are good)
The below drawing shows PA keyed, (oddly) and I added words to that effect.  I added meter details and that the keyed cathode voltage is 8.5vdc (1 volt is at R5) this is what makes fixed grid BIAS D2.
The 3 grid pins not show must not have broke welds inside the tube here. (or it will fail)
A Kenwood , simplified drawing. D2 zener is bios 7.5v? (I revised it tiny)

Quality Assurance: I worked in a factory and with say spot welds in some assembly we did QA work, we took samples of a lot, (if tube broke the glass) and did spot weld pull test if any failed the lost is scrapped. (just one test of many vast)  Quality is NOT FREE.
Hint 1: We took statistically relevant sample counts, to certify the lot. (based on how many product samples can be allowed bad out of 1000 part lots) (1? a trade secret for sure... )
To test any grid triad we pass 1 amp of current on the grid array, and measure voltage drop (must be near zero) using current limit PS, and then shake the tube, see of any grid pin bounces (spikes seen on a scope) if yes the tube is TRASH.  A very simple test and basic.
One more trick is to measure capacitance of elements and shake the tube hard, within G-spec and see if things change greatly, and means loose parts inside if fails. (NO HV present sure) end long complex topic quality. (my boss can talk all day on this...)
At one company we would hold a lot in quarantine for 30 days, if the samples from said lot failed the burn in lab tests (and thermal cycle tests) , the whole lot was scraped.  Again quality is not for FREE.

The PA needs no modifications, as many think, for sure using TUBE exciters  . (mixing TUBES with SS gear does not work if not modified) {mod 3 and 4 below protect your SS exciter (transceiver)
We call Vacuum tubes , Hollow State (a HAM joke) in England (UK) they call them Valves (Fleming fame) I'm a De Forest guy admirer ,  so valves to me fill the bath tub with water. (they really are electrostatic trans conductive modulators, sorry !
The below is what is  popular today not what I did ... but I did do 3 and 4 outside the PC and #2 inside and a pair of flashover snubber (clamps here)
This list is in no order it is only a short list of known and popular modifications (mod.s')

To use this PA on my Heathkit SB104 or my ICOM IC7200 it must be modified (#1,3 &4) just a bit. (but not my SB401, tube rig from 1969) 
  1. The  Gas/flash over protection) HV Glitch resistor? is too slow, as are HV added fuses. No! ,I use only GDT/TVS clamp protectors"in parallel"  that are vastly faster and better. .(seen here and below)  {used in 2 places, the ALC jack and the V1/2 cathodes.}
  2. Mod Power Surge is called soft power on or  In-Rush fix.? , to protect the old AC main  transformers  (I use a $1 fix for this called IN-RUSH a class of THERMISTOR) . (in series in the AC line) {maker has a free sample,ask?)
  3. Mod ALC,  is to allow this PA to use ALC on modern SS (Solid State) transceivers. this PA has 70k ohm source impedance that is no good for any SS gear. (the fix is here)
  4. Mod   SOFT KEY( link is mine) is  too protect your SS transceivers,    (My ICOM  spec. is 16v max! this PA reads 118vdc !)  ( there are many expensive Soft key cards sold, but mine is greatly more simple and  fraction of the cost)
  5. Mod. Meter protector is always a good idea, a simple Diode clamps seen HERE, The 1N5408, 1kv PIV, 3amps and  a whopping 35cents, and a 29cent Zener 20v diode.
  6. RE-CAP, HV section. for sure on 1960s Gear, but not newer rigs (1978) but  risk burning up that rare transformer is real.  My HV caps all test and look perfect, and I think are very top quality Japan factory made. (but can be replaced easy) (unlike these old dogs)
  7. Upgrading old tungsten panel lamps with LED, Not needed.
  8. Plate cap suppressors ,PFC mods (VHF 100MHz, parasitic oscillations? ( if someone lost the PFC devices,or using Asian cloned tubes sure. you may need this but I do not.) This topic is endless and I do not know why.
  9. Hard grounded Grid mods?
I do not use added fuses.
My SB200 suppressor kit is here. (with notes)

History :   (long long ago new tube gear, TV radio or ham, number 1 failures were tubes 90%)
Now:
Things that fail first are mostly electrolytic Capacitors, most makers stated, 20 year design life and may be 2 times that in the real world or longer,  I have gear here prefect, with 50 year old NEC CAPS.  A top brand for decades, is NEC. (or Nat. Panasonic)
Electronics in general can last 100years,  even some EEPROM have spec, data retention stated as 200years,  The other short lived parts are tubes and switches. (but R,C,L Q1 and IC1, are very very long lasting in general) if you get my drift.
The black ones in black cardboard not so good.
We do not have Domino silver mica caps so we dodged that bullet , yeah!  (any parts cracked ,leaking or bulging , or burned are bad) sure looking is first. (slowly)
Sure weak tubes:
Sure  gassy tubes if  abused,  my 1985 Amperex tubes are perfect, and I have tubes (cached) here that are 60 years old from the 1960s boxes full of them and all good. So old tubes do not just go gassy sitting in box. Abused yes. (overheated yes)
Good tubes like famous RCA they have huge books on this topic that I have read showing the extremes they went to, to not get gas, special glass (super special) and special pin wire plating ( matches glass thermal dynamic expansions) and hyrdogen gas , used to get all occluded gas out of the tube hot)
The tubes also have  2 stage getter, (mean get out gas) stage one  it is flashed off when new after that magic then the getter alloys park there, eat up gas for the rest of its life, for when you overheat the plates and gas migrates out, and the getter now eats up that new gas.
This tube the getter is a material that is deposited over the PLATE anode surface. As the plate warms up it is now a gas eater. (getter)  Do no over heat plates on purpose that is abuse.
All other failures are random. mostly.




ALC MOD, DONE OUTSIDE THE PA !  (SS = solid state , electronics)  MOST FOLKS JUST SAY I DON'T NEED NO STINK'N ALC,  some are correct.
Now some cures: AUTOMATIC LEVEL CONTROL (OR a better name ANTI-IMD controls) Tube based ALC will fail,  the impedance is dead wrong for SS is why.
(exceptions) If your transceiver puts out less than 140W-PEP, the use of ALC is useless due to the 3-400z pair can not be overdriven , the rule of  HAM AG6K
One more mod related and optional is the ALC jack clamps seen here.(SB200)
This mod is 2 fold,  #1 protection and #2 is make ALC actually work ( TUBE to SS )
I made this and it works great !

ALC on this PA is not safe for my  ICOM-IC7200 and any modern SS Transceiver nor functional.
Just R10 alone is no good for SS.
Many folks have no use for ALC, so do not read the below, but ALC does lower IMD. (your call and your splatter)

ALC is Automatic Level Control, this is the magic wire from Exciter to the PA that works as a team to keep the PA out of non linear PA tube regions and not create IMD (inter modulation distortion) or Intermod for short.

Problem child #1 ,  see the  signal is adjustable @ VR1 , most old rigs like this expect your exciter to have 1 meg ohm ALC input impedance, back 50 years ago. (SS modern  we see 1000 ohm input Z)   1000 times off and wrong.
 Z means Impedance in Electronics. (AC resistance)
Hark ! , even Heathkit ended that HI -Z inputs design in 1974 SB104 SS. (they cooked there own goose that day, yes progress moves on...)  New ALC spec. bingo. (about 1k ohm inputs not 1meg)
You will see ALC being dead, using (22+10+68k = 100,000 ohms source impedance into 1k ohm, SS transceivers.
 The ALC putput should be near, -4vdc but now is  dead as a door nail. (expected by me sure)

The ALC cure,
is seen below with my ALC card (board) ALC working is a good thing, for sure yelling into the SSB MICROPHONE. (hihi) or in RTTY mode.
Step #1 (cure)
Turn the POT below VR1 10k max clockwise and let my board do this job. But wait,  this is wrong, my TL922, puts out -28v at  max CW and the book for lies! (my scopes do not lie and is)  well not lies just they used different exciters...
Now that my ALC box works I set this pot to -8 to -10v keyed and mic yelled. See slide show just below for scope views (proofs)
One more fear trick is put a Transzorb at J2 to ground.  (so R10 explodes not things connected to ALC)   TRANSZORB Transient Voltage Suppressors these are modern parts not seen in 1978 and are fast ! Vast gear now use these modern devices.
Mouser has 6000+  Transzorbs to chose. SA36a should do the job. (super fast acting in uS time) but I put a SAC10 in the boxed external ALC buffer. (below)
I consider, R8 ,VR1 and R10 all 3 sacrificial (fuses) to any damage to any exciter or any transformers. (or even working as a team with the buffer below)
Public tube horror #1 is R10  just this one part kills ALC dead to all SS GEAR.  (that 68k R10 tells you this ALC only for TUBE exciters! (now the cure)    HIGH Z circuits for TUBES only.
If you
had a modern $3000 SS PA, that PA has a vastly smarter ALC built deep into itself, that prevents IMD even if the old Transceiver has dead ALC circuits and can not team up the this new PA.  ( the makers tell you this and is true)
( a $3500  ALS1300 cure? MOSFET PA !) It warns you that ALC is not optional. But in this  case its self ALC works much better.
The new PA EVEN checks VSWR and runs higher ALC to protect it self. (and arcing over tuners and band switches sure)

