Mystery Grid drive

[K9ACT]

The attached schematic is what my 8000 rig looks like and has for about  two years.

This setup requires less than 20 W to produce the recommended 45 ma grid drive at 2000V on the plates.

I had it on the bench for a few days making some changes to solve some unrelated problems.

I never did like  the floating cap idea so while I  had it out I decided to rework it with a grounded cap and bias applied to the CT, like my 810 rig.

After getting it all back together and fired up, I found that my  exciter maxed out at 30W could only produce about 30 ma of grid current.

This of course is nonsense and people I talked to about it said impossible... something else is going on.

Having no idea what that might do, I did the obvious and put it back the way it was and I am back to 20W and 45ma.

My single 810 rig requires 30W to produce 50ma grid drive and I just wrote this off to the foibles of the 810.

The 810 is configured with the grounded cap and bias applied to the coil CT so just for a lark, I reconfigured it like the 8000 and guess what?

It now only requries 15W to produce it's 50 ma of grid current.

In both cases, this represents a factor of two in grid drive requrements.

This seems like a rather profound issue but I  have never heard or read any dissussion about this and don't understand why.

Can someone throw a little light on this?

js

[Tom WA3KLR]

Jack,

I've gotten erroneous grid drive readings in the past due to the grid current meter not being r.f. bypassed.

Put r.f. bypasses across the 2 meter movements and tie the one grid meter lead to ground instead of to the cathode meter.

Are the bias rectifier diodes actually exposed to the PA r.f. also?  The present series order of the bias supply and the grid meter may make the bias supply sensitive to r.f.  The bias supply return could be put to ground, grid meter upstream.  I see that the 10 volts to the primary of the bias transformer is the PA cathodes node.  They are bypassed but I'm wondering.

Try the first 2 steps mentioned first and see what happens.

GL OM.

[WD5JKO]

Jack,

   You always have interesting circuits!

I would refer to the following link and copy Figure 27 G. This will require re-arranging a few things to get that done, and will likely simplify your circuit. Just add your grid leak resistor in series with the grid RF choke, or maybe leave that choke out all together especially if your grid leak resistor is wire wound. Keep the grid leak resistor close to the input tank coil, and RF bypass the 'cold' side of it going to the bias supply.

http://www.vias.org/basicradio/basic_radio_28_04.html

As to the meters, sure bypass them as Tom suggested, but I'd leave them hooked up as your schematic shows. If you return your grid current meter to chassis ground, won't that make the plate mater equal cathode current? Remember that cathode current is equal to the sum of grid current plus plate current.

Good luck with those big triodes,
Jim
WD5JKO

[k4kyv]

Sounds like regeneration in the 8000 stage.  I  recall reading in one of the pre-WW2 Frank C. Jones RADIO Handbooks recommending using grid neutralisation with the tap on the coil grounded because it produces regeneration that cuts back the grid drive requirements.  The only difference was that the Jones handbook circuit used a wide-spaced single section variable across the coil rather than the split stator.  Electrically that shouldn't make any difference.  The floating split stator acts as a single capacitor formed by the two sections in series, with the advantage that its  stray capacitances are more symmetrical making the circuit easier to balance.  I'd need to get out those handbooks and relocate that article, but as I recall, they claimed that the grounded tapped coil arrangement is inherently regenerative.  Off the top of my head I can't see why.

In my pushpull finals I ground the split stator grid capacitor directly, and feed the midtap on the grid coil through the grid leak resistor, with no rf choke.  Using an RF choke at that point tends to make the final self-oscillate at some oddball frequency (usually below the operating frequency), as the grid rf choke and plate rf choke work together to act as a t.g.t.p. oscillator with the chokes as the inductive elements.  The wirewound grid leak resistor itself is inherently inductive, and combined with its ohmic resistance, its impedance is high enough at the operating frequency to eliminate circulating rf current  through the tap on the coil.

The frame/rotor in my split stator plate tuning condenser is grounded through a mica transmitting capacitor.  The +HV is fed to the frame of the tuning condenser, and there is an rf choke between midtap of the plate tank coil and the frame of the condenser.  This lifts the rotor plates to the same (modulated) DC potential as the stator plates, so that there is only rf voltage across the variable condenser.  This allows the use of less than 1/2 the plate spacing than would be required if the modulated DC plus the rf appeared across the plates.

