The AM Forum

   Whilst perusing an online audio forum (diyaudio.com) I came across an interesting circuit arrangement. Similar to the old RCA "Type 807 in Special Triode Connection" circuit, audio drive is applied to the screen and control grids of a pentode — but with a difference.

   Similar to the RCA method, drive is directly to the screen, and through a resistor to the control grid. But this new method adds a second resistor from the control grid to ground. Unlike the RCA method, tubes like a 807 are not good candidates. However, TV "sweep" tubes show excellent results!

   Most sweep tubes have a relatively low screen voltage limit and a "Triode Amplification Factor" (as stated on the datasheets) of between 3 and 4 or so. Those characteristics are what make this drive method work well.

   How does it work? I can't explain it better that the author "smoking-amp" who has performed many hours of evaluation and refers to the method as "Crazy Drive".

   He posts many plate curves that he ran on various tubes. Unfortunately, these cannot be viewed by non-members of that forum. But I can tell you that he gets some fantastic linearity when using Crazy Drive!

   I plan to try this in the near future. Anyone else? Comments?

Here's a link to his description:
https://www.diyaudio.com/forums/tubes-valves/307517-ruminations-screen-drive-crazy-drive-21.html#post5531436

That is very interesting and will work great with most sweep tubes.
Depending on the tube his voltage divider ratio may need adjustment. Some sweep tubes don't even need grid drive if driven by the screens only and a little distortion is not a big deal.

He's going for linearity in the stage -those audio folks don't like much NFB, but for modulators NFB is fine.

I'd be concerned in the circuit below linked, with control grid dissipation - there is shown a divider having 1.78K between G1 and the driven G2, and I would not want to be that control grid with a loudmouth at the mike.
https://www.diyaudio.com/forums/tubes-valves/307517-ruminations-screen-drive-crazy-drive-post5513530.html


The special class B arrangement used 20K in series with the control grids to limit the control grid current.

Anyway, modulation-minded folks could join there and read that topic, there are lots of things to consider.

I saw that thread... didn't see the schematic though, or the specifics of the way to set it up.
Somewhat confused as to how the grid gets negative bias with a resistor to the cathode?
It is to the cathode, isn't it?
I could see it to ground, IF there is cathode bias... maybe.
Anyhow, unclear. Not sure where the drawing is...

The 6HS6(?) curves were scary linear.

And, I guess that this sort of linearity isn't really essential in a modulator, but it would not hurt.

I'm not a fan of feedback around transformers.
It causes problems in the way transients are handled, aka ringing and overshoot.
Again, probably not a real issue in modulators, but I don't like it anyhow.

                      _-_-bear

Is there a specific link to the schematic being discussed?


Phil

In "normal" screen driven circuits that I've seen, the control grid is simply grounded. Crazy Drive utilities that grid to linearize the overall gm. The resistors set up a voltage divider to provide a proportionate drive to the control grid. I believe that getting the ratio and total resistance of the divider correct will determine the amount of drive compensation and limit the grid current respectively.

Yes, just going to screen drive is the big step. Seems to make the kinks go away. Another advantage of crazy drive, as the author mentions elsewhere, is the it is easier to drive than screen drive alone. I'm sure a low impedance driver is still required (preferably a mosfet) but apparently less voltage swing is needed.

To me, if I were going to go with screen drive anyway, it only takes 4 resistors to make it crazy drive. The open question is the resistor values. Some work needs to be done to find the proper values. If that were known, it would be a simple matter to convert any screen driven sweep tube modulator (of which there are many) to crazy drive. Maybe not a great benefit, but certainly an inexpensive one.

Schematics and photos can be found in the other forum by its members. I don't want to post them here directly - they're not mine. But the circuit is an easy one to describe.

A pentode is screen driven (such as a 6DQ5). A resistor is placed between the screen grid and the control grid. A second resistor is placed between the control grid and ground. That simple.

You bet. There are some very, very sharp engineers on the board. I think there's a wide area of cross-over between AMfone and diyAudio. I know we each tend to snub our noses at each other sometimes, that's human nature. Tribalism they call it. But that's for a psychology forum! ::)

Phill,

   This is one link:
https://www.diyaudio.com/forums/tubes-valves/307517-ruminations-screen-drive-crazy-drive-2.html

I downloaded the LT Spice file, and converted it to a .JPG file. I attach it here. My library is missing the pots used, but it is clear enough to figure that out.

The Fet source follower driving the output tubes as been something that big George came up with at TubeLab many years ago. One of his famous typologies is a "weak" type 45 amplifier with a solid state constant current load, and that driving the FET source follower. The Power amp (single ended) can be then operated class A2 instead of A1. This dramatically increases the power while the low source impedance drive keeps distortion low. Then without NFB, the distortion from the 45 amplifier will cancel out much of the distortion from the Power output tube (perhaps it is a big triode like a 300b or 845).