My ALC  PA output (unloaded to a scope, turned VR1 full up is -28vdc (SSB,mic, yelled PTT) It's too hot this signal , so at the PA ACL pot see this to -8vdc max.
Then I adjust  my DIY home brewed R1 below for any ALC I need to my transceiver 0 to -8vdc. (tested now and works)

My fix  will stop IMD in its tracks, using a working ALC buffer board (This also worked on my SB200 PA circa 1964  to both my newer SS transceivers and my IC7200 (2008))
All parts from Mouser and the Wall wart here. ( I collect wall warts, all, and find them off old land line telephones at this voltage)

CURE#1 (my cure) I LIKE ALC so I fix them.  (But, I am told by experts below 140watts exciter input we do not need ALC)  I will use a spectrum analyzer to see if I get any IMD at full 100watt excitation. 
You can not use a DC output power pack or D2 goes dead.  
I will do this as an external box . (I may put 1 transzorb device in the PA at J2.)
This PA has very little spare room inside. (and keeping it stock is a goal too, "resale value" matters to many)
Goal 2 besides works is R10 in PA is sacraficial as is the whole box below, for  if Gas and flash overhappen; nothing expensive or  not sold gets damaged (T2 in PA too counts)
R10 will blow up over 200vdc.(at flash over) if one wants one can replace it to 1/4watt or less but is for me just a fuse. (as is)

If you want to use higher voltage AC power packs below, (limit is 35v peak, or 24v RMS) We all know that a door bell transformer or a HVAC home systems all have cheap 24vac transformers.
I have used 15vac pack found at a flea market, but the reads 19vac and gives me  25vdc my raw rails., the rails must be 19 to 35vdc, mine is 25v.
I just added D4b a "Transzorb" TVS Unidirectional, the band goes down to ground on D4b this protect my ua741 chip from the nasty -28v ALC the Kenwood can do. 
TVS can even eat up 100amps for 8.3 milliseconds and are used today on vast devices,  and expanding use and choices. I really like them and trust them. (and what is so cool? they are so inexpensive)
I used 5 protective devices below, just  in case V1/V2 in the PA went gassy or  VHF parasitic  oscillation went  mad. (TPTG )
All grounds below have large wire to BNC ground tab.  (make the grounds strong) #18 guage. for sure to TVS D4b

All parts from Mouser.com above.
A quick check of the box, I use : (and super easy to test live or even using 2 dry cell batteries as input to J1 pin with  -1.5v and -9v. Minus pin to jack  J1   and test with meter at J2. Easy peasy lemon squeezy.
To debug this new mod.card, just a DMM meter is all you need, But I also use my scope ,see slide show below.

Case in point. (references)  Looking how other gear functions with ACL inside and outside.
ALS1300 (Manual)
The ALS-1300 produces an ALC control voltage of 0 to –4 volts DC. (the meter dial goes to -7v) (and protects itself via both Forward/Reflected power in the amp "D8 and D9 " that creates ALC output)

Done:


I have added o-scope views of ALC out, transmitting full power 1000w + (100w input) into my RF dummy load  at this link: (slide show)

TESTING 1,2,3.
 
The best test for ALC is RTTY mode, as is yelling into a SSB microphone is hard to get consistent results, but with practice you can.(per slide show)
I last adjust the POT. above so ALC in the transceiver just starts to wink as I yell in the mic. (all manuals on any transceiver state the same or near)
The TL-922 PA surprised me with -28vdc ALC output @70k ohms.  So I reached behind the PA and turned ALC post to -8vdc output at 1000watts TX.
My ICOM 7200 is 3k input resistance at 700uA current so is easy to drive it with my buffer.  (the ratio is 70k to 3k, or 23:1)  this ratio may work direct?,but I'm to scared try this and to allow tube flash over in to my ICOM.
The buffer works  now.  100watt in ,
SSB , yelling in mic, this is what you get peak. (see modulation here)  (not yelling and silent gives no RF,  SSB is suppressed carrier.(LSB here) (nothing like "AM" modulation, at all) PEP= Vrms / R  {Vrms is 283v} 
This is RF OUT. (not AC line)





My SB200 does this in RTTY mode.. A nice comparison to the above.  The above TL922 is  a very powerful amp. 






I may also install into the rig ont he ALC jack of the rig a GDT dischange tube device 75v, to protect down  stream damage to any exciter, and with a 15 to 25v TVS "in parallel to each other"
What is nice is having lots of things upstream of the SS transceiver ALC jack that can blow up if the TVS/GDT fires off saving a $1000 radio.
I do the protections in 3 places to limit what blows up first.  even R2 in  my box.

End  ALC topic. (Canon EOS T5i, all photos by me)

Next up is:
AC POWER LINE IN-RUSH or SURGE:  (this mod. has only 2 purposes, to prevent flex/flux damage inside the T1 and T2, and save the rocker power switch from damage (hard to buy either now)
If say in one year,  you turn on a PA or many makers and models of PA in your shack, you will witness on some, that the transformer 1 in say 20 times makes this huge load GRUNT sound at 60Hz (USA)
That is you hearing SURGE happen. (and is not good for the transformer)  and mostly happens on cheaper low mass (under sized) PA transformers. (hint better ones do not do this at all due to high reluctance cores)
The power used just in standby is 2amps (RMS) Peak-to_Peak scope is here.
The peak AC line power used at full modulation is 50amps.
The current  surge is from 3 things, just powered on , the transform itself the the load , 2 things, the huge HV CAP.'s charging up and then see the filament currents as wiggling line.
There  are 3 main currents here,  Core flux (saturated) HV CAB bank charging and filament power. (standby current is only 2amps RMS)


Rumor has it the main TRANSformers have : E-I lamination core transformers used here. (and is)
The problem here is called transformer GRUNT, if you turn any large device , system or PA on with huge heavy giant transformers you will hear this loud grunt  sound and that is not good for the copper wire cores deep inside any transformer (not epoxy potted)
 
Even 1 in 20 times this core grunt  happens (it is random or more accurately is the core already magnetically charged) it is not really random in truth, and is caused by 60Hz crossing zero or not. (and core flux state)
I call this FLUX-FLEX. 
That is my job is to cure that and only that, or at least reduce its bad effects. (by at least 1/2)
Goal 2 is to avoid expensive complex cures, seen and sold everywhere, in the Ham world.   MJF makes and external box. (not cheap but works with a 10ohm resistor) 10ohms wins the most popular but sure varies by PA make and model.
The most common fix is adding relays. (for what is called  "soft power on") It is not soft it is just  staged, there are in fact solution that raise AC power in linear way. 0v to 120vac.(and expensive and complex)

The Soft power ON mod. using LOOK MA' NO RELAYS !;  as so  common seen,  the  6 to  22ohm pairs of surge resistors (you only need 1) nor the expensive PCB    and a complex fit inside any packed solid  PA.
No need for 2 resistors ever ! this smart  guy knows this too.! (and no relays either, he uses a SS relay.(a really good and  nice  design !)
Yelled at the sky (only)
Why in the world use 2 resistors?,  there is no ground lugs on the 2 transformers coil primaries.!   (the same current flows in one AC line hot as  flows out the 2nd wire (neutral) so there is no need for 2 resistors +relay contacts 2nd, ever)
The only ground is safety green wired ground.
Only 1 resistor is needed for 120 or 240 voltage, the current is the same on both AC input legs. (Kirchoff's laws are in force and not repealed yet, see first law on current)

This topic (Mine) is only about Flux-flex and or  huge cap charging currents too and only about  that the transformer saturation knee seen below and the huge power surge current, when it happens.
The fact is the surge happens only in 1 to 20 power on acts. (finger at that power rocker switch and that same switch loves to bounce and make FLUX saturation happen so very easy)
YES the power switch can bounce and is not good that.
This  is not new  we had
Globar thermistors in color TVs in the '60s for the same reasons. (not degauss coil reason, but inrush)

The FULL STORY ON IN_RUSH IS HERE (my SB200)

Will the True In-rush please  stand up, well there are 3 surges,  FLUX/Flex  ,  CAP charging and dead cold high current warm up.
Turning on power at the knee below is big time worse surge on 2kw up commercial gear, a zero crossing detector (a very fancy cure) avoids this knee, (a complex fix, avoided below)
The filament (heaters) do not worry me . But reducing cap charge rates do make for longer lasting caps too. not just the transformer.

