AB1 Modulator Drivers

[KE4LRL]

Guy's

I am putting together an AB1 modulator using pair of  4-65's.  (Input to a CVM-3 for modulation of a  240 watt carrier).

Any comment on a preferred method of driving an AB1 pair?

 - Modulator grids direct coupled to a phase inverter/driver with feed back from modulator plates to driver plates

- Couple the modulator grids via a transformer and use feed back to the last audio stage from the modulator secondary

-Couple the modulator grids to a pair of drivers driven by a phase inverter or transformer coupled driver with feed back from modulator plates to driver plates


Thanks & 73
Jeff KE4LRL

[w8khk]

Hello Jeff,

If you are running AB1, you should be able to eliminate the need for a driver transformer.  Phase splitter and a voltage gain driver stage should do the trick.

I read an article in QST for March, 1948, page 13, entitled "500 Watts of Audio from AB1" by Harry A Mandoli (ex W6HAM) and David H. Atkins, W6VX.  If you do not have access to this resource, I can help, just PM me.  The modulator uses a pair of 4-250s, with only three stages from mic to tetrode grids.  The authors designed the modulator with enough gain to allow application of inverse feedback, but it was not needed, so they left well enough alone.   Their design uses a 6J7 pentode for the mic preamp, a 6J5 after the gain control, then a pair of 6SJ7s for phase inverter and gain stage.  Of course this could be modernized, but at least this gives an idea of what can be accomplished driving the similar 4-65 tetrodes in AB1.  If you run a separate audio chain, your implementation could be much simpler.

The article provides several pages of theoretical design data, biasing, power supply details, and comparative phase inverter designs leading up to their final implementation.  The article has seven pages of information and diagrams.  The 4-65 will work with a very wide range of plate voltage, depending upon the power and load impedance desired.  Sounds like a fun project!

73,
Rick

[w4bfs]

sorta echoing what rick sez, ifn you don't have access to old Qst, download Bauer 707 schizmatic from Bama ...very similar...73...John
 ps. please note Bauer made an understandable error in schizmatic ... left off component values ... crazy voltage divider / feedback ckt  off the modulator tubes !  probably use something different, mysef

[steve_qix]

Hi !

There are numerous, resistance coupled methods for this, however if you're looking for something that already "works"  the RCA BTA1-R 1kW broadcast transmitter uses a pair of 4-400s in the modulator - resistance coupled AB1.  The transmitter has very, very good audio for a non-pwm transmitter, and negative feedback is included (you can use that part or not).  The schematics, I am quite sure, can be had over the 'net.

As I recall, the driver was not particularly complex, and it definitely worked very well.

Regards,

Steve

[AB2EZ]

Jeff

FYI... I used this approach to drive the 6550's in my EFJ Ranger. It works very well.

[As an aside, I'm using Zener diodes (a pair of 15V 5 watt diodes in series) in the path between the cathodes of the 6550's and ground... to set the bias on the 6550's at 30 volts... independent of the average total plate and screen current]

http://mysite.verizon.net/sdp2/id13.html

Best regards
Stu

[WBear2GCR]

Generally speaking I prefer to drive output tubes in AB2.
That's because in AB1 you hit a fairly solid "wall" at the point where the driver voltage hits the point where the grids want to draw current - which means a fairly hard clipping. Otoh, in AB2 the onset of clipping due to grid current being drawn is not there, and there is both some extra output power available and a "softer" characteristic as the tube runs out of "snot".

But, if you never hit clipping then the issue is moot.

The cathode follower and the transformer, or any DC connected grid connection is capable of running the tube(s) into AB2.

Another problem that can occur with capacitive connection to the grids is where the cap charges up due to large signal excursions and/or clipping and has sufficient time constant so as to "hang" the grids in conduction for too long. There is at least one very "famous brand" tube amp that exhibited this very problem - and was known for "eating" copious quantities of expensive NOS 6550s.

