Settings for homebrew amplifier?

[WB3JOK]

After two years I finally got around to smoke-testing my mostly-junkbox amplifier. One feedthrough cap started leaking which gave alarming screen current readings (fortunately false ones). Also the first time I drove it to full power there was a zorch and the plate/screen voltages disappeared, but it was the Variac, not the power supply 

Anyhow I was looking for some advice on bias settings, drive levels, plate voltage etc.

The circuit is four paralleled 4-125A's to run in AB1 or AB2, 4:1 balun on the input with a 200 ohm thick-film swamping resistor heat sinked to the chassis. So it should be able to input 25W or more, absorbed in the swamping R. Swing at the grid at 25W will then be 200V p-p across the 200 ohms. Output is the typical pi-net except the tank coil is made of three stacked T200 toroids. No bandswitch, I'm only going to use it on 75m and 40m, but there is ample room to add one later if I need to.

Plate supply is 3.5 kv but can be adjusted downward with a Variac if necessary. Screens are fixed at 425V with a 4-65A shunt regulator tube and zener stack. When in standby mode -125V is applied to the grids to cut off the tubes. Grid bias is adjustable separately for each tube. Depending on the bias setting there may be some positive swing (and hence grid current) at the top of the drive cycle (+100V).

So - assuming it's OK to run the full 3.5 kv on the plates, and drive it with 25W or less, which is about what my TCS and Command Set transmitters will produce, how much (total) idling current should there be with no drive? Right now it's only 60 ma total, which is 200W of heat, or 50W per tube, so certainly it can be set higher. Driving it with my HW-22 (normal uncompressed speech) kicks the plate meter up to 200 ma or so and the plates flicker a dull orange with prolonged sounds.
Looks OK on the scope but I have no idea what the spectrum looks like and don't want to splatter anyone... at least SSB keeps the duty cycle low, so I haven't tried it on AM yet.

Edit: I just revisited the 4-125A data sheet and it looks like 200V grid swing will be too much. Even at max plate voltage the tube is cut off completely with about -90V on the grid, maybe -100V at the peak of the plate swing. Without actually drawing a load line, from the graphs I estimate -65V bias will allow 40 ma per tube which is the limit for static plate dissipation at 3+ kv. So the max swing has to be limited to about 70V p-p (-100V to cutoff, -65 bias, -30V peak at which each tube is pulling close to 100 ma). At this grid voltage (25 Vrms) the driving power across 200 ohms is just 3 watts! I may have made it too sensitive... maybe get rid of the balun and use a 50 ohm swamping resistor. That still will only take 12 watts for 70V p-p...



I like the look of firebottles in the dark 


-Charles

[WA1GFZ]

Charles,
You would be much better off mounting the tube sockets under the chassis deck and have all the grid and cathode circuits inside the chassis isolated from the output circuit. Also you could put a small blower to pressurize the chassis and blow some air through the sockets.
Check the flux density of those cores you could be saturating them at high voltage. The cores should not get warm or the L is going to change as they heat up. Air core inductors are the best.
The lead length between the plate tuning cap and tube plates is very critical for stability
the shorter the better

Down load the care and feeding of power grid tubes and read it.
Nice amp

[WB3JOK]

Thanks for the tips. I had forgotten how useful the Care & Feeding manual is... I already have a copy downloaded and am now studying it again.

Looks like I want 50-70% of max plate dissipation at no load, so at 3 KV that's about 25 ma per tube, or 100 ma total. I may need to drop the screen voltage further in order to allow more grid swing for a given power output since I really don't need 3 watt sensitivity for full output  Adjustment is a simple matter of jumpering one or more of the zeners in the shunt regulator circuit.

The tubes are on top of the chassis because the 54 uf filter capacitor, rectifier strings and the various relays, metering and bias circuits are underneath, so only the 38 lb. plate transformer is "outboard". I figured at 3.5/7 MHz the stray coupling might not be that critical especially with heavy grid swamping - but I can mount on metal standoffs a sheet of copper, brass or copper-clad PC board material with four large holes in it for the tube bases if isolation is an issue, and still allow the fan to blow on the pins as well as the envelopes.

I"ll check the tank coil for temperature but three stacked T200 cores is a pretty hefty mass of ferrite which shouldn't saturate. I'm expecting at most 1000W PEP output...

The four tube plates are connected (via parasitic suppressors) to a 1" square copper sheet atop the plate choke so that is as low inductance as possible. But from there I have about 6" of 1/8" wide silver plated ribbon to the tank circuit. Should I use even a lower inductance connection than this? Or maybe put some ferrite beads over the ribbon?

-Charles

[W2VW]

    Maybe look into resistance in series with the cathode (filament center tap) return to ground. This would desensitize the amplifier due to inverse feedback. Allows for use of more drive instead of blowing the thing away.
    Another way would be to install a bifilar filament choke and apply drive through a capacitor. 