Jim
Wd5JKO

Referring to that schematic:

Notice the idle current is set to only about 38% of the Pd. I believe that crazy drive (if set up optimally) may lower crossover distortion enough to allow an even lower idle current than that to be used. More experimentation needs to be done.

-Low idle current
-Better linearity
-Easier to drive
-Inexpensive to implement

What's not to like?

Thanks Jim,

I had heard of screen drive for 6DQ5's but had never seen an actual circuit implementation.


Phil - AC0OB

Here's a classic 6DQ5 screen driven modulator from WA1QIX
http://amfone.net/Amforum/index.php?topic=30337.0
Methinks it could easily be modified for crazy drive. In fact, Steve even mentions driving both grids in the notes of his schematic!

Note that the schematic posted is DC coupled to the screen... so it is in class A2 or AB2.

Also, there is a cathode resistor - presumably that provides the (some?) requisite negative bias, which
IF it's in mostly or all Class A is not much... but maybe it's not really doing that much.

Trying to run that in more like AB2 might be impossible, as it likely biases where it biases,
plus or minus some cathode bias...

Solo screen drive that I have seen (no grid) isn't all that linear. Also, it seems that not all tubes
produce the same wonderful linearity that the "magic" circuit is tried with.

But it does produce a fairly wonderful set of curves for the 6HS6 tube.

             

The cathode resistors are 1 ohm. I think that's only used to monitor cathode current when setting the bias for the correct idle current. Note the +/- rails for biasing of the screen. I'm thinking of it as a triode with the screen being the active, driven and biased grid and ignoring the control grid essentially. The control grid is used only for its ability to help linearize the gm. The schematic states 20mA(9.3W) at idle per tube with 42W output so I suspect its class AB(?).

Again, the key to crazy drive is using tubes with characteristics that are common to sweep tubes - low triode mode amplification factor (3 to 4 or so). For reasons beyond my limited understanding(!), that's when the proportional driving of the two grids can effectively linearize the overall transconductance of the tube.

I suspect the crazy drive circuit has the ability to linearize a beam power tube better than the dubiously named ultra-linear circuit. Sweep tubes are a poor choice for UL mode anyway as the plate voltage must be kept very low (due to low screen voltage limits) sacrificing output power and basically wasting a good tube.

The outstanding characteristic of sweep tubes is their high peak cathode current rating (this at the sacrifice of using more filament power) and relatively low average plate dissipation. So, in a service where there's a high peak to average requirement, such as a modulator for voice, they're a good choice. Combine that with a way to make them very linear, their small physical size, the many different ones available, relatively low cost, and they become very very hard to beat.

More experimentation needs to be done with the crazy drive method. I plan to do some, and I'm hoping others will take up the task as well. Hey, if I'm lucky, one of you will figure it all out and save me the trouble! ;D ::)

I recall Pete W1VZR used 6DQ5's in a Valiant with Screen drive. Back in the 1980's, he had one of the best sounding AM rigs I had ever heard. Is Pete still around?

The Gonset transmitters sometimes use screen drive to the modulators. I believe a Harvey Wells transmitter also does so. I attach some ideas put into my G-76 to enhance the modulator. It would be fun to try adding some in phase audio to the control grids.

The mods I did allow for 130% positive peaks with a sine wave, and with voice peaks, I can go to 140% or more. I use a soft clip on the negative peaks that gradually kicks in at around 70% modulation, and gets progressively more intense at 90% modulation. I attached few documents about that. I have more, scope traces, 1 tone, and 2 tone tests as seen on a pan adapter. Seems very clean until the Negative Peak Attenuator goes into full swing.
Jim
Wd5JKO

 A few plots attached of the G76 at 50% modulation. The plots at 75% are a little worse, but not bad. The last scope image is with sine wave modulation adjusted up to 130% positive peak. See the attenuated, but not clipped negative peak that does not exceed 100%. Lower the audio level down to around 75%, and the modulation negative peak attenuator does nothing.

Yes, the G-76 may be a good candidate for crazy drive. I don't know if transformer coupling will be stiff enough to drive it but it would be easy enough to try (only 4 resistors). The control grid is grounded on TX but it's hard to see what they're doing with it on RX. You'd need to figure that out. Obviously it wouldn't be grounded after modification. It would be at the tap of a resistive voltage divider from the screen to ground. It would be interesting to wire it up with a switch or relay so you could A/B it, if you're so inclined.

Here is another dual grid drive circuit from an old Handbook. See attachment. For a mobile application, it had several advantages,: no bias supply, no screen supply, low modulator idle current, and high efficiency. This is similar to the zero bias 6N7 based 10 watt modulators.

