Instability in my dual 807 transmitter - Some progress - I think

[N4LTA]

I finally am getting around to packaging up the dual 807 amp and have run into some instability that is a puzzle.

When I key the transmitter it will cleanly key 90% of the time with a good sine wave but sometimes when keyed it will drop to about 1 watt out (down from 38 watts) and then slowly rise to 38 watts (in 3-4 seconds) and waver a few watts and then settle in to 38 watts again.

Looking at the output with a scope shows just a sine wave that starts at a watta or so  slowly rises to 38 watts. Nothing that looks like high frequency parasitics (the scope is a 400 Mhz scope)

The driver is a solid state driver and it seems to be keying cleanly with no slow rise in power. I tried an unused set of 807s and nothing was different.

Looking at plate voltage shows a fairly steady voltage on the tube plates dropping from 430 volts no load to 375 and then steady. The plate voltage stays steady during the slow rise in power.  

The screen is a different story - It drops to 250 volts at keying when things key properly but when it misbehaves it drops below 230 and then slowly returns to 250 volts with with RF voltage riding on it. When the RF voltage goes away the transmitter stabilizes. The screen current at 250 volts is 3.9 mA which is what it should be when stable.  

Is this a parasitic or something else?

The transmitter is a single band 75 meter transmitter.

Thanks in advance for any insite.


Pat
N4LTA

[w7fox]

Pat,

I have a pair of 1625s modulated by a pair of 807s.   I had lots of challenges with parasitics.  If you don't see them on your 400 meg scope you should be okay.  I used grid block keying, you don't say what you are using.  It sounds like you might have a leaky cap at the grid or screen, or even a bad grid-leak resistor.
Since you have a nice scope, it should be possible to poke around until you find the offending bug.  I became quite fatigued debugging my transmitter, and though it is now done and works great, I think I'll use a different final in my next project. 
Good luck.

Best regards,
Chris

w7fox

[N4LTA]

Yes,

I have heard parasitics are common with these tubes.

The problem is strange in that the scope shows a sine wave that is normal when keying the transmitter 9 out of ten times and then it shows a clean sine wave about 20% of full output voltage that stays synched and then rises to full output over 4-5 seconds and slowly changes maybe 5% in amplitude for 5 seconds and then sometimes drops as if something thermal is going on but only intermittently. The screen voltage get very unstable when this happens.

Pat
N4LTA

[W7TFO]

Hi Pat,

Is the circuit push-pull or parallel?

Parallel pentodes can be self-oscillatory (on occasion) if the grids are not built out with individual low-value resistors.

73DG

[N4LTA]

Two parallel 807s running at low voltage -about 370 volts with a screen dropping resistor. Good solid choke input power supply  - about 38 watts out.

Pat

[KA2DZT]

Is there one screen dropping resistor or one for each tube??  You probably should have at least some of the screen resistance split to each tube.  Something like the last few Kohms.  Make sure you have at least a .001 by-pass cap on each tube soldered at the socket pin.

[W3GMS]

Pat,
If your deriving your screen voltage from the plate voltage through a resistor, make sure the RF bypass cap is good at the base of the RF choke going to the final.  At the base of the RF choke where the bypass cap is, it should be clean from an RF standpoint.  Also with a very high frequency scope like your using, sometimes RF can get on the ground lead of the scope probe and you will see things that in reality are not there.  Try to keep your ground lead very short when taking readings while the transmitter is on.  I am assuming your pair of 807's are biased for class C service.  Does the peak in grid current match the dip of the final when resonated?  I am also assuming that you have good physical isolation between the input circuit and the plate tank network.  Good design practice would be to utilize parasitic suppressors in the grid and plate circuits on the tube.  If you don't have, but could borrow a good spectrum analyzer that would help to see whats going on as well.  Neutralization should not be a big deal on 75M with a screen grid tube but putting a scope on the output with a 50 ohm load, remove the screen and plate voltage twist the grid and tank condenser and see how much RF feed-through your getting. 

Have fun....
Joe, GMS         

[The Slab Bacon]

Do you have any regulation on the screens?? Either you might have slightly gassy tubes, a slight parasitic oscillation, or an instability in your screen voltage. 807s are pretty sensitive to sagging screen voltage, henceforth all of the screen regulator mods for old Viking 1s and 2s 807 modders.

If you are feeding the screens off of dropping resistors from the plate supply, and the plate supply sags a few volts under a load, the screens will sag even more.........