[flintstone mop]

NICESH 5 glowing 4-400 PUBES 
Really nice amplifier. But follow the advice of the others for the isolation and shielding of the 400's and some air across them would help.
I had a 50's style TX (a pair of 400's push-pull.....I tuned for 500 watts...ooops) the tubes were not recessed in the chassis or any air blowing and it ran without any problems.
Fred

[WB3JOK]

NICESH 5 glowing 4-400 PUBES
Really nice amplifier.

Actually it's four 4-125A's and the fifth tube is a 4-65A shunt-regulating the screens! They were much cheaper than a single 4-400 or 4-1000. The intent was just to add some "strap" to my PW 10-watt screen-modulated ARC-5 transmitters so I can join the fun on 75 & 40  In AM linear service, at a max 500W plate dissipation, that should allow at least 250W carrier output (at 33% efficiency). That's a nice 14 db gain if my calculator is working...

But thanks for the compliment anyway 

It's not obvious in the picture but there is a noisy 6" muffin fan (on the far side between the filament transformers and the tubes) which moves a pretty decent breeze across the envelopes (and the pins, another reason I mounted the sockets above the deck). Eimac says that cooling air is mandatory to keep the seal temp below 200C...

Another way would be to install a bifilar filament choke and apply drive through a capacitor. 

Sure, I could run it grounded-grid instead, but I'm not sure there would be any advantage (and it would then, as you point out, require a hefty filament choke).  A tetrode amp should have sufficiently low C_gp with the screens at RF-ground to be stable at 40m without having to ground the control grids too... if not I'll add a bridge neutralizing circuit. The air variable is already mounted on the chassis. Any suggestions always welcome.

Today I set the bias at 100 ma with 3 kv on the plates - the simplest way was to turn all four bias trimmers to cutoff and then one at a time adjusted them to increase the current in 25 ma steps. The plates show a little color at 75W per tube - is this normal? This thing scares me a bit, especially using a diddle stick within arm's length of a 3 kv power supply 

-Charles

[W2VW]

Sounds like it's working OK.
You need not ground the grid to drive the cathode. Leave all that stuff right in place. Just move the drive.
The reason I suggest this is you have a lot of extra drive available and it can be put to use with a cleaner signal than grid drive.
Did you leave enough room to plug in 4 4-440s?

[WA1GFZ]

4-400s need air flow through the sockets.

[W2VW]

4-400s need air flow through the sockets.

Same for 4-125.

[WB3JOK]

4-400s need air flow through the sockets.

Same for 4-125.

Not according to page 3 of the Eimac 4-125A data sheet. Although it does say that "free circulation" through the tube base is essential, it also says that the tubes should be mounted on standoffs above the chassis to allow this  They further state that if air is not able to flow through the socket and base, THEN forced air is required to the "stem structure" (pins/seal). Also note that below 30 MHz "when the ambient temperature does not exceed 30C (86F) it will not ordinarily be necessary to provide forced cooling..." (as long as the finned plate cap is used.

Of course MORE cooling than necessary will never hurt! But it appears that I would be within the manufacturer's recommendations at 125W per tube or below even without the fan as long as it's not too hot in the shack. 

Now the 4-400, on the other hand, has a 5.0 volt filament that draws a whopping 14.5 amps and requires a blower (even with only filament voltage applied) to keep the seal temps within limits... (the 4-125A filament is also 5.0 volts but draws only 6.5 amps). Yet another reason to use the plentiful and cheap '125's 

-Charles

[W2VW]

As long as you are satisfied that you will get your 2 CFM without any air movers then you would be correct. How many spares do you have on hand?

[WA1GFZ]

I don't think eimac was talking about RF applications when they saidt o mount the tubes on stand offs. A little air flow will help the tubes live a long life even if only while transmitting.
Ambient temperature will easily hit 86 degrees around the tube.

[WU2D]

Hi Charles,

Wow, what a great junkbox project! With everything above chassis, there will be significant feedback from plate tank to grid and the amplifier may be unstable. Some grounded Grid amplifiers need to be neutralized anyway. This one is a sure candidate. Their are several methods.

Tom, W8JI's website has some information on handling this if required.   

Mike WU2D

[W3SLK]

Mike said:

Tom, W8JI's website has some information on handling this if required.

Unfortunately, that's a waste of good bandwidth. I wouldn't believe anything that A-H says even if he swore it came from the 'burning bush!'

[W2VW]

JI is a very talented guy. He is right on the money 99% of the time. It's that 1% that can be troublesome. That's better than most. The same qualities that make someone good at these things are those that can potentially lead one down a wrong turn once in a while.

[WB3JOK]

Wow, what a great junkbox project!