The knee photos from transformer designers shows what to avoid , happens when the power rocker switch bounces (as this PA has and is easy to bounce with a fat finger action)
The saturated current will cause bad  transformer flex, (call it flux flex) actual  copper wires flex and move inside and wear happens, and is a bad act.
The first rule of curing this surge, is that you can not make it all go away, all you can do is limit the extremes.  (all engineering is a compromise in an imperfect world)

A better test would be to measure both T1 and T2 primaries separate, and to find the flux knee Saturated, next. (not saying our  Transformer does this, no me willing to prove it , me a chicken)
I am not saying our  PA does this knee thing ever., but we do have huge  currents flowing for up to 100mS at turn on. (as seen in photos of scope below)

I call this Flux flex (saturated core)  this does not happen on all PA, and is accidental and can happen with bounced power on  rocker switch.
My readings are here and all photos at the end of this document page. (10 camera tries to get the trace clear)
I first test my sensor using room space heater 1500watt (nichrome resistance wire type) and shows the setup I have is working good as seen here.
This simple test proves scope and sensor is working and scales set correctly The below is using digital storeage mode,  and play back.

This scope view is with the inductive probe
CT013  (donut ferrite caged)   and to 22ohm 1% resistor for   10mv per Amp.   This setup is what the maker of CT013 tells we to do.
 
The turns ratio
of T1 is step up 1:29  to 3000 volts. (SSB mode)
The turns ratio to T2 is reverse of that 24 :1 step down to filaments(heater).
(filament current is (5v/14.5a {x2 ] 29amp hot secondary side) so primary is 24 times less  when hot. (1.2amps)  But cold the tungsten is much higher current.
This heater is 14.5 x 5v =  77 watts each or 154w both  " a pussy cat ! when hot)  we do know that cold filaments use huge currents started up dead cold.
 
  • The use of RS1 in-rush parts work good, the device warms up super fast,  (seconds to 0.1 ohms) and cools in 5 minutes time .
  • The device really works good with that rocker switch bounce, that is why it grunts, switch bounce
  • The only down side that I find is that cool down time, so turn off the PA and wait 5 minutes before turning it back on.  (no problem there no?)
  • No relays to burn up, or or its contacts fail.
  • No huge device that I would be hard pressed to fit inside the PA. (must not be in HV sections ever)
  • No delay timers.
  • Low cost $2?
  • UL rated for direct line AC installs. 
After testing for long periods of time (finding grunt point is not easy, if ever ) and Cap's at 0vdc to start.
I see 100amps max and is effective Z= E/I   120/100 = 1.2 ohms so a 5 ohm RS1 is plenty of protection. I grunt goes higher it works better.
The TX current is 12amps.
I am really happy with my RS1.


Next the 3kv T1 side.
The HV side works safer, set rocker to CW mode, for less power turned on. (do so) and less HV. SSB charges the caps to 3kv.
A capacitor at 0v the current flow is near infinite at time zero, then TC= RC happens, the charge rate time constant. 
The limit to current is only the T1  P to S ,  transfer charcteristics.(and not know on our T1) but yah I see 100amps (primary)
This T1 Transformer secondary sees all 4 caps per phase (full wave) as 50uF , as a short at time zero AC on. (so this is a huge surge but is limited by the transformer total impedance (input to output)
The caps in the HV bank are rated for max current,  spec, indirectly via Ripple current max ,spec.  over this value current the cap plate foils flex and low life happens. same is true of Transfomer widings flex, not ever good that.
A new Kemet 220uF@500v cap Imax is 2 to 4 amps max,  over that the caps life shortens. (or transformer does about 4amps max here, so is marginal)
What won't work is connecting one cap like this here, to a 500vdc huge battery bank inlimited current on tap (pretend ) this damages the cap, and if you short the cap when charged the same thing happens but in reverse.(all bad, this) ask rail gun makers this.?
But is not of any matter now, I have a cure for all overload. below.

Today with my test jig and with a new RS1 in-rush thermistor  I get this.  device   .  SL32-5R020
I tested this PA with a box full  of in_rush thermistors from 4 ohms, then 5 , 10 and 15ohms. from the maker Ametherm Inrush Current Limiters
The PA AC power line tests show it  can do 15 to 100amps surge, (seen it and scoped it and recorded it all.)
This is what the RS1 can do below:  Mine runs at 120F in standby RS1. in second time.
and cools in   5 minutes may be long enough.  (and does!)

I find this, data measured by me RS1:   2ohms =100f, 3ohms=84f and 4ohms = 80f and 5ohms at 78F. (SL32)

The RS1 IR gun shows 120F in center and 175 in the edges, so not an easy thing to do. But I used center of my device for checks.

5 ohms 20 amp rated.  (even a reduction to 1/2 normal is good)
The RS1 heats super fast and goes to low ohms.  (R= E/I   R= 1.2v/ 2amps = 0.6 ohms  short seconds after turning on the PA, this is very good, fast warm up as after all the surges is 100mS long or less.
The cool off time is 5 minutes, so  don;t turn on any PA unless off for 5minutes.

This is a cure. Clearly worth $1 to $2 and some work putting it in.
In theory the T1/2 combined input Z is about 1.2 ohms for 100mS max, so in theory ohms (RS1) that becomes 5 ohms(ball park)is 20 amps max or 5 times less in pure theory. but the RS1 must be cool to start.
In fact I can not get over 30amps ever with RS1 in place and 4 to 5ohm cool,  repeated many times now, 10 or more, and is  100 amp to 30 amp reduction 3 times less.
Looking out into the world of ham:
I see others using 47ohm NTC thermistors. (such huge value)CL80 rated for 3amps and 0.5ohms fully hot, this will never work on any 12amp PA. a 300watt sure. but not real PA's like this.
OTC (over the counter) MJF.com  IPC-120 is 10ohm startup. (it is a nice turn key plug in fix!)   (and more design by (good) by W8EXI  wingfoot = 10ohms) and seen here DIY.(2.5ohms)
I like 5 ohms. (have many ohms (boxes of NTC) in my huge attic cache of parts ) so before it was 100amps peak before is 30amps max now.
 (DIY HALL) Love my HALL current probe. ($4 cost + box)  120kHz band width too. nice. RS1 warms up in like 1/3 of a second. (the filaments tell me that)


The DIY  Hall sensor is super cheap and accurate and fast. (the days of $500 current probes are gone , finally)

I made my own HALL sensor and  did more testing (click below link), with it, it is 120 times faster than the blue color sensor above (120khz bad width)

SEE MORE SURGE HERE


I want the most simple device, to stop most of the  surge, even 1/2 is good goal.  (First the theory at Ametherm, for inrush )
I like to use this  INRUSH LIMITERS:
This RS1 below is the bees knees good.

There are 2 solutions 120vac and 240vac,  the latter uses smaller cold resistance device, as current flow is 1/2 of 120vac setup.
My wall power is 120vac 60Hz and 30amps power. (special power jacks here in the shack just for PA , only)
The choices are best discussed with the maker  of the limiter.
Goals are:
  • Not overheat the device RS1, is first.
  • The device SR1 must reach full temperatures, aand low Vdrop, at my stand by current of 2amps. must so that the hot OHMS is super low quickly just ofter power on happens. (mind does, and is 1.4v drop)
  • The Device can cut surge by 50% (1/2) easy even more. how much depends on above and the actual device you choose. I get 3 times less surge, 50amps to 15amps. now. I am happy here.
  • Make sure PA Tube filaments never run below 5vdc, or the tubes spec, in standby or keyed up.  (loss is super tiny, (1.4 / 24 = 0.06v loss volatage at filaments.)
SB200 below is only 1 example and wired 120v and finding room on some PA can be hard.  Exampe #1
I added  a new barrier thermal stip to my T1 side spare space and wired in RS1. (can be connected to F1 input or output this device is UL rated for both ways) It just needs small air space to do its job mine runs at 100F standby.

This is my choice for SB200 and TL922


Power on standby is 2 amps or 240watts.  Key'd up way more for sure RTTY max, is . 12 amps line current RMS. 
Surge current was 50Amps  and now 8.  With SL32 below.


I  (Ametherm SL22, 4R014 4 ohms cold and , 0.12ohms  hot and 14 amp rated, 214C max. temp.    I tried this and get 1.3vdc drop  at 110F standby.
The Ametherm  SL32-5R020    5ohms cold and .08ohms at 10 amps (of 20 Imax) is reading 1.4drop at 85f in stand by. This SL32 drops to 0.7ohms  R=E/I  as it warms up fast to normal temperature (85F IR gun measured)  MY CHOICE is here.
The 1.4vac drop will only cause a loss of filament secondary voltage loss of 0.06v (24:1 turns ratio transformer) means not a problem at all.
I like the 5R020 the best.
I did all testing out side the  PA, using my DIY test box, (AC power box) faster to do for sure. (1hr work total)  Color me done.
Tested MS32-10015  and has 3x less surge that one but is too slow to warm up. and same with Ms32-15012 even slower to warm up. 
SL32-5R020  wins in my book.


See the fast hall sensor I made here.
All my test data is above and at the end of this html page.