Yes, that uses the same connection (20k - 22k resistor between grids) to the 807 as the circuit that I refer to in the Title. I've attached an image of it from the RCA Transmitting Tube Manual TT-5.

While crazy drive does not appear to help much with the 807, it seems to work very well with sweep tubes.

Don

The worth of sweep tubes is in those that require only a low G2 voltage for maximum plate current.

The driving voltage to the screen (and grid combo) has to be quite high for the 807, and much lower for the sweep tubes.

The tube manuals show which tubes of various kinds use a low screen voltage.

With a small OPT run backwards 700-800V peak grid to grid voltage drive for 807s shown in the handbook, and other screen-grid tubes like them, is not too hard. There are some class B driver transformers that might work but those seem a bit rare to get.

For small stuff and when it's wanted cathode followers to eliminate the need for a driver transformer, the peak screen to screen voltage of 300-400V is easier to get.

Also consider the headroom wanted from the driver as 'good practice' and sweep tubes are even more attractive. It may not be unreasonable to use a 6080 for those who prefer tubes to transistors.

Hard to beat a mosfet source follower, and I may use them, although the "purist" in me would like to use a cathode follower. Yes, the 6080 comes to mind (I have a bunch of them) but I'm looking at the 6BL7GT for a future project. It has a 500V plate voltage limit, lower filament current, slightly higher gm, and similar Pd vs the 6080/6AS7.

I have one NOS 6BL7GT in my "inventory" which I inherited from my father that he had for his TV repair shop back in the 1950's. Always a favorite of mine to put some of his old stuff on the air — so there may be a cognitive bias in my choice. ;)

6080 seems like WAY too much tube for a driver... way way way.

How many watts do you think is needed?

I think the object is to have a cathode follower with a low impedance output as a driver. Both the 6080 and the 6BL7GT would have an output impedance of about 140 ohms. If you know of any tubes that can better that, I'd like to what they are.

I much prefer a cathode follower to drive the grids positive instead of an inter stage or driver transformer.  I never looked at the 6BL7 before.  Nice tube.  Reasonable filament current at 1.5 amps, and nice voltage capabilities too. 

I generally use a 6SN7, and I have used the 6AS7/6080 as well.  As we get bigger, we need substantially more filament current.  Another step up would be the 6336, but that is probably way overkill for driving sweep tubes.  Nice they all use the same pin configuration.

Unless you have really big sweep tubes, then you might consider a pair of 304-TLs as drivers for each phase. 

I don't understand an objection to the 6080 for power reasons, OK on objection for voltage, it's officially not rated so much. The 6080 has been used at > 400 supply volts in ham stuff but that's off label. It also has a +/-300V heater to cathode voltage rating.

Aside from the heater power, the anode drop for a given current (30V@100mA) is low compared to the 6BL7 (200V@100mA) and it may compensate the power budget a little there.

Either way the probable 6-10 Watts drawn by the modulator screens is comparable with the heater wattage of either tube (10 or 16W), and either cathode follower tube has to run class A1 or AB1, in order to get away from cutoff/crossover distortion and make it easy to bias the modulator accurately so there is some idling plate current there, and that's a few more watts. For either tube, probably some negative supply through a cathode resistor to get it right (attachment).

For power all I can say is 2-3x whatever is required as "good practice" when driving class B-like grids with variable impedance. The voltage swing for the G2's of the modulator is more or less fixed, it only remains to supply the peak current without distortion. Power supply is cheap and many tubes are cheap and junkboxes are filled with both. Don't skimp on what you build because you might have to do it over later.

I have no accurate idea what the peak screen grid current would be but it will change upward with drive as the instant plate voltage lowers and the upswinging of this effect gets sharper and sharper as the Modulator is driven harder or the load impedance is lowered. Too many variables. Whom is building it will have to decide what to do.

But that's just my opinion and my way of doing. Most audio or RF circuits I build would not be cost-competitive for sale. I agree that there are many cost-down to be made, but the worst thing that will happen because someone used a larger tube is that it will last longer.


How about a separate filament transformer for the 6BL7?

Sure, if the swing excedes the heater/cathode voltage rating, a floating or biased filament transformer may be needed. Fortunately, if the 6BL7 doesn't cut it, a 6080 can be swapped in without rewiring the socket, as you've written, so it's easy enough to try them both.

It's my understanding that crazy drive requires less voltage swing than screen drive alone. So, maybe I get away with a 6BL7 as a CF and save 6W of filament power.

I plan to use MOSFET source followers for my initial trails to set a baseline of performance in my DX-100 test bed rig. For a future, new build (where there's more room) I'd be looking at the hollow-state CF possibility.

I'm a little confused. I think you mean 304TL output tubes. They'd work, if I could ever find them, but being triodes the aren't compatible with crazy drive which requires a sweep tube pentode. Push-pull-parallel sweep tubes can be used for greater power.