This is known as series feed, since the PA plate current flows through the tank coil.  The requirements for the rf choke are much less severe than would be the case of shunt feed, where an rf choke is used to feed the +HV to the PA plates, with a blocking capacitor to the tank circuit.  In the series fed circuit there is very little rf across the choke.  Its function is to avoid circulating rf currents between the midtap of the coil and the variable condenser, since it is virtually impossible to tap the coil exactly at the null at the rf midpoint of the coil, which may actually shift slightly with different loads.  With shunt feed, the rf plate choke must withstand the full rf voltage output from the tube without affecting the tuning of the tank circuit - a task very difficult to achieve.  That's why the plate choke is such a source of trouble, and why the T-368 feeds the high voltage through the tank coil.

[K9ACT]

I would refer to the following link and copy Figure 27 G. This will require re-arranging a few things to get that done,

http://www.vias.org/basicradio/basic_radio_28_04.html

That is exactly the circuit that I am saying requires twice the ammount of grid drive as the one I am now using in both my transmitters and shown in the schematic I posted.

I have solved the problem but would like to know why it works this way and why anyone would ever use the grounded cap version.

I thumbed through my only pre-war Handbook and can find no reference so, Don if you ever run down that article, I would sure like to hear more about it.

js

[steve_qix]

I use (or I suppose, more accurately said, used) the RF grounded coil center tap version of the circuit, and floated the capacitor frame.  I used this or similar circuits in all of my triode RF ampfliers (I like using triodes).

Some tests would be very interesting.  An oscilloscope with propes connected to the tube grid, and to the grid-circuit-side of the neutralizing capacitor will surely show the issue.  There must be a difference between two circuits, and this would reveal part of what is happening.


In looking at the circuit, I can see several possibilities as to why lifting the coil center tap, and grounding the cap frame might yield a different result than the other configuration.

With the center tapped coil (transformer secondary), I believe you are more likely to achieve circuit balance unless the coupling is very, very loose.  I am guessing that, in the floating coil center tap configuration, there is less voltage appearing across the grid and more at the neutralization capacitor.  Did the neutralization change when you reconfigured the circuit?

The additional drive power is going somewhere.  I suppose the coil could be dissipating more power (although that is a lot), particularly if the circuit Q were different (were the 2 capacitor sections is parallel in one configuration, and then connected in series in the other)?  If not circuit losses, there may be simply be more coupling through the neutralization capacitor to the output network in the new configuration (or more RF voltage appearing across the neutralization cap).

Inquiring minds need to know 

Regards,

Steve

[K9ACT]

I am affraid that I can not answer the quesiton on change of neutralization with the two methods because I was doing other experiments and not quite sure where I left it.

However, K9WEK who is working on the same issue from Indiana says there was a change.  He also modeled the circuit in SWcad and this shows an increase in distortion with the CT grounded.  I do not see this in the real circuit.

We have also had discussions about the use of Pi networks with triodes and I am now thinking there is some connection between the two issues.

I found the attached paragraph in the 1947 issue of the Radio Handbook which touches on what KYV recalls reading but is much more generalized.

This caution would apply to any use of grid neut on single ended amps.

The reality is though, that either hookup has worked well for me from 80 to 20 meters, just requiring a bit more grid drive with higher freqs.

If regen and degen are reasons to avoid it, I would like to know why?

js

[steve_qix]

I always use a PI section output with triodes !  Works great 

Grid neutralization, just like you have it - and all was well from 160 to 20 meters (the transmitter did not cover 10).  This was with a pair of 450TL triodes.  Also did this with pairs of 250TLs, 833As and 812As.  Same deal with the PI section output.

Regards,

Steve

[k4kyv]

The Gates 1 kw transmitters from the 1950's on use grid neutralisation with a pair of 833A's in parallel.

I believe grounding the tap in the grid coil is supposed to cause the circuit to be regenerative, while grounding the effective tap in the capacitor (the common point between the two capacitors, or sections of a split stator capacitor, wired in series) makes it degenative.

Regeneration helps boost the output from the woefully inadequate parallel 807 rf driver stage in the Gates, saving the cost of a larger driver that would have reserve driving power. 

The only issue that should be of concern with regeneration or degeneration is the effect it has on modulation linearity of the class C final, and achieving adequate grid drive.  As long as the final stage modulates linearly, regeneration is an el cheapo way of driving the final, with modulation linearity at the extreme positive peaks further helped by modulating the driver stage up to 20% or so,  as the final is modulated in the vicinity of 100%.