NOW HOW TO FIT RS1 INTO THE KENWOOD.  ( one can find  AC power barrier strip and do all this in ring lugs.  or likejust below here ) Be sure, the RS1 do does not thouch any thing that can melt or catch fire. (mine only ran 80 to 100F)

I will use  a real barrier strip and mount it to an aluminum tiny plate that locks down to the nut (6mm) seen on the left. space is tight here. no lie. Not drilling holes is  goal too. (metal filings not  good)
(120vac only) 240v needs different part #. RS1
Now HALL (DYI) sensor results.

I used this for a barriers strip to my metal DIY plate (photos coming)


KEY LINE FIX   (my way)  DONE OUTSIDE THE PA. (best is this)
This ends 100vdc from wrecking the your  solid state transceiver.
This is only for SSB and is not QSK and for sure hot switching your TR relay in the PA.
  (PTT push to talk or SOFT KEY{tested and now working great} This is super easy to do. (and in an  external box this time)
There is a ham called 
JAMES  COLVILLE.  K7RY $9.99 cheap empty PCB, who makes PCB like this and is seen here. (but mine is all MOSFET'S)
The PROBLEM is 3 fold as is the cure. (mine not his)
  1. Keep in that nasty 118vdc key line (PTT) voltage out of my very expensive SS transceiver.
  2. Keep that 35mA current on the same line much lower, in fact at 1mA now, I= 12v/12,000 ohms (super safe for all gear)  Do not arc your transceivers relay contacts, for sure. (very hard to repair that)
  3. Fear (justified) of PA TUBE  GAS discharged (a.k.a flash over) out of PA to your exciter (Transceiver) {or that nasty VHF Parasitic oscillation surge}
The TL922a  has +118vdc key line and at 35mA the 2 relays (RL1/2 combined current) is too much for most SS gear.
See  many PA's  that  will blow up my ICOM (SS 7200)   there are may key buffers (my name) that flat do not work, some are low voltage input only.
Using and optical coupler (3000v isolation is very great idea) but has problems, it needs to be tuned to your input voltage (from 12 to 120vdc ) and many sold like that do not work at all at 120vdc.
How ever one could put such a device OC between Q1 and Q2,

I did this with an external box, and with one cheap 12vdc WALL WART. ( I have spares for my IP web cams that I have )
With D4 below removed and jumpered. and no CUT. (the cut below is for inside PA installations)
I used 2 BNC jacks and one  Japan (style) DC power socket jack to the box. J1, is keyline Send, , J2 is RLcont out to Kenwood 922, and J3 is DC 12vdc wall wart ( or hide all this in side the PA, the Kenwood has no real space inside for this...)
I had to use BNC to RCA(phono) adapter to fit my PA RCA RL_cont. and RG58 cables.
I used 2 MOSFETS, My current to the exciter is now 1mA, 0.001 ampere's,  (nada) at 12vdc. safe and sound.  The Icom send line uses a weak voltage PTT relay.  D1  can be a better  15v TVS.
My PA this NO FLASH-OVER  mod too that helps me.
(in effect this is a custom SSR, solid state relay.) All parts common as nails, (electronics world common) and I used no RCA jacks (loath  them I do, but audio they are ok) The use of only MOSFETS (N-chan-enh-mode) is best, GDT and TVS devices, yah!
The below is not for QSK modes.(full break-in) Do not do the cut below if using my external box build. (like mine)
I like the MOSFETS, they have a nicer grate voltage Vgs , gate threshold and can be poor and still work. (for weak logic low, not perfect 0v but at near 1v and bipoler (BJT) fail easy. Yes one can stack a diode on the emitter to cure that (BJT bad act up)
See my PCB version. (lay out for PCB )  I tested this and it works below.
This is my design if you want to get rich and make cash sell them I do not care, too old to care that. (give me small credit would  be nice)  Note a 12vdc wall warts are cheap for video web cams sold cheap everywhere.
Click to zoom.                   

MOST SS GEAR have weak send key lines (the relay used is weak , in the world of TUBE PA weak only) one would never want to damage RL501 ever, try to find one, in Japan, much less USA, ouch.

The below is ICOM IC7200 actual and Send wiring and parts details, the manual warns you "not for TUBE amps".  Tells you use below Relay to run  external boxed 12vdc relay then to the SS transceiver. bah !
Limited to 15volts !  note the cute 0 ohm resistors. (joking black black black?)
[Key] line is 0 to 4.7v and diode clamped and goes CPU in Radio..  NO HV must land here.
Done: deal: Softkey works, next is ALC card DIY to make. Ignore error in drawing. see above for current versions.
No PCB needed , this works just as well. Note anti static mat for MOS


I like MOSFETS for sure  runs at wilder range of voltage easy. 



Caps checks: (more here)  Most of these CAPS are still good and were not junk class at year built, not 1960 junk Caps, as my SB200 had.  I just  pretend to show what I would do if I needed to do this....
I do not have a full spec on these caps,  no diameter or length , nor % accuracy, +80 to -20% is and was common, so if true that, the  200uf can read, in fact. 360uF  or 160, if Caps are 20% spec then  160 to 240uf is normal.
The new caps can be had in 25mm diameter (gringo 1 inch) and for sure are shorter (less long) TODAY, (evolution of caps) and will fit if you make up cap shims in the rack below, use high density foam, or custom machined Teflon blocks? (do what you must but non conductive)
I did  not reCAP my  HV Capacitor BANK,  mod.'s not shot gunned, unless the gear is from mid 1960s, (junk caps  were and are (R-leak huge !) 
I test the caps with my ESR meter and make sure all resistors related are in spec. (not ruptured, electrolyte spills from caps, no bulging or burns or any thing odd , seen on any CAP)  Eyeball inspections are very good and first.
If you see the 8 caps there the total series capacitance is 25uF (200/8) and the resistors must be lifted 1 by one to do ESR tests.
I do close inspections and tests.  ( ever see relic 1960s parts?)
The caps here are $35 a set. (mouser.com) do not let others charge you 10x that cost. oK? ( I dislike scalpers) 25mm diameter caps will fit ! modern and are now SMALLER ! (evolution happened)
 New caps are smaller and do not leak anywhere like 40 or 50 years ago.
Those resistors run very hot, some Hams do exchange them with 100k ohm versions lowering heat greatly (P= I  x R) and increases future life of CAP.'s.
the Kenwood caps 40 years old now , most are Panasonic (NEC) and are of super good quality, unlike this 60s junk.
OOPS:   failed HV caps. or GASSY tube damaged.  or VHF parasitic damage.


Looks bad is bad, looks good, can still be bad, testing solves all these kinds of problems. (cap meters, ESR meters and live testing solve this, or ripple tests,  even scopes used to measure for noise or ripple and more)
Hot useful links : (mouser.com tops my list of useful.) I do not get paid to say that, ok?
Mouser mica caps.220pf example ,vastly more.

C14-C22 HV CAP bank rebuild?  200uF at 500vdc. 220uf is more common,,, use that.(the diameter is a secret but 25mm would be my guess.) Mouser 25mm diameter so does digikey
I'm told 500vdc caps are now on the re-surge, due to modern SMPS power supplies for PHOTO CELL power systems, many are using them for Green lower carbon earth (90% efficiency).  GOOD! NEWS ( see this list the tall ones and skinny may fit)
This TDK cap comes  in 3 sizes (in mm). Diameter/length as you can see modern day caps are vastly smaller. In Japan they use mm and never Yankee, imperial inches.  25.4mm = 1.0 inches. (easy no?)
2555( 2.2" tall is very short and very modern...)
30
40, mm all.
35
35

Most the parts I need are at mouser, huge selection or at Digikey.com (both are good) Mouser has BOM support and even shared !
Make up a  bill of materials "BOM"  you need first then buy , or shipping gets too high) order 20 things, or for 2 or more projects at once, they have a BOM on line builder at mouser, there to do all that.
Take your time on lists. and order LAST.  (do not buy one 3 cent resistor and pay $7 to ship it, that be so wrong. (if you have 3 projects going buy all 3 project BOM parts at one time)
Keep a good perspective, you are not  buying a $3000 PA,  or $8000  that for and ALPHA Amp. so spending $100 is wrong? on OLD PA? How can this be?