Take a look at the 4-65 curves and see what it calls for in terms of power in AB2 to drive it fully... probably not much. If you're driving in AB1, capacitively coupled no real power is needed anyhow, just voltage swing sufficient to drive the grid.

                 _-_-bear

[WBear2GCR]

Sorry, I'm confused.

Bias the tube up until it draws some current?
Or bias the grid until it draws some current??

Honestly, doesn't seem like either solves the problem, unless I misunderstand something here.

            _-_-bear

     

[KE4LRL]

Guys

Excellent posts with a number of points to ponder.  The 6SJ7 seams very popular for the AB1 driver arrangement - I wonder why? 

Jeff

[WU2D]

The headphone guys have some good low Z driver circuits which should drive the grids nicely. The White Cathode follower circuit is very effective and it is a self generating push pull type so the top and bottom tubes source and sink. I built one of these for a friend who wanted a headphone amplifier and I could not believe how low the distortion was.  6S4's top and bottom.

http://www.tubecad.com/2007/04/blog0104.htm

And in an application driving a big triode:

http://www.plitron.com/PDF/GLASSkoby.pdf

Mike WU2D

[WBear2GCR]

Sorry, I'm confused.

Bias the tube up until it draws some current?
Or bias the grid until it draws some current??

To quote myself "bias up the grids on the 4-65's where they just start to draw a little current"

I'm quite confused, a sheet says that they are biased typically at -100 volts, dunno if that is for Class C or what, but no matter, it isn't a zero bias tube. So, I don't see how bringing up the bias voltage (less negative) will draw any grid current at all... (i'll be able to check my RCA  or Eimac book this weekend, it's not nearby now)

Also, as you make the tube conduct more, away from class B, you're lowering the headroom because you're moving closer to class A.

Whereas if you run in AB2 you gain headroom... which was my original point.
Distortion due to clipping or running out of headroom can not be made up by negative feedback or biasing.

Afaik, the point at which the grids begin to draw current is the point at which you drive them positive, and that takes power which can not be supplied through capacitive coupling.

Perhaps I have some misunderstanding here, but that's what I thought about how it works.

                    _-_-WBear2GCR

[Opcom]

The RCA 1K uses the BTA-250 transmitter for a driver. The schematic for the BTA-250 using 6J7 pentodes to drive 828 tetrodes in AB1 is here:

http://www.bunkerofdoom.com/kd5oei/bta250L/rcabta250L.html

The whole thing is an input transformer, two 6J7, two 828, and the mod iron. Global feedback is also shown. The rig was very good sounding.

[AB2EZ]

Its ok to drive the tubes into Class AB2 on occasional voice peaks, even with a capacitively coupled driver provided:

A. The percentage of time that the tubes draw grid current is small (as it would be for voice modulation peaks)

B. The coupling capacitors are large enough, so that, in conjunction with the resistors that lead from the grids of the modulated tubes to ground (directly or though the bias supply) the associated RC time constant is greater than around .1 seconds. For example, if the grid resistors are 10k ohms, then the capacitors should be 10uF

Why is this okay?

When the modulating (output of the capacitively coupled driver) grid voltage on either tube gets big enough (i.e. on a positive or negative modulation peak) to cause grid current to flow... that will cause the associated coupling capacitor to charge up by an amount that produces a small voltage across the capacitor (i.e. small relative the the peak voltage of the modulating waveform). The next big modulation peak will add more charge. However, between modulation peaks, the grid resistor will continuously discharge the coupling capacitor. As long as the modulation peaks occur only for a small percentage of the time... the coupling capacitors will not build up a significant offset.

Note: If you apply a 100% sine wave modulating signal, then the coupling capacitors will charge up a little on each cycle of the modulating sine wave.... and the resistors will not be able to remove this accumulating charge until the coupling capacitors charge up enough to keep the grid current from flowing (almost zero average grid current). Thus, the modulator will not behave the same way, when "tested" with a 100% modulating sine wave (which will shift the biasing of the tubes) as it will with a normal speech waveform.