Just my $.02 worth

[N4LTA]

Lots of possibilities to look at and thanks for the ideas. The screen is derived via a resistor - I have parasitic suppression in the control grids , screen grids and the cathode in the form of low value carbon resistors as well as a typical plated parasitic suppressor (100 ohm carbon resistor with a coil in parallel)

Two screen dropping resistors on for each tube plus a 22 ohm carbon at the tube as a suppressor.

Tubes were old so I swapped them with 2 NOS 807Ws and I got a few more watts out , but the same problem.

I didn't regulate the screens in order to make the modulation simple (no screen choke)  The plate voltage is pretty stiff as it is a choke input supply with a large choke and 110uF of filtering. (two 22uF 450 V in series.)

I haven't put in on the SA  - I will do so tonite - I got rid of my homemade power attenuator so I'll have to kluge something up.

I'll also check the feedthrough with no voltage on the screen or plate.

Thanks everyone.

One more question - Since the cathode is 33 ohms above ground (a carbon resistor parasitic suppressor)
Should the screen bypass go from screen grid to ground or from screen grid to cathode - I have seen it done both ways. It is now SG to ground - I'm thinking SG to Cathode is maybe better?

Pat
N4LTA

[W3GMS]

Pat,
I would definitely put a bypass cap from cathode to ground since you have a 33 ohms from cathode to ground.  If you do that you then should be able to put the screen bypass directly to chassis ground.  I generally prefer putting the cathode of tube right to ground or at the most through a very small non inductive resistor to act as a shunt so cathode current can be read.  Even with that small non inductive shunt resistor I always bypass the cathode to ground with a .001 ceramic cap with short leads.   I guess you want some cathode bias, hence the 33 ohms rather than developing bias with a negative grid bias supply. 
Joe, GMS   

[N4LTA]

I have a regulated protective bias supply (-62 volts)  as well as partial self bias with a grid resistor.

The GE Ham notes recommened a low value carbon resistor in the 807 cathode to add some degeneration for stability. That is why it is there. I pretty much did everything the old Ham Notes article said to do to prevent instability because I have had problems with unstable 807s in the past.

The RCA notes say to bias the tube more into cutoff and drive it harder if you have instability. The tubes must have been buggers to design around.

I am beginning to wonder if all my additions to help stability may be causing problems.


Pat
N4LTA

[The Slab Bacon]

Keeping on mind that an 807 is nothing more than a 6L6G with a plate cap, you are dealing with an old design tube designed back in the early '30s.

And.........keeping that in mind, I have a commercial rig at home that uses a pair of them (807s) in a parallel amp at 50 Mhz (52.525 to be exact), so it is doable.

[KC2ZFA]

The first post describes an intermittent behavior of the screen current: some times, on turn-on, the screen draws too much current until it settles to what it should be given the amount of rf drive, grid bias, and plate voltage.

Is there such a thing as a damped grid-screen parasitic with a 3-4 seconds time constant that tends to "eat" rf drive ?

[w7fox]

Pat,

I've been brainstorming what could cause an effect that takes a few seconds and I just thought of something.  Filaments.  Any possibility of loose contacts or cold solder joints?

Best regards,
Chris

[N4LTA]

It happens on drive. The amp is on with power to the plates and screen and biased into cutoff - not pulling current.

The plate has about 425 VDC and the screen near the same as no screen current is drawn nor plate current.

When it is "keyed" the bias voltage rises from cuttoff to -62 volts and the rf driver comes on and the RF drive combines with the grid resistor to provide the correct grid voltage   - if I remember correctly -85 volts

The RF drive is adjustable from .4 watt tto 3 watts.

When the amp is keyed -  9 out of 10 times it delivers a clean sine wave at about 38 watts when the plate current is dipped and the load cap is peaked.

Then on occasion - I get a few volts RF out and after 2-3 seconds the RF out rises slowly to 38 volts but then moves down a bit to 34 watts or so and is unstable. Keying the amp again usually goes back to normal. When the weird stuff is happening the screen voltage  seems to be oscillating but that may be the scope ground lead being to long as mention previously.  Since the screen voltage drops, the plate current has to be rising considerable - but I did not have the meter in circuit. After the plate current reaches the critical value for the choke input filter it is pretty stable - so the plate currint may be going pretty high. I need to get a permanent meter installed.