DAY1:
Inspections are next.:  (first is to consider removing MODs others have done wrong or with wrong parts, or poorly.) (4 decades old , expect anything) look first, do not turn it on day 1.
  1. Open the rear case 120/240AC access port (pull rivets easy) and strap it for your power feed (AC) ) { never hook 240AC strapped to 120 or BOOM you just blew it up} <<< first things first !
  2. All grounds in the case and screws related tight and clean.
  3. The PA AC chassis ground must be properly grounded to the HOME power service entrance ground at the ground stake.  Never let any PA float ungrounded for any use or testing. (3kv is lethal)
  4. Check all solder joints for cold solder, or some dweeb used low temp Pb free,low temp , junk class solder,  if the joint does this on a band switch it burns up (or other rare parts) do not cry later. (suck out the old solder and use 60% tin lead solder. (costs more , is more)
  5. The Tubes  Plate(anodes) have  cap  and with PFC  VHF suppressor,  the resistors are bad now this old , warped ,cracked  and read 47/2  (23) ohms, and are not sold now, carbon comps are dodo bird now.  See below for sources of parts.
  6. Check visually that no seen damage, no missing parts, no variable cap fins  bent,  nor tank coils bent, for sure band switch cap.'s crossed over and shorted. Look for hacking gone wild.   See the 3 tests here.
  7. Look at parts inside for obvious burn damage. (look for 30minutes all hatches open (no AC line connected)
  8. Do the NO HV BULB CHECK below this is for Gross OXYGEN gas in the tube (if any Tube has white frosting inside it is bad)
  9. Make sure the PA HV meter shows HV correctly and if  turned off the HV drops to below 100vdc in 2 minutes or less , if not the bleeders inside are bad. (get the PA serviced if this is not clear what to do or how to do this work safely )
  10. Once turned on  make sure there is no seen TUBE gas, or boom happens or blows up you expensive SS transceiver. Do this test standing far back from PA with hands in pocket, (ok this is a test with interlocks, bypassed) if not sure, ask your HV TECH for help.
  11. Never  work on any PA with tubes lacking a working  chicken stick. ( and it grounded to PA case metal)
  12. The T2 no blow up mod.
Most of my mods are to keep GAS /ARC from killing dead other expensive things, like my ICOM IC7200  ! (I have zero fear of grid inductors L7/8 blowing up, as they are sacrificial.)

The so called Bulb tests (3-500z): (and first a crude and safe Oxygen test is first things first) then the HV gas test real is last.  Bad gas looks like this.
No distractions , TV , cell phones,  tired or drunk. (no kids near by either! or wild Pets?)
This first will fail if oxygen is present in the tube and oxygen will kill  filament dead fast. (the test is short , seconds)
  1. The test is this (with NO AC line connected first) Big top lid off, then the screened cover removed next. (HV is not present now but, if you don't understand this say the heck out of the dog house or use a CHICKEN STICK)
  2. Remove both the 2 PLC1 and 2 cap HV devices totally removed both, so the tubes can not get HV, (that tank coil in the center there is lethal,  with HV present turned on and after turned off , lingers and is lethal)
  3. Put the V1/V2 BULB cage screened lid back on with at least 2 screws, in center.Left and right, to activate no death safety switch #1
  4. Now hold the 2nd safety switch engaged (press it) on the left,  on case top left side hold it down and... 
  5. This test is very  dangerous done wrong , HV is present but not to the 2 tubes. V1 or V2.  (this device COIL  will have HV present now , DO NOT GET NEAR THIS AREA WITH YOUR BODY WHEN HV IS PRESENT)
  6. Push the power on  rocker switch, and see the tubes glow like this. (only seen with eye through the lid screen , only see normal filament heater glowing)
  7. If one tube is dead, it is dead. (or Filament wiring bad,   under socket is bad or pins damaged. (etc)
  8. If both glow the  same way,  that is very good sign , if not then one tube has gas. (if heater glow looks odd){means gross oxygen present"  We do not want oxygen inside the tube. (EVER) Do not proceed if this test fails. !
  9. Power off the rocker switch and pull the AC line from the wall., let the PA rest for one hour, or chicken stick it. (or both if you the person living a long time is good )
  10. Put the front panel knob to HV and meter left shows 0v zero volts, if not? wait or fix your bleeders now, death awaits you. (trusting my life to 1 meter is folly. I use a 2nd tool and the chicken stick)  <<  a real tech has seen meters stick all his life, moving coil meters do that !)
  11. Next we put back #2 above PLC1 and 2. and last put back all panels but the one large painted panel
  12. All is safe now only the top huge no grill  panel is still missing, so we can again see through the grill. (top right side)
  13. The PA is ready with AC power and you turn on the power and meter set to HV and CW.
  14. You turn it on and you see 2000v to near 3000vdc in that HV meter needle , next  set rocker to SSB and see HV is max and last look in the grill for this. (ignore this odd blue glow here, It is Video camera lamps or shop black lamp present, just look in the grill holes.)
This poster shows the safe way to look ,grill in place is that.

Here is SB1000 with bad gas. HV is present now. ( this tube was abused,  over powering it got it too hot.)
The unsafe way is here HV turned on, but  a very good photo  GAS Blue is Oxygen , purple is Nitrogen ionization a mix here(a.k.a AIR). (as seen here.)
GAS 101:  (no tube on earth is 100% gas free  , and if  I do a crazy  high voltage test at say 10KV ,  this tube must be gas free at 3000volts. (only)
Factoid #2 is that the PLATE  deep inside the metal has gas lurking, waiting to sneak out  (TRUE) I am told the plate has top surface getter coating !
The #1 cause of gas is owners overheating the PLATE (anode) over 500watts and migrates out of the metal, (so don't do that, so run you AMP under driven a bit ) Some versions of this tube  rated at 600watts.(plate)
#2 is buying fake tubes, from China or Russia?. (with bad glass or very poor factory full de-gasification  technique and no GETTER , and not trivial to do right (RCA can teach you!) USA and French are best up to 1995 made.
I think the Russian tub rumor is just that, a dream they had, never executed.
 
VACUUM TUBES, Parked on a  shelf tubes mostly do not gain gas, but the getter inside can fail (seen on cheap receiving tubes , that huge silver coating on the glass, mirror like, bingo Getter)  This 3-500z does no leak gas, for fun or age, sorry.
(I have vast tubes in boxes, here, all gas free (means not effecting normal operation) back to 1960 , most 99% are gas free at 60 long years,! and counting)
GAS is ABUSE ! period. (or sure cheap poor grade tubes)

The best book on this is from old Collins radio man,  Walter H. KOHL, 1961 seen here, find all words (out gassing easy and learn )
The glass in tubes are not like window glass or drinking glass or even a light bulb glass,  (there are 1000s of grades of glass , this is  commercial grade PA TUBE very special glass)

All tubes made use a very special type of  Pyrex Borosilicate GLASS and KOVAR seals,  that does not let AIR gas walk past the molecules inside this glass.(or at the pins) {great pains and engineering went in to any tube designs USA/EU,  but China I bet not)
The Pryrex codes are 1500 types. (only few work with vacuum tubes)
The #1 glass then was Corning 7720 called "NONEX" for transmitting tubes ended 1995(hot running)
Link2:
Some glass rated at 500C . temps.
Not withstanding , dropped tubes or pins bend hard over, and other abuse.(the top hat can fail and for sure overheated (allowed)  (the PCF plate cap device, on mine acts as a heat sink too, so don;t throw that away)
The overheating PLATE will cause gas, the illegal temperatures, cause liberated deep gas inside the plate, call  Occluded gases and if you overheat the plates (over 500watt) you forced out that hidden gas. ( called COCK PIT ERROR).
Some Hams bake out the gas with 50vdc on the plate, even making up a  bench bake out test bed station (socket , transformer and 50vdc PSU to plates , 5v to filaments )

No GAS , good glow , just so you know what good glow looks like.



Test for GAS 2 ways, first with no plate caps, and power for 1min max, see if filaments glow right, or act odd, (zero plate voltage, done carefully.  BBI, does the AIR BOTTLE TEST, hands in pockets or death.
Test two, if above passes !,  put the plate post caps( PC1 and PC2) back and do test 3, no GAS test.
Put the screened top case lid back on, closing safety interlock and look for any clues of gas, while  turned on 3000vdc in HV dial, SSB  mode 
This is BBI again DOING IT LIGHTS OUT (BLACK AMP ABOVE I GUESS) AND SEE THE TUBES GLOW AND NO GAS. (The  below is normal filament glow)


Test 3: Oh my golly gee, see this fast test. 3kw peak out. (200watts excitation)(a short test never long)

One of the longest lasting tubes (electron vacuum tube) was: BBC's main Northern Ireland transmitter at Lisnagarvey. The valve was in service from 1935 until 1961 and had a recorded life of 232,592 hours.
Some transmitter tubes (never abused ) can last 50,000 hours.  (overheated tubes drive gases deep inside the iron/nickel plate (anodes) and now is gassy. So do not run 100% power.
Quote from Kohl, 1961 page 24 up, in his great book,
Page 27
"for argon, oxygen, and nitrogen the permeation rate at the same temperature becomes less by a factor of 100,000 or more; hence glass may be
considered impermeable to these latter gases for practical purposes, "  (he is talking glass tubes USA made using top grade glass not China made trash.)
I say myth busted, that this  glass leaks gas at all. (abuse or china knock offs will fail)
The tube makers, worked directly with Corning glass to discover and control that this fine glass does not leak.  (if the fan in dead the tube gets to hot and that is abuse or shack at 120degrees F ambient?)
The trick used by many makers of PA, paint the dog house inside flat black. The tube emits IR, energy, (heat ) and lands on the black shields and they get hot,  cooling the tube, not just the air flow, this is something to consider at rebuild time. (cheap flat black spray paint is amazing)

The French tubes are marked 3-500Z/8802. and need up to 20% more drive to get full output.
The prime supplier now is RF PARTS  or RFP,  if you can not find a good USA/French tube go there.  Buying direct from China, will not work well and no easy returns if allowed at all.
Cheap consumer radio receiving tubes were  $1 then (Dumont,etc) and were of  softer and cheaper glass, leaking glass, I am only talking about this one commercial TX TUBE by Amperex.
On all tubes with the getter inside dies,  the tubes life is near the end, for sure overheated.  The PA tubes many have the GETTER covering the anode plates, looks like black paint.
Note no  GETTER seen inside our tube.   (no getter silver mirror coating on the glass (barium),  but I am now told it is a coating over the PLATE.)

