[Amps] GS-35B Efficiency Varies with frequency?

[Jim Thomson]

Date: Mon, 10 Dec 2012 16:50:42 -0800
From: Paul Baldock <pbaldock at frontier.com>
To: amps at contesting.com
Subject: [Amps] GS-35B Efficiency Varies with frequency?

I am building a GS-35B amplifier. The efficiency drops off a lot 
between 80M and 10M. Although the gain is quite constant. Does any 
body have suggestions as to why this is or whether this is typical.

The input PI network has a Q of around 2. The PI output has a Q 
between 10 and 20

Here's some results with 70W Drive:

3.80MHz. Q=10, Output Power 1.1KW, Vp 3000V, Ip .54A, Ig .26A, Eff 68%
7.10MHz. Q=13, Output Power 1.1KW, Vp 3000V, Ip .60A, Ig .28A, Eff 61%
10.1MHz, Q=20, Output Power 1.0KW, Vp 2900V, Ip .68A, Ig .30A, Eff 51%
14.2MHz, Q=13, Output Power 1.1KW, Vp 2900V, Ip .70A, Ig .29A, Eff 54%
18.1MHz, Q=16, Output Power 1.0KW, Vp 2850V, Ip .78A, Ig .31A, Eff 45%
21.2MHz, Q=13, Output Power 1.0KW, Vp 2800V, Ip .82A, Ig .37A, Eff 43%
24.9MHz, Q=19, Output Power 1.0KW, Vp 2850V, Ip .79A, Ig .35A, Eff 44%
28.5MHz, Q=16, Output Power 1.1KW, Vp 2750V, Ip .85A, Ig .38A, Eff 47%

You will note as efficiency goes down the Grid current goes up (in 
general) for the same drive. Is this normal?

Things I have tried:

1. Checking that the bias voltage is not affected by RF. No signal 
Vp=3600V, Ip=140mA, Vb=34.5V.
2. Rebuilding 10M PI Output in a different configuration. (It is 
actually an L-PI)
3. Checking that the metering is not affected by RF

Any Suggestions?

What is a safe Max Ig for the GS-35B?

Thanks

- Paul KW7Y

### what are you using for a wattmeter, have u tried the same wattmeter on another hf amp ?
What does the tuned input consist of, separate pi nets, bandswitched, or relay switched ? 
How close do you have the tuned inputs to the cathode ?  Ideally, you want them right next to the cathode.
I  forget what the input Z of that tube is.  What are you using to connect  from tuned input to cathode,coax or strap ?
One method to increase eff on the upper bands is to use parallel 50 ohm coax to make 25 ohm coax... and use the
25 ohm coax from tuned input to cathode.  That has the same effect as putting more cathode to chassis C in parallel
with the tube’s cathode to chassis C.  ( you can also parallel some RG-59  to nake 35 ohm coax.) 

##  You can also try adding a fixed C, like 50-200pf,  from cathode to chassis. 
Try increasing the tuned input Q a bit, try this on 15M, which has the lowest eff in your case.  

##  If you use a T net for the tuned input, eff will suffer, and  you will have a hard time getting a flat input swr.
That entire eff issue revolves around the tuned input config.   I gave up on individual pi nets for each band...
and instead opted for a pair of broadcast variables and a tapped coil.  4 x section types, typ 17-540 pf per section,
with all 4 x sections strapped in parallel with cu strap.  No wire used anywhere on the tuned input bandswitch,
nor the taps to the coil.   I used 6 ga solid cu wire for the tuned input coil.  .187”  tubing or .25” tubing will also work.

##  a  parallel tuned input is not much better than a T input.  Unless you have some way of tapping the coil on the cathode
side, you will find that min input swr and max grid current do not coincide.  

##  Your loaded tank Q is way too high on every band except 80m.. ( which has the highest eff).  ( output kw pi net )   

##  Loaded Q’s  of 8-10 provide for WAY better eff...and slightly less harmonic redux.

##  You not only need an L-PI... you also need a vac cap for the tuned input.   Then you can easily get a lower,loaded
Q.   Check the rotor contacts if an air variable used for the load cap, they can easily heat up. Also, you don’t want the
rotor grounded in several places, only once.  Use a shaft isolator on the shaft, and float the front end of the load cap
frame.  Only ground the back end of the frame.  Ideally, a vac cap should be used for the load cap as well. 

##  what are you using for tank coil material ?   I’d suggest 8 ga solid wire or .125”  tubing for 80+40+30m. 
And .25 -  .375”  tubing for 20+17+15m..     And .75 – 1.0” wide strap for 12+10m. 

##  How big are the bandswitch contacts ?   Beware, the contact ratings that  mult-tech industries use are dc-60 hz ratings
not RF ratings.   If you have spot heating  going on anywhere, you can usually find it with a fluke 62, point and shoot IR gun. 

##  do u have another tube to try out ?   I think the gs35b has one side of the cathode tied to the fil, and I forget which side. 

##  what are you using from load cap to TR relay ?    Is it a piece of 50 ohm coax...or is it wire/strap ?   You might be introducing
a huge Z bump in there ?    What is your bypass  SWR....esp the upper bands ?  What are you using for a TR relay setup. 

##  You could also try increasing the idle current, like 180-250 ma.   That gs35b is capable of well over 60% eff on 6m. 
I suspect the eff problem is in the tuned input /  and or the pi output, or wiring in both pi nets. 

##  IF your tuned inputs are setup for 50 ohms,  you might want to try measuring the power output of the tuned inputs
with a wattmeter, then into a 50 ohm dummy load.  But if your 100 w dropped way off to say 60 w, then your grid current
would also be way down on the upper bands, and its not, it’s higher. 

##  any bad switch contacts or rotor contacts  will generate a lot of heat.  If input Q is too low, and a SS xcvr used, eff will suffer.
If input Q is too high, losses in the tuned input will increase.   What size coupling cap are you using between tuned input and cathode ?
2-4  x  .01 uf @ 1-2 kv disc ceramics in parallel will handle the drive power.

Let us know what you fine out. 

Later ... Jim   VE7RF