[W3RSW]

Jack,
What are the small inductances shown series tied directly to the grids of the 8000's?
Value of them?  Might they be the cause of grid instability and/ or oscillation particularly at UHF rather than acting as chokes?

[k4kyv]

Wouldn't they be the parasitic chokes?

[K9ACT]

Wouldn't they be the parasitic chokes?

Roger.  Just the traditional 8 or so turns around a 50 ohm mud resistor.

Doesn't seem to make much difference in or out.

Seems like the "mystery" is no longer a mystery.

Been a fun journey.  When my new HB mostfet driver would no longer drive the 8000's, after I made the switch, it became more than an academic interest.

Thanks for all the ideas.

js

[k9wek]

What an interesting topic!

I have been working with Jack (K9ACT) on his Parallel 8000 amplifier for about 2 years now, and finally got excited enough to build one of my own.  I put it on the air about the first of this year, and have been using it every day since then to check in to the 75 Meter Noontime AF Forum hosted by Jack.

Jack asked me for design help back when he started building his amplifier.  I had drawn up schematics using a pi network output with a balanced grid circuit about 20 years ago, but never got the nerve to build it based on the discouraging information in several different handbooks.  When Jack said that he had this 8000 triode museum piece he had been keeping for many years, and wondered if it could be used with a pi network output circuit I saw my chance!  I designed the input circuit and protective bias supply based on the tube data sheet information and used one of the many pi network programs to come up with the output circuit.  Because of the worry over the input circuit I suggested to Jack that he use a toroid for the grid inductor.  I thought that it would force a better balance than an air wound coil.  Jack built the amplifier and then wanted to add a second 8000 in parallel for some reason.  I tried to discourage him but he did it anyway.  Needless to say, he made it work.

After Jack had his amplifier built we came across the schematic of the Johnson Viking Courier linear amplifier.  It uses a pair of 811A triodes in a similar layout.  I got a copy of the Courier manual and saw that it used an ordinary looking solenoid wound air core grid coil.  Also, in the manual was a detailed neutralization procedure.  This is, as far as I know, the only commercial ham rig ever to use grid driven triode tubes with a pi network output circuit.

When I built my Parallel 8000 rig, I grounded the rotor of the grid tuning capacitor and used a coil similar to the Courier's.  For some reason, to be discovered only after this thread started a couple of weeks ago, my amplifier took almost twice the power to get the same grid current that Jack's did.

Jack finally convinced me to try moving the ground over to the coil even though I was sure it wouldn't make any difference.  To my surprise the drive requirement dropped from 40 watts to somewhere between 20 and 30 watts.  (I can't afford a wattmeter!)  Now I was really confused.  I turned to my new friend LTSpice and discovered that indeed the Spice model agreed with the real world.  The drive needed is substantially less when you tap the grid coil vs. the rotor of the capacitors.  Also, the grid circuit is badly unbalanced by the feedback from the plate circuit with either hookup.  When the plate voltage is removed the circuit is only very slightly unbalanced by the grid loading.  This is due to the Q of the grid circuit.  But when you power up the plate supply the grid becomes quite unbalanced.  This seems to be due to the fact that the plate has a very big voltage swing and the coupling to the grid through the Cgp opposes the grid voltage while the coupling through the neutralizing capacitor is in phase with that end of the grid tank circuit, so it gives a boost to that end.  Grounding the center of the capacitor seems to give a better balance point than the coil, but it is still not obvious to me why one is regenerative and the other is degenerative.  I'm still looking for an answer to that one.

Can you get something for (almost) nothing?  I added a 2 uH inductor in series with the center tap of the 40 uH grid coil to ground to let the circuit regenerate some more, and the drive dropped to about 12 or 14 watts.  What is the limit?

Something curious happened, however.  My plate capacitor started arching over when going from transmit to receive.  This only lasts until the plate supply filter discharges.  This is the same problem Jack had a few weeks ago.  Coincidence?  Could it be that the regeneration causes the amplifier to go into oscillation when the drive is removed and the relay switches the antenna load off the amplifier?  This didn't happen with my amplifier until I went to the tapped coil arrangement, and then it happened after only about 20 minutes of QSO.

Does the extra drive power required with the tapped capacitor show up in the antenna?  I can't see it going anywhere else.  If this is the case, it seems like the degenerative arrangement is not so bad.