The production plant heats the tube very hot at pump down time, getting all abnormal gas out of the tube. (and even flushes the tube with Hydrogen gas using and inductive furnace.
One expert told me this: The gray coating on the tube's anode surface is a so-called "active getter". When the tube's anode gets hot enough, any gas molecules that get close enough to that surface become chemically bonded to it."
When the tube was made they....  (watch this video, it covers the induction RF heater clean out very well,, the words second inductive heater is the Getter activation step "Barium loop seen here".)
Considering all that one way to get rid of huge gas is run only filaments, then if possible run low HV on the plate , in  test jig 500v for 6 hours then 1000vdc for 6hr. more  and work up to full power only if the blue or purple glow is getting less.

Next, up, kids do not do this at home, but this is super ON TOPIC !
Sees this Video too.(thanks Mr. Carlson !) (he tells you how to clean up gas a the end)

One other manufacturing trick is the  baking a new tube with hydrogen inside to eat all that oxygen for breakfast. (there are many methods, all very good)
Not only that trick but  the envelop (glass ) flame baked, the metal inside is INDUCTION heated and run super hot to get all gas out of the metal, as you can see this is non trivial how they do that. What matters is not abusing the tube NOW.

PRC (jelly beans) You must buy 12 , min. order.( direct buys from PRC)



More abuse is letting the plate cap get too hot +250C, or by throwing away the cap PFC coil that works as a heat sink too. oops , why do that?
Or using too much torque on the cap screws.
end gas.



GRID folly & myths and
the like (all for gaining 50watts , really , but I read others don;t like burning up the inductors with bad gassy tubes)? 
There is nothing at all wrong with this factory design, if you have gas deal with gas.
HAM AG6K )read this first, how the grids really work.

If you work on  GG (grounded grid) Amp.'s. PA read what Collins radio said.,
Quote:
"To change the grid connection from inductive to capacitive, provide a measure of negative feedback, and to neutralize the tube"


Facts: the 221 cap's put the tube grid at RF ground. (at 30MHz and below) The COIL inductors put it at DC ground.
Do not guess, do not put in mods that have  no purpose.
If worry about CASCADING GAS flash over is the motivation to take actions then why not use GDT and TVS devices that only cost $2 a set !
Yes blowing up T2 would not be fun, if that is a goal?, then address that.(click link above bingo)
Just to be clear, here is an UN MOLESTED 922a below (there is nothing wrong with Kenwoods design !) The yellow wires are heater filament and the other 3 are grid.  SEE THE 3 Maroon  mica caps for Grid on both tubes, those are RF GG grounds.
   see one hacked up here.
One of my Amperex tubes is 1974  date coded and has no gas. (46 years old today) "think about that " (or Eimac tubes)
The real tubes,  have very very special GLASS envelopes that does not leak gas, (something those guys in China do not understand !)
Look to at the fantastic , aluminum ground plane there, and the spot welded ground studs, why on earth wreck this beauty.?
Then cry that the grid grounded (hard strapped to case)cures gas? (I see many folks do have gas and is cause by running the PLATES too hot) Self inflected pain.
This PA has 10 grid components, not for fun or giggles and was on purpose. (by clear intent and design and everyone knows, 10 extra parts costs more...)

Each of the 3 grid pins have there  inductance you can not change. (the tubes each have 3 grid pins and will look like 3 tiny parasitic inductors and parasitic, capacitance to cathode and plate..
This is GG (RF) Called a Class AB2 amp.  (grid BIAS on all tubes,(Fleming valves, De Forest tubes) comes in 2 classes.  Fixed and SELF , or a mix of the 2)
This PA uses FIXED bias, by placing a plus +100vdc voltage on the cathode that causes the tube to be more near cut off in standby. (this is indirect fixed grid BIAS) keyed up bias is 8.5vc..
Then when you key the RL-cont. jack the RL2 relays energize, and R7 is un-shorted and R7 now sets the new DC BIAS. (again not by accident but clear intent)


Do not go to the dark side and hard metal strap a Daisy chained Grid pin bar, to the grids and then last  to 1 case ground (each tube). < doing  this makes ground even more inductive. (why do that?)
The grids are tuned, to prevent parasitic non HF , VHF self oscillations. Leave them be ! Please. and leave the normal grounds there alone,  the chassis is a ground plane (best),adding or doing some wild daisy chained inductive bus bar ground is not wise at all.
 
The folks claiming all this voodoo have bad tubes, china cloned graphite plate with non spec. HIGH 240 Mu tubes and  gassy, bad sockets bad relays, etc ,etc., ( its only 50 years old: LOL !) bad socket pins, loose screws? oops, and bad grounds in this area. (Carl @KM1H   nails voodoo .)
Some folks do fight VHF 110MHz parasitic oscillations in this amp for sure higher mu tubes.

On this PA I say, leave the grids alone.
The only grid bias is done indirectly to the cathode. The Kenwood manual states " a zero-bias Class AB2 ,   
 via L20 cathode  inductor and R7"  
" Is connected to the filament return circuit biases the cathode positive during transmission (makes the grid negative to the cathode !). "

!yes confusing that line in the manual. What they meant is this.

Thus. the grid is biased indirectly and this voltage is low (key'd) to achieve low  distortion."  DO NOT MESS THIS UP ! As Carl; states.
The manual is more clear.
Mode Selecting Circuit
Zener diode D2 connected to the filament return circuit
biases the cathode positive during transmission (makes
the grid negative to the cathode). Thus. the grid is biased
without floating it. Bias voltage is low to achieve low
distortion. Plate voltage is 2.2kV. idling current 100mA
in CW mode and 3.1  kV at approximately 200mA in SSB
mode. Therefore. plate dissipation power always exceeds
200W.
In SSB mode. the zener diode is shorted by S3 lowering distortion.!!!

The grid is at RF ground at HF freq.. 3 caps one for each matching grid pin, not by accident, but to match each and every grid pin internal inductance (fixed). Xc = 1/2πFC and XL = 2πFL (when Xl = Xc they are in resonance)
The grid is at DC ground with the 2 choke coils 470uH. and go to higher XL at wrong VHF freq. XL for 2π fL (as F rises so does XL) F= frequency, L = inductance, π = Pi or 3.14.  C= Capacitance.
All tubes have parasitic capacitance, (from plate to grid and cathode and even out side the tube, to other near by structures, that is why you need to not change things in the design by Kenwood.
this basic parasitics,  it is far far more complex but these 3 are primary elements. (this is a kids dream the truth is here)

Do not hard ground strap the grid to ground level the 5 parts alone (10 for both V1/2)   but new grid caps may be needed, Silver Mica.

They did good here keeping grid inductance low. (how low, not sure)   The grid is a cage matrix and very good design. (and 3 grid pins)
 Fun is seeing inside, what is there. UI see cathode to Grid C-stray clearly now. The ruler is very nice to do, scaled.
w8ji  Tom';s photo Thanks tom for your words , and about  Laws of conservation of energy on VFO explodes !  and behaving like a "tuned-plate/tuned-grid" oscillator ! TPTG

Wow seriously thick glass there. 



Plate (anode) Parasitic suppressors:   (of the 3 kind , plate, grid ,cathode suppression or even all 3)
 ( the prevention of VHF parasitic Oscillations)   "the plate supressors below are not mine, they are seen on other ham sites" Mine is factory stock.

Only some PA fail and need this work,  not mine but I have Amperex tubes and normal Mu.
The cause is todays high Mu tubes (say "mew") GM gain of the tubes are not 40 not 100 but now 240 Mu (Siemens _SI)
This high gain tubes from say China are too high and love to oscillate at 100MHz, we are told (unless new cures, happen)
My Amerex tubes work just as good in 1985 and today 2020, amazing and no mod.'s inside my rig, yet.
There are also 10 ohm cathode VHF suppressor mod.'s too and the stock Kenwood grid suppressors are/were there day1.
Thanks to hams and QST , AK7M and AG6K
See the words "
These incognito inductors" 
and words "unavoidable VHF tuned circuits"
"For example, the grid of a 3-500Z triode, directly grounded, self resonates at about 95 MHz."  and "At frequencies above grid self-resonance, the grid exhibits inductive reactance and is no longer grounded." Just as Collins radio told us long long ago.
 