Pat
N4LTA

[w1vtp]

I'm pretty sure I posted this before.  Got pics of this xmtr?  The author of the "Torpedo Twins dual 807 Final" points out the importance of isolating the input circuit from the output

[W3GMS]

Pat,
I don't think its a voltage regulation problem based on how your describing the behavior.  Lower screen voltage means lower output power and less plate current.  You should be able to vary your plate and screen voltage at least 20% and true the output will change, but it should still be clean.  As mention in an earlier post, you want to make sure that the raw screen voltage is clean and since its derived from the plate voltage at the base of the plate RF choke make sure things are decoupled well at that point and then the source of your screen voltage will be clean.  That together with good screen bypassing and bypassing that 30 ohm resistor in the cathode circuit should get you to a circuit that is RF sound principle wise.  Personally I would not have the 30 ohm resistor between cathode and ground.  I would put the tube well beyond  cutoff with the negative bias supply.  

You can have the best circuit on paper and if its not laid out correct it will be very problematic.  Make sure you have good isolation between your grid tuning network and the plate tuning network.  They should never see each other.  I would look at the feed through power first and if that looks good, then proceed to the other stuff.  
Fun stuff....
Joe, GMS      

[N4LTA]

The grid tuning is under chassis and the plate tuning is on top so isolation should be pretty good - and on oposite sides of the tube.

The "torpedo twins" article is where I got the ideas for the parasitic suppressors.

Got a guitar amp to finish tonite for a friend  before I can get back to the transmitter.  The part I hate the most - mounting the chassis in the cabinet - screw up and you ruin the cabinet - I hate it.

Pat
N4LTA

[w1vtp]

Pictures might help in the case of a layout question.  Sounds like you are doing the right things.  How about more shielding around the tubes like the Millen xmtr does?

[VK7ZL]

I read that for RF applications it is recomended that the bases of 807's be mounted below the chassis to prevent feedback which is the way I mounted the 1625's in my transmitter.

Bob

[The Slab Bacon]

Next question............... Why bother with as difficult tube to tame (at RF) as an 807, when there are so many good choices out there nowadays.

Even something as simple as 6146s would have made a far better choice.
Is it just nostalgia? ?

[w1vtp]

Bob

Yes, that is an option and yes it was necessary.  Millen 90800 did it another way - much like the torpedo approach except there was a can going part way up the tube's side

http://eshop1.chem.buffalo.edu/images/Millen/90800-exciter/90800_3.jpg

I have been aware that 807s had this problem occasionally.  Didn't think it would be so problematic as this particular case.

Yes, Slab, I'm thinking the same thing but then, I have the Bendix TA12 with three of those puppies inside just waiting to jump up and down looking for my attention. So, I may be scratching my own head when I try to use it.  It is fun getting these old bottles to perculate again

Al

[W2ZE]

If the 2 tubes are in parallel, then that means you have about .6 pF of g-p feedback. If you were running a 6L6GC, you would have 3pF of g-p. Is the amp nuetralized?

[N4LTA]

Not neutralized.  At 80 meters , I hoped it wasn't necessary

I tried to find some Millen 807 tube shields during building but that was not likely and didn't happen.

Don't really want to redesign this thing now that it is complete - I guess I could change the sockets and put a couple 6146s in if I have to.


I ordered the parts to complete a 30dB power attenuator and they will be here today - simple and I like to have a close to calibrated input to the SA. If you ever need one -  take your  whatever power rated dummy load and put a T connector on top - Power goes to one side of the T - the other side I mount a small box with a connector that fits the T connector  on one side and whatever you need for your SA cable on the other. Run a 53 ohm resistor across the SA connector and a resistor from center pins of the two connectors 790 ohms if I remember correctly but check  - you get a 30 dB attenuator with an output Z of very close to 50 ohms for cheap.  I think it works out to 30.1 db at 49.8 ohms out with an input z of close to 48 ohms. Not perfect but not too bad either.




Pat
N4LTA

[N4LTA]

Got my attenuator assembled and hooked up the transmitter to the SA.  Keyed it about ten times and all looked great - 2nd down 30 dB and the 3rd in the grass. Keyed it about ten more timed and got the final answer. Lower than fundamental frequency parasitics. Was a surprise to me.
They moved back and forth in frequency as the power slowly crept down and then back up.

Seemed to get worse as the transmitter warmed up. I could retune both the grid and the plate and stabilize it, but repeated keying brought it back and every now and then the parasitic  frequency swept past the fundamental up to 7 or 8 Mhz.

I am thinking about swaping the 807s for 6146s since they have very close to the same characteristics for class c phone. Maybe try a few tricks first.

Makes me want to line all the 807s that I have up and get out the .22 .

Pat
N4LTA