Stu

[Opcom]

Note: If you apply a 100% sine wave modulating signal, then the coupling capacitors will charge up a little on each cycle of the modulating sine wave.... and the resistors will not be able to remove this accumulating charge until the coupling capacitors charge up enough to keep the grid current from flowing (almost zero average grid current). Thus, the modulator will not behave the same way, when "tested" with a 100% modulating sine wave (which will shift the biasing of the tubes) as it will with a normal speech waveform.

Good point, always adjust the input level from the sine wave generator both low and high just like you would by varying its frequency while testing.

A triggered gate in series with the test signal, synced on the signal (zero cross or peak), can be made to allow a few cycles through for checking 'peaks' and yet not be overcome by the effect. A simple setup would be to use a 1KHz tone and a 10mS gate, allowing 10 cycles through, then 90mS of silence (rest) for the caps to discharge/recover. The scope trace for observation would be run at 10Hz rate and 2mS/division.

A good use for the old $5 HP pulse generators at hamfests. I've done this for various things with a 1-10% duty cycle, and used the pulse generator to drive the cathode of a 12AX7 amplifier, making it into a switched amp so to speak. It was a necessary technique also (but with different gating circuit than a 12AX7!) for testing a KW amplifier with a 100W dummy. I believe this technique is much overlooked. Someone makes a pulser for ham radio linear tuning, but why buy when make is more fun, unless you are in a hurry or simply want an OTS device.

[Opcom]

A triggered gate in series with the test signal, synced on the signal (zero cross or peak), can be made to allow a few cycles through for checking 'peaks' and yet not be overcome by the effect. A simple setup would be to use a 1KHz tone and a 10mS gate, allowing 10 cycles through, then 90mS of silence (rest) for the caps to discharge/recover. The scope trace for observation would be run at 10Hz rate and 2mS/divisions you are in a hurry or simply want an OTS device.

I wonder if you could construct the same setup with a freq sweep function enabled at the same time?

One thing I have also found in the past too is that the resistor values used across the grids make a difference. That is why I normally don't select resistor values and tie them to ground just to bias the grids at the right voltage. The resistors are the load for the phase inverter. Granted, they are in parallel with the grids and that resistance will divide and go down some, but it still makes a difference. I connect the resistors in a center tapped configuration and connect the bias supply to that center tap just as if it were a driver transformer.

Since a time constant of 0.1 second was mentioned, a sweep does not seem practical for lack of time.

Isn't the object to determine the efficacy of the RC coupling circuit's ability to drive the grid momentarily positive in response to a voice peak? My suggestion was to use a simple method to repetitively simulate the average grid current conditions present during a voice peak. It does not have to accurately reproduce a voice peak, just cause the same capacitor charging function within a very small portion of the time constant. Then the RC coupling setup could be evaluated in terms of gross distortion.

In addition, a triangular wave would be better than a sine because it is linear and the oputput can easily be compared to the input on the scope (or a second scope) in XY mode. I realize I am assuming the experimenter has some test equipment, but nothing out of the ordinary or exotic.

[WBear2GCR]

According to the RCA sheet on the 4-65, in AB1 you can get 175watts with 1750 on the plates. Whereas in AB2 with 1800vdc on the plates you get 270 watts out.

Bias in AB1 is -90 volts.
Bias in AB2 is -50 volts!

In both cases they call for "180 volts" driving signal grid to grid.

Screen in AB1 is 500vdc, whereas in AB2 it is 250vdc.

Driving power in AB1 is 0 watts, in AB2 it is 2.6watts (higher at lower B+, btw).

Clearly, for this tube there is significantly more power available in AB2.
I'm unclear on what happens to this tube if one sets the bias like in AB1 but can send power to the grids - perhaps this exceeds the current capability of the tube.