We now have newer hard cures.

Topic grid caps: (why) Quoted and with my colored coments.
"Another method of improving stability is to tune out some of the inductive reactance in the grid structures  by bypassing the grid to the chassis with small capacitors. (as Kenwood did)
 'This increases the self-resonant frequency of the grid circuit to a point where the amplifier-tube will have less amplifying and oscillating ability." Oct. 1988 QST 
This forms a parallel tank circuit.{Xl - Xc} (effects mild they say on 3-500x, but wise to use) (but  all  good suppression effects add  up !)
This tube has 3 grids and all 3 have thick wired connections and make the grid super low inductance , unlike old 811 tubes.
At 220pf per pin 100MHz would (calcs) be 11 nano henrys (nH) per pin (x3)  and that new Z below floats the grid pins to high Z state only at  around 100MHz. As you can see the grid goes inductve below Fr.
A floated grid causes it to go greatly negative (electron cloud charges up the grid) and the tube shuts off or goes super low gain at 100MHz. (best I can tell that is what QST article topic above is talking about)

The above quote box, I think they mean the grid caps form a parallel a tank circuit. (x3) Tune out explained.
end grid topic.

Cathode suppressors. VHF.
Some use  10 ohm cathode restore see here as suppressors.
End cathode topic:
NOW TO PLATE CAP SUPRESSORS:

The newest cure seems to be not using copper at all and for sure not evil silver. (silver raises Q a tiny bit at 100MHz. way more are UHF freq.'s.)
The cure they say is high permeability coils of some kind, SST , nichome and more.  (and impossible calculations for that reason , endless alloys exist and no 2 the same )




Is the real or Memorex, or Voodoo?, well gas is real and so is VHF parasitic oscillations if present at all, ever.. 
What is Vooduu is the physical cures. (the act of making 100MHz devices seems endless.) for sure if it messes up 6 or 10meter bands. (if have em')
   The cure is Low Q at VHF and High Q at HF.  
No proofs by me only what others did. 
Most of this is using China or what some call Pacific RIM tubes that have Mu of 240+ way too high.  ( I've been all over the Pac.RIM and the EAST side makes ZERO this TUBES , )
My buddy calls those tubes, C4, cheap china copied crud. (or worse puts in commie)
Cheap glass that leaks air, or pins leaking or crazy wrong Mu. (or tons of gas deep in the plates and peculates out at normal power levels of the PA , called Occluded gas in the business) or omg element spacing all wrong inside.

Q means quality of merit in tank circuit.  and very simply
Q= X/R 
The R factor if high , simply burns up the power wasted as heat. (it simply means low Q is a very in efficient device. (or this case we want Q to drop like a rock at Freq. 100MHz.) no small task that. and at the PLATE (anode) for rapid gain reduction.

The below sure is cute, but for the life of me can not model the long ribbon from 1.8MHz to 100MHz. the coil part is easy to model but not with SST wire, I gave up trying to model SST .  (yes over my pay grade sure.... blush)
No BOM no SPEC, no help from mean (bill of materials, and wire spec and maker of wire) SST has vast alloys.

SST is made in near endless alloys, some are high permeablity (some super) others supper low and all points between, I see no reason ever to use SST.  (vast alloys) Unless actually named and was not.
Pure  nickel is 100 -600 Rel. perm. Ur. Why not use this,? so we can do effective calculations on 500 Rel. Perm.
A common nichrome alloy is 80% nickel and 20% chromium, (and unknown Permeability)
The stated goal below is HF high Q and at 100MHZ lowQ.  I think the below is 100%  done with only  Empirical testing. (and impossible to quantify or qualify other than a snap shot photo and simple tests.)
AG6k?  Wish I had real mech. drawing of this,... or could buy one. (materials are easy for sure) (and I lack bravery)
 
Odder fixes?:
(as seen by google searching the www)

  SST wire ? 
 this 2mm OD Nickel in the wire has 400-600+ times more permeability. (I see no computer models that can do this.) I see this drawing.
Nichrome wire is mostly used to make resistance type room space heaters or moms toaster. (not RF parts but who knows?)
NO Nichrome wire told maker and p/n# giving for vast choices and alloys. (my guess is to force direct sales) (top secret alloys)

More VOODOO?

more unknown parts.

Next brass !   Getting down to Brass tacks?  Wins for sure that ART award.


r. Brass strap is same as copper /w low permeability (Ur = near 1) {no calculators cover a 1/2 loop}
 
 

Next up the PFC resistor Rx1 or Rx2. Carbon comps are not made now, only rare and NOS parts found if at all. (I use  MOX)
Resistor data sheets cover newer class resistors to use here. (or buy no longer made CC carbon comps from scalpers at $6 each) the cure is MOX.
 
 The OY series, (20kv rated) and or the MOX series,  now replaces the old crap carbon comp's (infamous for failing and aging poorly)
 OY series show no traces of resonance but....(and used on Ameritron PA) keep leads short ! yes the end caps are metal, and make tiny inductance   (RCL effects matter.


The factory used 2x watt resistor for 4 watts total and 22 ohms,
Lx1/2 is not told, guess at 22uF.
The silver on the Lx coils do not render them useless at 100MHz  the skin effect is very tiny at 100MHz using any computer simulation you want, silver at UHF very effective, (and off topic that )
If you did see "Parasitic Oscillation" why not run the PA with one tube. (then switch to see if it is tube related or even find a real Amperex/Eimac tube and try that.?)

PLC,  Parasitic L&C:
- Kenwoods PA, V1/2. 22.5 ohms is PC1/2 Rx that, and Lx is UNKNOWN.  20nH?
 
This drawing can be heard to find but here it is now.  Not the low inductance double grid pins.

Link  below are for Inductor calculators:

By: Serge Y. Stroobandt, ON4AA (THANK YOU !)  one of the best sites on topic of RF coil making. (there are 8+ variables so...)

Mine looks like this, and is kenwood stock and if yours is missing , wow.
The stock coil here is a huge  heat sink for the plate cap so it will NOT OVERHEAT,  vast stock TL-922 Run perfectly even mine now.
I am told  a GDO , grid dip oscillator can measure this inductance or self resonance. (z / freq plot?)  and the Q reads (loosely coupled GDO) of 5.5 below. (better is Q=2 we are told at 100MHz.)
I suspect mine works so good now is the 2 Amperex tubes 1985, and Mu = 100 not 240 as seen today on CLONE TUBES.  (now who makes clones)?
Factory TL922, coil 2-1/2 turns of copper flat bar, the silver does not matter ! here!, at 100Mhz. 22ohm CC=carbon Comp. resistor inside, silver soldered down, Do not use cheap lead solder.


So
me are 3 turns other 4 which is best, IDK. I guess seen from 1935 on.   RX = 50 ohms ( 100/2) (this wrong for 922, use 2x 47ohms)


We see many materials used here, copper , SST, Nichrome heater wire, and silver over copper.  The ferrious matherals have huge permiablties.  (SST is vast alloys and unknown exactly what)
I am not saying above what I know or have done only showing odd things others have done (the magic of RF) Parasitics
I will never be here if my 922 blows up back to SB200 I go. (I flat refuse to buy China tubes via direct or from 3 middle men all making fast buck) as time goes on all these parts will be gone, both sellers and buyers (SK)

PFC Resistors choices, many:
MOX resistors are non inductive. (metal oxide)



First the Bad.
  (even buy one and scrape off the coating and LOOK for lazer cuts forming a spiral inside.)
"Metal film resistors are (can be) trimmed to value by having a spiral track cut through the resistive film" - they can be extremely inductive!   The dirty dogs say non inductive but when pressed, oops at audio freqencies) wow.
Better is to use MOX resistors. and not laser trimmed in a spiral. use 10% or 20% values to avoid precision and those so wrong laser cuts.
One metal film rated for HighFreq use is Vishay  CCF0247R0JKE36 at mouser, look up all value for CCF02 for what you need.
The Metal Oxide (MOX) resistors are the better class to buy.

Many resistors sold are listed as non inductive but is a lie.  (XL = 2πFL )   Non Inductive at audio or at RF 100MHz is not the same thing.
Most makers of resistors are not even aware that HF exists now or just  do not care now as in. ( what's a radio?)
In a computer the only radio inside is a WiFi  chip the size of a babies finger nail (pinky). (2 to 5 GHz parts)


SOLDER ISSUES (and heat) (for  the stock Kenwood PFC, copper plate cap. reworked)
Some of these parts get super hot. and some need high temp solder (vast alloys of solder are sold now, complex) but some need silver solder, the plate (cap) seal is 250C max rated and hotter than most common lead free solders rated low temp are useless here .
Some solders go very weak and sag at 250C,
A nice chart here shows what you can buy. (1.5% silver, may be best on plate caps)

60/40 Sn-Pb, which melts at 188 C (370 F) (Sn = tin, Pb means lead)  one of the most common solder used for 50+ years, but here is Kester list that is very good and maker famous longer.