Opcom, I like your CQ article with the 6AS7 drivers.  Interesting use of the split secondary driver transformer, very similar to the way many broadcast transmitters interface from the 600 ohm line to the push-pull voltage gain stage.  The RCA BTA-1M series of transmitters do this, followed by an 807 for a cathode follower driving the 833s.

I got a basket-case Valiant (supposed to be complete/working) at the Huntsville hamfest.  It was so bad, with bent chassis, that I gutted the audio section and re-designed it to use a UTC S-21 modulation transformer, and the original modulation transformer (all windings in series) as a mod reactor.  Instead of the pair or 6146s, I put in four 1625s in push-pull parallel, with a 6SN7 cathode follower for a driver.  Another 6SN7 for a gain stage, preceded by a 12AU7 phase splitter and UTC A-10 for 600 ohm line input.  That worked out really well, great response, low distortion, but I wish I knew about crazy drive then.  This thread got me thinking about how I might resurrect my other BTA-1MX.  Maybe eliminate all the 807s and audio transformer, and redesign the audio from the ground up.

Sorry, I did not mean to confuse you.  Most of my post was straightforward.  But the comment about a pair of 304TLs as a driver for each phase was just my failed attempt at humor, as if a 6336 was insufficient.  I guess I better not give up my day job, comedy is obviously not my forte!

In any case, very interesting thread.  That article from CQ hit the nail on the head as far as cathode follower drivers are concerned!

Fellow member BEEFUS taught me that a low Mu tube might not be the best as a cathode follower. Here is why:

The output impedance can be approximated as:

Zo = 1/gm

The gain can be approximated as:

A = mu / (mu + 1)

Reference:  http://www.valvewizard.co.uk/accf.html

This means that a 6080/6AS7 in a cathode follower will gave a gain of about 0.66 whereas a higher Mu tube might get up to 0.90. This might not seem like a big difference, but getting a bigger swing of audio to overcome a 1/3 drop can be a lot.


Now if there was a tube with high Gm and High Mu that also had some peak current cathode emission capability.

Look at the 6DJ8....or maybe the Russian types D3A (triode strapped), 6n6p, or the 6C45P. Other more common tubes like the 6BQ5, and 6W6 (triode strapped) might have merit as well. The 6W6 is a sleeper here..itself a real candidate for Crazy drive.

Jim
Wd5JKO

ask and you shall receive.

I had not looked at the socket pinouts but that's interesting! When you do the experiment, can you please look at the modulator grid current in the crazy drive hookup? One circuit had just over 1K Ohm from the screen drive voltage drive to the control grid, then about 10K from the control grid to the cathode. It concerned me about the grid current being too high there.

I did, at first, laugh. Then, I wondered if it was a typo. I need to learn not to take life so seriously.

I'm not at all sure that the resistor values on that schematic are optimum. There probably needs to be some jockeying around of the ratio between the two and the total R of the two. Obviously, one needs to make sure all tube limits are adhered to when doing this. Hopefully this "tuning" process can be refined down to plug-and-play resistor values for any given tube type. It'll take some time.


6BL7 mu=15

A=15 / (15 + 1)=0.9375
Not too bad.

By the way, Patrick, your pass regulator chart made me totally loose it! 3CX3000F1 in deed. :o
Take a lesson, Rick. When exaggerating go way, waaaaay, over the top! ;D

Don

Don, I did not see that chart until after my post.  Then I got to looking at the tube numbers.  Something looked WRONG.  Opcom, 6080 - 6SA7 - maybe a pentagrid converter instead of a dual triode 6AS7.  I recall a professor long ago who would leave out some letters when writing on the board, then go back and fill them in - and he would say "Don't mind me, I'm Lisdexic".  Patrick, it is ok if you switch AS and SA, we know what you mean.....

So, I do have some 304-TLs.   They have enormous amounts of emission.  But yes, I do have a couple 3CX3000 and one glass 3X3000, going into my next AM transmitter.  The power supply is all finished, 5000 volts at 1 amp CCS, from a Collins FM broadcast rig.  Unfortunately, if I used the 3CX3000s as drivers, I would have nothing to drive, so I showed a bit of moderation in my post.  Not like the new transmitter under construction....  I can't hear you, will you please "turn up the wick?"

On a serious note, I think more should be done and documented on the crazy drive, as well as screen drive.  This may open up many options for clean modulation with simpler driver circuits, and it also just makes it fun and interesting.  Looking forward to seeing your documentation when you perform your testing.  Kudos!

It's an unusual grid that will draw 250 ma! (the article calls for 45 watts!)
I guess in that case a large tube will be needed, but that was for an 810 in PP for a kW!!
I seem to recall the 805/838 only needs less than 10 watts, for about 300 watts PP.