WRT to the proposal to set the capacitive time constant of the input to an output tube's grid so that the grids "hang" and draw current ( I think that is what is being suggested?), I personally do not think that this is a good method. Perhaps some tubes will handle this arrangement, but I doubt that it will work as advertised (push the tube into AB2) except to a very minor degree at best, and the down side is two fold: potentially dumping useless LF energy into the iron (saturation) and possibly causing damage to the output tube due to excessive plate & grid dissipation.

In the case of the 4-65 RCA shows the grid being driven positive to 100volts, from the initial bias of ~-50volts. Merely changing the bias voltages to the one of the AB2 case and capacitively coupling doesn't make it run AB2. Since that swing requires power from the driver to be transferred, I don't see anyway that capacitively coupling to that grid will make that happen... of course maybe I am just misunderstanding something here?

                 _-_-bear

[AB2EZ]

Bear

I believe you are misunderstanding what I said.

I won't restate it, since I said it as well as I could.

Stu

[KE4LRL]

Bear
Good point on the screen and bias voltages in AB1 as compared to AB2.  While I have not worked the math out I world believe the higher bias voltage (equal to the audio input level on the grid) and the 500 volts on the screen prevents the tube from drawing grid power and the tube would not go into AB2 when the exiting single over comes the lower grid bias voltage.

I fact one of the articles I have read on AB1 modulators states “ when the grid input (audio level) is greater than the gird bias voltage the signal will clip as it would in a low level clipping stage”.

It will be interesting to find out what happens when the high frequencies along a pre-emphasis curve over come the DC bias voltage an clipping begins on peaks.

Guy’s
Tthanks for the comments on other BC transmitters that run AB1 modulators. It appears that the 6SJ7 was a real favorite.  Perhaps a 5879 would wok as a “modern” option for the 6SJ7.

Stu
The RC time constant as related to audio has to be a consideration for any circuit.  Off had the only thing that comes to mind is a low impedance path to ground through the bias supply.  That may get the caps to discharge and prevent the chargefrrom hanging due to the higher grid impedances.
 

Rick
I had a chance to read the article today. Several really good points surfaced:
High and low voltage supply regulation
Well designed low level class A operating points
Lower current demands for a given set of tubes (of course lower efficiency as compared to AB2)
In particular each plate of the 6SJ7 is loaded with a cap/resistor to ground prior
to the coupling cap used to feed signal to the grid of the modulator.   (As if to provide a constant load to the plate much the same as adding a resistor to the secondary of a driver transformer.)

Mike
I think I have the schematic of one of the original White audio amplifiers.  I am looking for the article that accompanies the schematic.  Interesting circuit, if I remember correctly the actual work (the physics) of the audio output transformer was used to determine the electronic circuit parameters.

I hope to get started on my modulator just after the Shelby hamfest.

Jeff
KE4LRL

[AB2EZ]

Attached below are

1. A schematic
2. The results of a simulation using LTSpice/SwitcherCAD III (0 => 1 second) [2nd slide]
3. The last 0.1 seconds of the simulation (0.9 => 1 second) [3rd slide]

for one half of a 4-65A push-pull modulator, operating in Class AB2.

A. The 4-65A is biased with a fixed -25 volt bias supply fed by a 10k ohm resistor (R3)

B. The 4-65 grid is modeled as a diode (D1) in series with a 1500 ohm resistor (R1). This is obtained by examining the RCA 4-65A Data sheet (http://www.mif.pg.gda.pl/homepages/frank/sheets/049/4/4-65A.pdf) which has a curve of grid current vs. grid-cathode voltage, for values of grid-cathode voltage between 0 and +100 volts. I picked a worst case value for R1 (largest grid current/grid voltage) of 1500 ohms, which occurs when the grid-cathode voltage is +100 volts, and the corresponding grid current is 68mA.

C. The grid input signal source is a capacitively coupled voltage source. The value of the coupling capacitor is .00001 Farads = 10uF.