The winner here is Sn5Pb93.5Ag1.5  (Sn = tin,Pb = lead, Ag = Silver ) top of Kester list. 301C rated. if rich get the 2.5% Ag.
 Lead is SAFE not licked and used/worked at, with a simple fan on the bench. (no need for commercial EPA vent hood)
Always wash your hands before eating after playing with solder.(lead)

Silver (Ag) solders  for jewelry can be crazy expensive. (AVOID THIS)

Do not use acid core solders ever.(in electronic work)
 


The Flashover mod:, by Mr, Kenwood. (solved)   I do in fact use 2  cures here, one is direct cathode protections with 230v GDT (x3) and GDT1 below.
I do have  protector in my ALC too but is an external box ALC buffer show on my page.


This is a worthwhile  mod. to protect T2 from flashover.  (kinda  like seen here but for SB200 reasons and is paired up. "parallel")
R7 is named 2 times, one in the meter box and the R7 bias resistor near, D20 and worse  the R7 shows 100v (100k is true) on the schematic,  R7 is the BIAS resistor.  (this practice is common as they use same names on many modules)
The 100vdc (118v on mine no key) you never want this to go down the KEY line to any SS RIG  of any kind (most).  This is what I call the Virtual GRID BIOS voltage. (and IS via cathode bias)
R7 is cathode BIAS that becomes grid bias.  The keyed up BIOS is 8.5vdc, cathod to grid. The extra voltage  of 1volt is via the meter grounding resistor 0.68 ohms.
D2  btw is 10watt stud mounted zener.  7.5v zener. 1S265 ( Nippon #) Toshiba , common maker.
 I use GDT1 and  TVS device in parallel both.( and only costs $2 to add)
I do not use the X cut,  I added 2 parts, non subtracted.   TVS and GDT in parallel and wired from ground to D1 cathode.
BUY: TVS HERE.  BY GDT HERE.
I did not use the X cut wire hack, the reason they do that is to protect T2 from flash over,  but clamps work too as in the GDT +TVS paired in parallel. The clamps do not change the BIAS (so wins in my book) Do not change the BIAS, leave it alone.
T2 blow up mod1:    This is all to protect the Transformer from damage if tubes go gassy and flash over.
 
This paired devices used with the plate glitch resistor, will afford even more protection . but I do trust my cathode GDTs too.   (non off these parts existed in 1979 (mods) In the navy we had them in 1970 but cost $5000 each. yah  then.
In the old days , protections were  non or nill, and today we do all that, we also prevent cascade failures,  that is what  I do here, prevent #1 not blowing up that $1k transceiver and then try to protect T2.



Next up A.C. line current Surge testing. (100amps possible here)

Line current Surge testing using slow mag donut probe below 
My photos of TL922 current using the CT013-000 blue device below is  (100amp, 50mA out) sensor device.   (the performance plots are here)  the device
PA in operate , CW mode, turned on. After about 1 full second current is 2.2amps on 120vac line RMS current 60Hz.  The surge is same in CW and SSB at turn on. 50 amps.
This sensor has bandwidth of 1000 Hz.CT013  This old slow sensor show this, on scope. 25amp hit first try.
Repeating this over and over I can get 15 amps to 50amps easy or even more

The below is with PA turned off for 12 hours,  and I captured worst case cap charge current on the AC line. CT013-000 sensor below. 10MV/amp.  (this hit  is rare)

If I do the turn on PA with no line cord connected then play nasty tricks with the line cord plug , I can in fact get a 150amp surge. (here)( to do that 20 more times to get the FLex Flux pulse, I  go chicken.) and stop here.
Next below is my DIY hall sensor box, no more crude donut  mag sensor  below is 100% HALL. (2% accurate)
     20mV /amp  scaled.



 
back to mAG:
  CT013-000 sensor below. What does line current do, in SSB mode. yelling in to the MIC.  this.  (10mV per AMP)

Transmitting on to huge 1500w watt dummy load, and tuned for SSB 40meters, 120vac line 60Hz, 30amp RV class service, (super powerful dedicated AC power feed, to the PA. Input is 100watt from ICOM, excited.
My power feed 120vac  30amp, service ,  NO voltage sag here. (as seen on cheap 15amp power feeds in some homes)

End mag sensor start Mr. Hall.


Goals (this page and rig)
Get the full RL1/2 relay spec. (coil V and Ma)  118vDC mine unkeyed, and at 35mA keyed, SOLVED: I'd say the 2 relay coils are (mine stamped on side as 100vdc) 3000 Ohms in parallel or 6000ohms each. Done.
Gen up my new RL_cont,  Soft key, schematic (done)
Get best RS1 thermistor P/N for this PA  only.  (soon) 5ohms would work. SL32-5R020 (done)
Lots of scope photos of surge (mine) and RF power out and a nice AC current scope view with me doing max SSB modulation. done.
Build and test my 2 external protector boxes,  SOFT KEY and ALC.   SOFTKEY DONE,  ALC  done.

Dutycyle,  this is no commerical amp, or broadcast class,
TL922:
SSB. continuous for 30 minutes (FogHorn-leghorn is in trouble)
CWR and TTY. key-down continuous for 10 minutes

The Heathkit SB220(read about heat here and black paint cures)
SSB: continuous voice modulation.
CW: Continuous (maximum key-down 10 minutes).
RTTY: 50% (maximum transmit time 10 minutes).


Conclusion and status:
I only added the in-rush device last,  after testing it outside of the PC first.  SL32-5R020
and added  my flashover clamp  pair here.
That is all I did inside the chassis of PC.


Overheating, avoidance.
  • Keep the AMP bottom side on to say  a grill stand to allow air flow to the bottom. Get as much cool air to the bottom side.
  • The fan working right and grill behind  not packed in lent.
  • even paint the (mods) HV cage inside walls in flat black paint,  really helps any PA,  the heat is IR, (infraRed) energy and best for sure to land on black surfaces, heating the case is good first , then air flow gets the rest of the heat.
  • Do not block fan exit to a say a wall. (or other large object)


Transconductance terms.
Gm =  ΔIp /  ΔVg   (say the change in plate current divided by the change it Gate voltage. ) MATH Gm is now called (SI units) Siemens. 
Mu  = ΔVp / ΔVg  
Mu = Gm x Rp



rev 8   ++++ 12-25-2019 (day1 +)  This website K5JXH.com  is not affiliated with or sponsored by Kenwood or any other named device or named product. (we do not track you ever) 
Do  not work on any PA if you lack HIGH VOLTAGE TRAINING OR FOR SURE RESPECT OF IT.  (If unsure? ask for help or get your rig fixed by a real tech, that knows HV power supplies or join a HAM club and find a HV Elmer.)

E mail ,   k5jxh  (at)  arrl dot net (you know the drill)  73 to all.

Answers to questions from other Hams.
From :W1Gee.
  1. What thermister RS1 did I use (add) for surge (inrush) protection, answer best is: SL32-5R020
  2.  (not asked but..) where did  it put the SL32 , answer is  here.
  3. Where is the GDT1 go and how is it wired up, here, and is 2 devices paired !
  4. What kinds of  transformer protection is there?, The GDT1 is  one, and are there more? (sure!) The big danger is gas flashover, and to T2, I used cathode protection on my SB200,  but for Mr, Kenwood.  I used 230v rated GDTs. See last below."notes:"
  5. What did Kenwood use to protect T2? , answer (hope and prayers) They simply hope gas never happens.T2 does not have 5kv insulation inside, like T1 has, and in 1978 there were no fancy dancy GDTs nor TVS made then, these GDT parts are MODERN now.
  • There are many things on board that will blow up at 3000vdc on the whole cathode side. But T2 is not to be a sacrificial lamb , like L3 is and all those LV Caps and RL2 and the 2 DIAL meters blown up, (the  meters are upgraded with clamp diodes)
  • (most caps in here are 1kv rated all can blow up with gas. ( surge protector in the plate feed path are also popular)
  • The other protector is at ALC jack If C40 blows up (shorts over 1kv) BOOM you just sent 3kv out the ALC jack. This never pans out well ,oops. (the 68k R10 resistor will blow up then arc over and I used a GDT clamp on the ALC jack pin.)
  • If ( IF's matter) C42 shorts, the band switches and all that blow up, so cathode clamps really are better in this regard !  The normal highest Cathode voltages are BIAS   The cathode is only at 7.5vdc keyed up and RF drive is 100v peak or tad less from exiter, I use 75% drive from ICOM)
  • I think the cathodeGDT devices are the best  fix.

Notes:  (I say use 230v GDT here) x1 one for each cathode pin1&5  I use Bourns 2073-23BLF ( long leads 0.8mm not bent) axial type.  on at least both cathodes V1 pin 1 and V2 Pin5.