The project being considered is somewhat lower in power, so I'm back to smaller
tubes for the driver... :D

                             _-_-

There's whole lot of information on crazy drive in the DIYaudio.com thread, which I referred to in my original post. I spent a great deal of time reading that forum to try and get a handle on all of this. I learned a lot from it but honestly, there's still a whole lot of it that way over my head. There are people in that forum, and this one as well, that are true experts and I'll never be in their league. :-[

That said, I'll do my best with trying to make it work myself and contribute what I find. The responses that have been posted here are very encouraging and I'm glad it's generating interest.  I had hoped that better minds (than mine) would get involved, and this has been the case. I'm pleased (and relieved) that there are others moving forward on this.

I wish I could dedicate more time and money to this, but both are in short supply here lately. So I'm plodding along slowly, but my enthusiasm is unabated!

Don

Bear,
I hope you're right! Smaller is better.

I'm not sure how much voltage or current will be needed for the crazy drive method we're discussing here. It will be similar to screen drive, but the driver needs to work into the resistor network that constitutes the crazy drive circuit, as well as any drive needed for the tubes themselves. The resistor network will certainly draw current, but the tubes might only require voltage. I don't know. Until examples are built, tested, and evaluated, we won't know to a certainty how much drive is required.

My first attempt to build one will be using MOSFET source followers. With this, I should be able to make measurements that will answer the question of how much drive is required.

I hope that others here are working in parallel and can provide some insight after having run some tests.

Don

Don, we're into "class 2"  here, the screen and grid are going to draw current.
So, that makes it simple.

Look at the value of the resistor to ground... I didn't take note, that will be the
limiting current draw from the resistors. Everything else will be the tube. We've
got a good ballpark on what the tube's electrodes are going to require from the
tube manual and the results in the thread... the tube is still limited by the max
screen current no matter what. Or else it will look like a roman candle inside that
bottle! :D

But the series resistor is already limiting or more properly scaling the control grid
current, that leaves the screen's current draw to be considered in addition... now that
I think about it the fact that there is a resistor to ground from the control grid scales
the voltage, whereas a series resistor alone would limit the current... if I'm seeing
this properly. (ha ha)

Bear,
Yes, I see it as a voltage divider. The control grid getting a scaled-down voltage. I "think" that's the dominant factor. Hence, the ratio of the resistances is a factor in tuning for minimum distortion. If so, connecting the wiper of a potentiometer to G1 would make it easy if the overall resistance value can be approximated.

Certainly bias set point will matter too. Reading the work posted on DIYaudio, it is suggested that the idle current is relatively low using crazy drive. That's one of the benefits.

I've got a pair of 12GC6 tubes being shipped here that I want to try in the DX-100. I think they will be big enough to modulate it without massive overkill. The DX-100 will be my first crazy drive victim.

Here is my hand-drawn schematic for the DX-100 crazy drive modulator that I'll be building. The section shown will be preceded by a 12AY7 and a 6CG7 LTP.
Comments and suggestions are welcome.

Don

Don,

What is the expected DC voltage at the Mosfet end of the 68ohm resistor??

How are you biasing those mosfets, to what gate voltage??

Not sure, but I seem to be missing something or other...

Bear,

I've been asking myself those same questions. I don't know where the 12GC6 want to be biased. I'll have to determine it experimentally. The MOSFET biasing circuit is variable from about 2 to 25 VDC (to the gate) given the values shown. That will set the bias voltage for the 12GC6's. The potentiometer at the grids will be used to determine the proportional G1/G2 drive - tuning it for minimum distortion. The 4.7k value is more of a guess than anything else. I'm sure there'll be some smoke-chasing before I get this to work.

I'm uploading a revised schematic here.

Don

Just trying to accomodate K5PRO.. he uses some pretty big stuff at work smashing atoms and stuff. never know when an amplitude modulated particle beam might be useful!

Useful?
Useful for what - world domination and enslaving the human race???
Oh wait. Yes, I see your point.

Don,

The zeners on the grid may take it OUT of "crazy drive" and act like fixed bias.

If the 400volt supplies slide around, the bias will slide, how much will be the question.

Bear,

Those diodes are schottky's. Smoking-amp suggested them in one of his posts. I'm not sure I'll keep them in the circuit but I put them on the drawing anyway. We'll see.

Yes, the sliding bias is a concern due to the unregulated +/-400V. That voltage was chosen only because it's available in the DX-100 (I can get the negative rail by using a bridge rectifier). If it does become a problem, I'll have to regulate it or build in a separate dedicated supply. I'm sure I wouldn't need anywhere near +/-400V. If the current demand isn't too much, a couple of VR tubes may work with the stock supply - need to find that out.