I cheated a little... my junkbox didn't have all the parts at the beginning, but the Hamvention fleamarketeers helped a lot
The second most expensive part after the pi-net loading (vacuum) cap was the 17x17x3 BUD chassis! I bought it new, for something like $50, two years ago! I don't remember them being that big a percentage of the project cost 30 years ago 

With everything above chassis, there will be significant feedback from plate tank to grid and the amplifier may be unstable. Some grounded Grid amplifiers need to be neutralized anyway. This one is a sure candidate. Their are several methods.

You make a good point, all things to watch out for, except... 1) it's not a grounded grid amp, 2) the tank coil is a toroid which radiates very little, 3) the grid leads are almost entirely below chassis except for very small vertical portions which are at right angles to the plate lines to minimize coupling, 4) the grids are heavily swamped with a 200 ohm non-inductive load resistor, and finally 5) the screen grids of tetrodes reduce the Cgp to a trivial value, at least at 7 MHz where I'm operating it.

My researching shows that at HF, neutralization probably will not be needed in this circuit, and it is perfectly stable so far. I agree that it's better to have all the grid "stuff" below chassis but that wasn't practical with my layout. If I see even a hint of instability then I'll mount that above deck too, as I mentioned in a previous post. I can twirl the plate and load caps all I want, and run the input drive up and down, and nothing so much as flickers... thanks for the warning though.

Tom, W8JI's website has some information on handling this if required.

I have read his neutralizing info. I think it's a useful site. Some may disagree

Progress on the bench tonight. I hooked up my 75m Command Set (with homebrew screen modulator) as a driver. With no modulation using a 10x probe on my scope, I measured 60v p-p at the 50 ohm dummy load which is a whopping 9 watts carrier. Then I keyed up the amp at 75% plate volts. At first the 4-125A plates would get more orange and it didn't want to load. I couldn't get more than 50W out with 2.5 kv/250ma (625W) going in. 

I finally figured out my pi-net loading cap (1000 pf max) was grossly undervalued, at least for 75m. So much for my ARRL Handbook calculations... anyway I kept adding shunt capacitance until I could tune through a broad peak with the variable cap. The additional cap that worked was 2500 pf! I had a .0025 6 KV disc cap from my Tek scopes spares which worked but it got rather warm. The Handbook claims that they are big enough for this power level but it appears marginal... A transmitting mica cap (about 1"x1.5") of 2500 pf worked as well without heating palpably.

I'm probably going to need a bandswitch after all since I'll bet that's too much capacitance for 40m - I was hoping just to retune the plate and load caps without switching anything. Oh well, at least there's plenty of room for a switch, but there isn't one in the junkbox that can handle the required power level.

With the amp properly loaded, I raised the plate supply to 3 kv (3.5 kv no-load, my "big-iron" plate transformer especially driven by a small Variac apparently isn't quite manly enough. It's an old Thordarson and I think it's rated for 500VA). Without readjusting the bias pots, the idling current rose slightly to 120 ma and the plate dip was 240 ma with the full 9 watts of drive. At that point I could smell the dummy load getting hot (a hollow 100W noninductive resistor sitting in the exhaust air from the fan) and sure enough I had 300v p-p on the same scope. That works out to 225W carrier output with 720W dc input, about what I've been led to expect from an AM linear, and the tubes are not overloaded (125W each). Only a slight red-orange color to the plates key-down.

I can probably get better efficiency and power output with further adjustment of the bias. Should be a lot of fun on SSB too. I've ordered some thick-film power resistors to make an 8 db pad so I can run my monobanders into it and stay around 20W PEP at the amp input.

However, there are a few problems. I blew another screen feedthrough cap when I accidentally turned off the filaments before the plate voltage. (Since the screen supply is derived from the plate transformer center-tap at 1/2 the HV, when the filaments are all off including that of the shunt regulator tube, the screen supply can rise momentarily to half the plate voltage, and that was over the 1000V the feedthrus are rated for!)

Also the plate tuning cap tends to "zorch" when the BC-696 is first keyed up (a design anomaly of my simple screen mod circuit that initially puts out 100% carrier before quickly settling to 50%). The plate DC is already blocked from the cap, too. At high PEP levels it'll probably arc over on voice peaks. Also I noted that the toroid stack gets quite warm, I guesstimate 150 degrees, with prolonged key-down. I may need to redo it with four cores instead of three. Or I can just stop trying to stuff 10 lbs of sh!t in a 5 lb bag and be happy with a "half-gallon".

-Charles

[WA1GFZ]

Those loading padders will drift in value as they heat up. broadcash mica will be the best.
Even door knobs will drift unless you parallel a lot of low value units.

[WU2D]

Charles - I'm a Command set nut. I have run several stock units and have a couple of nice ones. I have also hyper-modded a few "holey" ones. My latest is a 20M conversion done on a completely stripped out 4 - 5.3 MHz chassis that a ham left for dead.