This is about the most interesting problem I have gotten myself into for quite a while.  There doesn't seem to be much in the literature about it.  I would appreciate any ideas or any reference material anyone on this forum may be able to provide.

Dan K9WEK

[W8ACR]

Hi Jack,

I was looking through an old Taylor Tubes manual from 1937 and came across this schematic. Although it doesn't have a pi network plate tank circuit, I thought you might find it interesting that they recommend grounding the center of the grid coil.

[W8ACR]

Hi again Jack,

As I look at this schematic again, I see that it is actually a different circuit altogether. Nothing is at DC ground. The coil is at RF ground with bias applied to the center of the coil. I should also mention that there is some discussion in the text of the manual about plate vs grid circuit neutralization. There is no mention of regeneration. The text does mention that the tube grid puts a resistive load on one half of the grid coil. This results in the other end of the grid coil being other than 180 degrees out of phase and it may be difficult to accomplish neutralization at the higher frequencies. The text does state however that grid circuit neutralization is preferred in unbalanced output stage.

Ron W8ACR

[K9ACT]

Hi again Jack,

As I look at this schematic again, I see that it is actually a different circuit altogether.

Well, it couldn't work anyway because there are no envelopes on the tubes.

Seriously, that is most interesting.  I will be anxious to hear what Dan has to say about this.

I am still running mine with the floating cap and he is running grounded cap and we both think we have good reasons.

js

[Gito]

Jack

I'm encouraging myself to give my opinion (don't laugh at me,if I am wrong).
The way I see it ,if you ground the the cap,you put one half of the cap(one of the two 40pf variable cap) directly in parallel with the input of G 1,So at a glance the circuit looks the same,but I think it is different .
Just a thought.


Thanks

Gito.N

[WA1GFZ]

Jack,
Maybe the RF choke bypassed to ground on the grid may have had too low reactance and was absorbing RF. Moving the bias to the CT may have eliminated the extra load on the grid. Since you now have extra drive a load resistor in series with a .001 to ground would shunt load the grid and make the amp more stable. A reactive load is much worse than a resistive load hanging off a grid so use the extra drive to your advantage.  I would start at around 2Kohms gfz

[Gito]

Jack


AS I wrote before ,When you ground the cap,you put one of the two 40 pf cap in parallel with The C input of the tube(call it C 1),maybe the C input of the tube is around 7 pf,when you tuned to resonance the Var cap has a value of 20Pf (for instance at resonance),so the value of C at the input of the tube (G1) is 20 + 7 = 27 pf,and the other var cap is 20 pf(call it C2) .So if this var cap has equal value.And for instance there are 100 Rf volt.across these var cap.it will divided  equally across these var caps ,thats 50 Rf Volt with the Ground as reference,But since the value is not equal the voltage across C 1 is 20/47 x 100 volt = 42.5 volt,since You need 50 volt to drive,the current you get at G 1 must be smaller than you expected.

Thats also happening  if you move the taps of the ground of the coil ( these is used when the var cap is floating/ungrounded) If the tap is in the middle/center of the coil,there are equal voltage at each end of the coil with ground as reference .but if you moved the ground tap from middle of the coil the voltage at the ends is no equal anymore (ground as reference).depending which way you moved it.So you can put a higher voltage or lower voltage to the grid 1 of the tube.by moving the ground tape on the coil ,(the turns ratio of this coil is not 1 to 1 again/not at the middle anymore)
 
I don't know if I'm right,please correct Me if I'm wrong.

Thanks

Gito.N

p.s  sorry my english  language is not good

[KM1H]

Im another who uses nothing but parallel triodes and a bandswitched pi net output. I detest messing around with plug in coils as well as open wire line.

The input is a HB version of the Bud turret input assembly and I never use a grid choke for the reasons mentioned. I do use parasitic suppressors in the grid and plate leads as a precaution. Id also suggest a .0047 or .01 plate choke bypass for 80/160M use especially if using electrolytics in the PS.

I went with the grounded cap on the 2x  250TH, 810, and the in process HK-454 amps since the Viking I, II, and HT-9 have an abundance of drive and I still have to swamp with inline attenuators.

Now all I have to do is finish cleaning up the 12/5/08 ice storm tree and antenna mess and fix or build a bunch of antennas. The only bands I have now are 30 & 20M.

Carl
KM1H