There's a lot of testing to be done and measurements to be made. I plan to start by figuring out where the 12GC6's like to be biased with G1, G2 essentially strapped. Then I can work backwards to create a source follower to do the job. After that, I can look at how much audio drive is needed. I haven't done anything quite like this before so I'm sure I'll learn something the hard way. ::)

Thanks for your comments and interest, :)
Don

I am slow walking a similar setup. Looking at a 6BL7 (1/2 each side) cathode follower with a constant current source (CCS) instead of a pull down resistor to a minus supply. A good CCS won't care if the minus supply has ripple, or moves around in voltage.

On a somewhat different note, I ran into this W5BM (sk) article about screen modulating a 4-1000. Schematic attached here. He is using a form of Crazy drive on the 6550 screen modulator tube, and is using a 6GK6 as a CCS to the minus supply....still on topic.

The full PDF is too large to post here, but if anyone wants it, I can email it directly.

Jim
Wd5JKO

Jim,

I somehow missed your post for a few days, sorry.

I was pondering using an IXCP10M90S CCS for that purpose. I have no experience with them but wondered if it would work in that situation, (sure, a vacuum tube CCS can be made also, but it's a bit cumbersome and I've already given-in to using MOSFET source followers). So, you're validating what I had imagined, or guessed... that a CCS might work. I'm on the steep edge of the learning curve with much of this so I appreciate the help. The IXCP10M90S is rated for 1-100mA and 900V. I still don't know how much current would be needed so they may or may not be appropriate here.

I'm pretty much slow-walking too. I'd be interested to see a schematic of your 6BL7 CF w/CCS idea if you have a sketch.

I don't follow the W5BM schematic very well. It looks like maybe he's coupling audio to G1 through a .01 capacitor, not a voltage divider. I don't know. There's a 306 volt zener string between G1 and G2. I can't figure it out.

Don

I have no idea what his "Beta Network" is supposed to do. Makes some sense right up to
the diode in the feedback loop!

Looks like the audio is fed to the output stage via pins 8 & 3 of the driver tube.
The driver is "standing" on 306 volts to keep the filament to cathode breakdown within
limits...

Looks like the lower 6550 also receives audio to its grid from the cathode of the
input tube, via that 0.33ufd cap...

It's a bit unclear how this screen modulator is coupled to the screen of an RF tube, since
the circuit presumably swings between the -300vdc and +700vdc, making the quiescent
point likely not where the screen of a big RF tube wants to sit? So maybe cap coupled and
then a DC screen bias on the screen side of the cap??

An unusual approach?

I made some progress today...

The quiescent bias for the pair of 12GC6's is:

With G1 and G2 at 0 volts:
   DC Plate Voltage      800 Volts
   DC Cathode Current      50 mA
This yields 40 watts, about 114% of rated Pd.
That's too hot!

With G1 and G2 at -3 volts:
   DC Plate Voltage      900 Volts
   DC Cathode Current      10 mA
This yields 9 watts, about 25% of rated Pd.
That's in the ballpark.

Now I have a clue as to where to start. I'll have to set-up a crazy drive voltage divider and run this again. I'm guessing that I'll need to keep G1 near the -3 volt level and G2 will run closer to about -10 volts, depending on the division ratio. It looks like there will be some juggling involved to find the sweet spot - when I get that far.

Don

Pins 8 and 3 are the outputs of two identical cathode followers a 12AT7. There are two available inputs, the BNC jack and the jack from speech amp.

R1 is not shown except in the op-amp diagram. It would the the third input to set a DC level. In the tube diagram, the 20K 10 turn pot probably does this.

The control grid of the 6EJ7 is the (-) input of the op-amp diagram. It is the phase inverting voltage amplifier for all the voltage gain for the 6550.

The 12BH7 has a high plate voltage but also a large voltage drop and has to push those zener diodes around as well as the grid of the 6550.

The diode portion of the feedback network gives the modulator more gain when outputting positive voltage than negative voltage.
R4 in the network sets the overall gain.

C1-R6 set up the frequency response compensation via feedback and the overall gain at frequencies passed by C1. At some point the combined impedance of C1-R6 becomes more than R4 value and has less effect. at high frequencies it has much effect.

The lower 6550? the 6GK6? also has some gain and it is called a 'sink' there. So it is to shunt current to the -300V supply I guess, not having seen the whole article. As we might guess, a screen grid on a HV tube like the 4-1000 will, if left alone, gain charge and voltage rise toward plaste volts with bad results. This may be to prevent that, to eat negative screen current, or to just be the bottom half of ther totem pole output and do so without feedback, as it does not use the network.

That's the best I can answer for you without the article which I could not find.