D. The voltage source is assumed to have an internal resistance (R2) of 100 ohms (e.g., a source follower or a cathode follower). The signal is a sequence of square pulses: 100 volts peak, 0.0001 seconds rise time, 0.0001 seconds fall time, 0.001 seconds wide, with 0.03 seconds between pulses (i.e., duty cycle = 0.0011/.03 => 3.7%). The pulses start (voltage source turns on) at T=0.0 seconds

Note:

The peak grid-to-cathode (ground in this model) voltage on the 4-65A is around +56 volts. This compares to +68 volts* for DC coupling (i.e. no coupling capacitor, and bias supplied via the cathode). 

*I.e. (100 volt peak pulse height x 1304 ohms /1404 ohms) - 25 volt fixed bias = 68 volts; where 1304 ohms = 1500 ohms in parallel with 10k ohms]

You can see the 10k ohm resistor removing charge from the 10uF capacitor, during the the gaps between pulses, by observing the sawtooth waveform that varies between around -43 volts and -39 volts.

The pulses (representing modulation peaks) cause the 10uF capacitor to develop approximately 16 volts of extra negative bias (i.e., it charges up to ~ 16 volts) for this specific (illustrative) example. [(43 + 39)/2 = 41; and  41-25= 16.]

A key factor in avoiding "clipping" of the positive peaks is that the source resistance (100 ohms) is much less than the forward-biased grid resistance (~1500 ohms).

The last .jpg slide [4th slide (see my next post)] shows the effect of increasing the coupling capacitor to 100uF (compare to the 3rd slide of this post)

Stu

[w4bfs]

a very interesting thread ... as I mentioned in a parallel thread its all in the source and load impedances ... following what Stu has modeled, as the load goes non linear with a 100 Ohm source Z and a 1000 ohm load Z implies a 1 order of magnitude difference or ROUGHLY 10% distortion ... increase this to 3 orders of magnitude (100 Ohm source and 100k Ohm load) and the distortion drops to .1% distortion, all other things being equal ... hope this helps ...73 ...John

[AB2EZ]

Here is the simulation (0.9 to 1.0 seconds) with the coupling capacitor changes to 100uF

[WU2D]

Jeff,

Audio people are nuts...

I have now built the White circuit with 6CG7's and 6S4's. The amp that I dropped off was a single 6CG7 cathode follower with a standing current of 12 mA. The preamplifier is a cathode coupled 12AX7 long tail with a FET constant current source.

I dropped off the little White output headphone amplifier wagging my head that I did not think it was going to drive my buddies various headphones as well as I had hoped. Testing using sine waves into resistors, it was looking like the output impedance was around 600 Ohms. After that the sine wave starts to round off and I figure it messes up the bias. There is only so much standing current you can run in a small tube. As you know, doubing this current should square the output power which is why you see 6080's and multiple tubes in parallel being used. There is nothing less efficient than class A transformerless low impedance circuits with tubes!
A little voice in my head that sounded like Timmy kept saying - "Use a transformer dummy!"

We switched the 300 Ohm Senhiessers (he had three kinds) from his 700 dollar solid state headphone amp to the little homebrew tube amp. Now Roy Orbison and Blondie are apparently his standard test sources. He said that the tube amp easily blew away the solid state amp and then used he weird language that I did not understand. As a reward he gave me a pair of AKG K601 reference phones "so I could hear stuff" and he said that they are 120 Ohm phones so If I can drive them, that may be a good test. 

Mike WU2D

[Opcom]

I think the primary key to this is to use a driver circuit that has a very low output impedance so that any charge buildup on the coupling capacitor during the AB2 portion of the cycle would be eliminated immediately upon transition back to AB1. In either case the driver would supply the required current to charge and discharge the capacitor in a way that the voltage presented to the grid would be regulated to the input voltage regardless of the current. Some feedback would probably be needed around this circuit to make sure any charging artifacts from the combination of the capacitor and any shortcomings of the driver stage WRT the ability to discharge the capacitor were eliminated.