I removed the tertiary fil winding on the VFO coil and wired in a 12SK7 ECO in a conventional Hartley using the existing components - less heat and mucho stable. Several plates were removed from the main tuning cap so the the bandspread is 1 MHz. The front cap was disconnected and removed because I used a powdered iron doubler low-Q tank which was wide enough for the whole band.

The doubler is a 12A6 in the Magic eye Socket. The center socket takes the VR-150 which is on the VFO screen.

The roller is removed and in its place a full Pi network is installed using the ceramic form (sans link assembly). BTW I have also used the Roller as the Pi coil with great success on my ARC-5 AM Mobile Station. A plate covers the window with a TX and spot switches, and tune and load knobs.

With some key click filtering and neutralization, the results are a full output 20M Command set with a great CW note. I ran it on Straight Key Night and made a few Q's.

Mike WU2D

[W3RSW]

Oh man !  Whether it works well or not, it sure is purrrty, what with all those glowin' tubes.

[WB3JOK]

Oh man !  Whether it works well or not, it sure is purrrty, what with all those glowin' tubes.

Still looks purrrty but not working so good right now 

My thick-film resistors arrived from Mouser so I installed the T-pad (20 ohm series arms, 50 ohm shunt). This is 7.7 db power attenuation. Hooked it up to the HW-12... made a contact with a guy who kept telling me I was "transmitting and receiving on different frequencies", he's "not into building stuff and doesn't like those old radios", and finally "no difference in my S-meter" (with the amp on or off. Scope with cliplead antenna showed a big difference on my end). Gave up on 75m since everyone else was too busy belching and complaining about carriers  and tried dodging foreign broadcasters on 40m. First contact was so weak he couldn't copy me well enough with or without the amp. Finally found Neal W1FT calling CQ from north of Boston who was S4 which should be a perfect amp/no amp test ... but right then the damn amplifier stopped working  So we had a nice chat without the amp.

Anyhow, the idling current is the same (100 ma), screen (425v) and plate (3 kv idling) voltages are unchanged, relays are making contact, RF at the grid is swinging - but there's actually less power coming out of the amp than is going in! Measuring the RF directly at the pi-net loading cap confirms it. Can't tune through a shallow peak with the load cap, either. I suspect the old mica loading caps have given up and are leaking severely. Doubt the tubes could be fried with all the voltages and currents looking ok, and with the attenuator there isn't enough power reaching the grids to melt them. I'm not sure what is going on but it's late and I'm tired, not a good time to be playing with 3.5 kv plate supplies. And of course I have to leave town for four weeks starting tomorrow. Oh well, at least it's only a hobby. I bought some doorknob caps on Ebay and have to wait for them anyway...

-Charles

[W2VW]

Fast toob test is to look at max plate current with drive and the plate tuning severely mis-tuned. Obviously test for only as long as it takes to measure. Record value for later toob comparisons.
Some of those old micas sometimes have pretty low R.F. current capabilities.
Your results could have been from plate circuit tuned to a harmonic too. That would account for puked loading padders.
Good fun!

[WB3JOK]

Your results could have been from plate circuit tuned to a harmonic too. That would account for puked loading padders.

Mistuning probably is not it, since I actually was able to QSO on 3.898 with the amp inline or out. And at 40m the plate tuning cap is near minimum (with the Cp of four paralleled tubes + stray + fixed tank coil I can't get a small enough capacitance to tune the tank to 20m). But I'll be sure to crank up the scope sweep speed and freq counter, and double (pun intended) check.

I wonder if there is enough circulating current at 40m to hurt an old mica cap that did fine at 75m? Since the L is fixed the Q has to be different by a factor of 2, right? So if the I doubles, the I²R losses quadruple... 

Further experiments will have to wait a month. Anyone know of a source for an affordable bandswitch?

-Charles

[W2VW]

I was thinking of the load which might appear to a second harmonic. Probably quite a bit of reflected power.

[WB3JOK]

OK... let's talk bandswitch for a moment.

I have read that it is not a good idea to use a simple tap switch (to select part of the tank coil) since the unterminated "free" end can develop very high voltages. But progressively shorting ceramic high-voltage switches are not inexpensive or very common as surplus.

Would it be possible to use a tap switch with the wiper connected to the loading cap/pi-net output AND the output end of the tank coil, (with the switch contacts connected as usual to different turns)? In other words, an increasing number of turns will be shorted out by the switch as it is rotated. The question is, will there be much voltage appearing at the unterminated taps of the shorted-out section? 

-Charles

[WU2D]

The guys here on AMFORUM are amplifier guys and I am sure that they all have pet tank circuits.

But here is an old idea. The switch shorts out the coil as you go higher in frequency, eliminating your concern. Note the extra loading cap which automatically comes in at the lowest band. For 80M this may need to be 1000 pF or more and it should be a high voltage type.

Mike WU2D