They are not the same application but the feedback network in the screen modulator is interesting. I believe the 6550 would do well in the 'special class B' circuit, which is not crazy drive. Anyway being a modulator those circuits are not perfect for minimizing distortion and the modulated stage is hardly a linear load, so some feedback is needed for improvement in all of them.

Patrick,

I do expect that feedback will be beneficial in a crazy drive amplifier. I see feedback as being the last thing to apply to an amplifier after everything else has been done to get it right, using it to better the amplifier rather that cure its ills from the outset. So, as feedback goes, I haven't even been considering it at this point, but that time will come.

Along with modulated stages, I believe that loudspeakers are a non-linear load as well. While "perfect" is a good target to aim at, "practical" will inevitably win the day when the dust settles. But I don't mind getting my shoes dirty kicking up a little dust. ;)

Thanks,
Don

Well !
 This is a subject near and dear to my heart. The triode connection of tetrode tubes. rather than make an old buzzard transmission on the radio I will make a text presentation here to dispel untruths and old wives tales.
 My experince with triode connection goes back to 1966 when I was a boot. I had the basic build on the Crap Box 180 Originally a pair of 6146s modulated by a pair built on an old TV chassis. The plate voltage was about 700 volts. I had a copy of the  RCA transmitting tube manual. I saw the triode connected 807 modulator. I have a copy of RCA ham news or ham tips.One was GE the other RCA. This was dated from 1947. There was a whole write up on the glowing performance of the merits of triode connection of the 807. I fegured it would work well with the 6146 s . The audio driver was a pair of triode connected ( low Mu) 6BQ5s Plenty of audio power for the job. In triode connection mode the idle plate current was in the order of 50Ma. or so. I fed audio into the line level input from my microphonium preamp. It modulated but when I listened to it on my receiver it was loaded with grunge distortion. I was not happy with it. I decided to try a tube much like the 807, the 6BG6. Essentially an octal version of the 807. The idle current was about 15Ma. The audio driver did not quite have the peak audio voltage output to drive the 6BG6 s to full output. What I heard at lower modulation levels was the same type grunge distortion to a some what lesser extent. I gave up. I configured the modulator for the usual class AB 2 tetrode mode.It worked OK FINE !!!. I still was not to give up on the triode connection concept. I had started building my 4-1000 transmitter I chose a pair of 4-1000s as modulators in  RC coupled AB1. While all of the Ts were dotted and the Is crossed it would NOT make 100% modulation at low power(1500 volts) At medium power 3Kv it would barely make 100% mod. At full strap(6Kv) it made somewhat better than 100% positvive,if you could keep things from blowing up with that kind of voltage. Little did I realize that the problem was the ratio of the modulation transformer was the limiting factor. I had an 811 modulator deck from another transmitter. I pulled the plate caps off of the 4-1000s and placed them on the 811s  I made sure that the gear shift was in first gear I would image 6Kv on 811s would turn them into flash bulbs. To my surprise the 811s fully modulated the transmitter 100%. I disconnected the pair of 807s audio voltage amplifier and grabbed my HH Snot stereo amplifier and a good push pull audio transformer. I tried a few configurations. I tried tying the screen grid and the grid  in  in parallel at each tube. It appeared to work. I could make full modulation in first gear. I went into high gear to make some heavy duty changes to the modulator. As it is I had a 1500 volt power supply for the RC coupled 807 audio voltage amplifier. I also had a series screen regulator using a pair of 211 s as the pass tube regulator. I ran 850 volts on the screens of the 4-1000 s. I carefully diassembled the 807 audio voltage amplifier from it's sub chassis and built it on to the screen regulator chassis. The 211 s were replaced with a pair of 845 s . This was years before the audiophools discovered the 845.  I needed a driver transformer i I had an old UTC VM-4. I took the core apart and reassembled the core in the cross lamniated configuration. This it to obtain maximum inductance being that there is no appreciable unbalanced DC on the transformer. If you are familiar with that piece, Terminals7&12 to the 845 s 9&10 tied B+ CT. Secondary: 13 &14 to the screens of the 4-1000 s.2&5 are connected to the grids. driving the grids at much lower peak audio voltage.  3@4 tied. CT of secondary This is connected to chassis ground when transmitting. I used the bias voltage for the 845s about - 250 or so as cut off bias during standby.This configuration worked very well. No problem making full modulation. However I was getting comments about excessive band width. This was due to the poor performance of UTC vari- match modulation transformers. I went to an S 22-somewhat better. While trying to figure out the distortion issue I tried a few other configurations: Screen drive only Grids at audio ground , Grid driven,screens to ground. These schemes did not yield good results . I decided to try the RCA method. I came up with a pair of 20K 50 watt wire wound resistors. The modulator performed like it did with as I will call it the dual differential drive method. The RCA configuration suffered from the same grunge distortion the afflicts the 807 s. I went back to the differential drive method. Years later I came across a UTC 250 watt linear standard modulation transformer. This cleaned up the HF audio distortion.  