One enhancement that might help to present a more linear load to the driver would be to put a series combination of a resistor and rectifier from grid to ground so that it would draw the mirror image of the grid current when the grid was cut off. Since the grid draws about 25mA at +55V (taking from previous posted AB2 operating conditions and the grid current curves) the resistance in series with the rectifier-resistor combination could be about 2200 Ohms. A caveat is that the ressitance of the grid to cathode of the 4-65A is not linear with respect to voltage. So find a vacuum diode with the same curve.

The above paragraph serves mainly to point out the very low impedance of the driven grid. The tube manual also reccommends a bias supply source resistance of <250 Ohms for AB2. The driver has to deal with this.

The current waveform through the capacitor would be distorted in AB2 but the voltage output of the driver at the driven grid would/should not be distorted. It could be done in any of the ways suggested, either by a source-sink (does anyone say "totem pole" any more?) or a source with a low resistance value to ground. As to the selection of the coupling capacitor, I'd leave that to the audiophiles except to say that an electrolytic would not be my first choice.

This is all kind of hairy. I think I'd rummage for a transformer, or if that's out of the question, consider a DC coupled driver. It could also double as the bias regulator.

[WBear2GCR]

Driving power in AB1 is 0 watts, in AB2 it is 2.6watts (higher at lower B+, btw).

I think you mean 0 mA grid current not 0 watts for AB1. It will require more than 2.6 watts of drive for AB1 operation.

The original post said they wanted to build AB1, but wanted to know which type of AB1 to go. AB2 is more efficient, but it will require a driver transformer. The discussion was that you could draw some grid current to get things started without taking it up to full AB2 operating parameters and minimize or not produce distortion, etc.

Sorry, that is zero watts[/u] in AB1. You need to swing volts, and yes a very very small wattage to charge and discharge the cap, and of course the tube has its own dissipation when swinging those volts, but none of that power is used by the output tube in AB1.

                   _-_-bear

[WBear2GCR]

It's semantics.

In this case you can't break the relationship between the current on the grid and the voltage on the grid, and since we're talking about the maximum required and P = I*E, that's the same as watts. RCA specs it in watts to drive, and provides the curves that show the current for specific grid lines at given tube element voltages (of course).

My point is simply that I doubt very very strongly that any capacitive coupling will actually drive a tube like the 4-65 into any sort of AB2 condition like the one previously described.

Personally, I prefer to run with a coupling that will permit AB2, if you want to go out of AB1 or not is another issue - I don't want to hit that AB1/AB2 threshold limit if it can be avoided by simply building a "better" driver.

Stu, I haven't had time to really look at your simulations, but I suspect that the tube model doesn't include the shift from zero grid current draw to current draw as you go from AB1 to AB2? Don't really see how a diode can model a pentode properly.

          _-_-bear                     

[Steve - WB3HUZ]

And he made this assumption.

The voltage source is assumed to have an internal resistance (R2) of 100 ohms (e.g., a source follower or a cathode follower).

You ain't gonna get that with a high gain pentode. I doubt any triode based phase splitter would do it either. A cathode follower could but why not just direct couple it and forget all the work arounds? The main point for using RC coupling and AB1 is simplicity (check  out the skizmatic in Opcom's post - dead simple). Once you starting doing things more complex, you've lost the advantage of the approach.

It's semantics.

In this case you can't break the relationship between the current on the grid and the voltage on the grid, and since we're talking about the maximum required and P = I*E, that's the same as watts. RCA specs it in watts to drive, and provides the curves that show the current for specific grid lines at given tube element voltages (of course).

My point is simply that I doubt very very strongly that any capacitive coupling will actually drive a tube like the 4-65 into any sort of AB2 condition like the one previously described.

Personally, I prefer to run with a coupling that will permit AB2, if you want to go out of AB1 or not is another issue - I don't want to hit that AB1/AB2 threshold limit if it can be avoided by simply building a "better" driver.

Stu, I haven't had time to really look at your simulations, but I suspect that the tube model doesn't include the shift from zero grid current draw to current draw as you go from AB1 to AB2? Don't really see how a diode can model a pentode properly.

          _-_-bear