In about 1977 after moving here to Skowhegan, Maine, a friend of mine had a Junkston Valiant that had a number of issues.  Chernobyl resistor melted down in a bad way. After cleaning up that mess I decided to work on the audio. I went overboard. I pulled the modulation transformer out, took the core apart and cross laminated it. In place of the 866 s after solid stating the HV supply I made use of the space for a 10Hy 300Ma filter reactor as a modulation reactor. I decide to triode connect the 6146 modulator tubes. I had a 15 watt modulation transformer that had two sets of primary taps as well as the full primary connection. I connected the full primary to the screens and the set of taps colsest to the CT to the grids. I used a single 6550 as the audio driver tube utilizing the secondary of the 15 watt modulation transformer as the primary. I fired up on 75 expecting to get glowing praises as to how good it sounded. I was crest fallen when I was told that the grunge distortion levels were prominent. I listened to it on a good broadband receiver. It sounded like crap. Surely some mistake was made in the speech amp. I pulled the 6146 mod tubes out and took an audio output transformer,hung it across the screen pins( full driver transformer secondary)and connected it to my large speaker system I had hanging on the wall. I patch an FM radio tuner into the line level input. The audio was clean and robust. So where is this distortion coming from ? I grabbed a 6.3 volt 20 A filament transformer, A pair of 4 pin sockets and a pair of 811 s. I JS ed it together at the side of the transmitter and tied the grids of the 811s to the taps that went to the grids of the 6146 modulators for the right step down impedance. I key up the transmitter. The audio was very clean and green. Hmmm I said. I had a pair of used TV sweepies ,a pair of 6KD6s. I happened to have a pair of 12 pin sockets kicking around as well. I hooked up the 6KD6s in differential drive method like I did with the 6146s . The grunge distortion was WORSE. The idle current was about 150Ma so I did not want to have things key down for any length of time. I tied the screens to ground and just drove the grids. The audio was very clean but low modulation because the tube would not pull much more current with audio. Finally I grounded the grids and drove the screens . That was the magic bullet. The 6KD6s worked every bit as well as the 811s. I then addressed the high idle current. I found a  Radio Shack multi-voltage wall wart. I lopped off the DC connector . I tied the + to ground and the - to the grids. About -6 volts brought the idle down to 60 Ma. I operated with the transmitter for a while. I returned to the 6146s with screen drive only. It worked it was clean. What I noticed is that it had trouble making full modulation. I figured I had spent enough time on it. Overall it sounded good so it went back to it's owner Jack, now W4PPT, the founder of the Boat Anchors list. Shortly thereafter P.W. Fallon,WA1IWQ was doing up a Valiant his way .He liked the idea of the triode connected 6146s. He was probably the first person to build up a solid state audio driver . A pair of LM 383s in bridge mode. I gave P.W. a push pull audio transformer to couple to the screens of the 6146s He used a UTC S 21 120 watt mod transformer. Much better suited than the stock transformer. It was set up in the 3:1 impedance step down ratio to deal with the low impedance of the modulated stage in the Valiant. While it sounded great he could NOT make full modulation.The solid state audio driver bucked and farted to try to make full modulation. P.W. had a Dynaco MK II 60 watt audio amplifier. He used that to drive the 6146 s The bucking and farting was gone but it STILL would not make full modulation. A new pair of 6146 s was placed in the modulator. No improvement. We were both at an impasse. While PW can up for a visit I told him how well the 6KD6s worked. I had a pair of 6CD6 s . Basically 2.5 amp filament tube. About twice the peak emission of the 6146. The 6CD6 is like a lighter version of the 6DQ5. P.W. wired them in and biased them correctly.Now the transmitter easily made full modulation. The solid state audio driver behave very well after that. I puzzled over the reasoning with the difference between the two sets of tubes. Certainly the 6146s in tetrode mode make 120+ watts of audio. Why not when in triode connected mode ? I gave it a lot of thought. I may be right or wrong here. Beam power tubes which we are using are beam power tubes due to the fact that the screen grid is wound at the same pitch as the control grid. This is to get maximum gain and minimum screen current draw. In the triode connection ,the screen being the only driven element, the control grid is a useless appendage. It is analogous a lunar eclipse with the shadow of the grid limiting the potential emission of the cathode. As it is many others have used the 6DQ5 in the Valiant and other transmitters in triode connection mode. I have a viking Valiant that I did up for a friend in 1984. I used a pair of 6KD6 s in the modulator and a Chicago CMS 1 250 mod transformer stuffed into it. I eventually wound up with the transmitter. I used it as a 100 watt class daily driver until something crapped out in the power supply. I have used it for
 better than 10 years before it died. The game plan over the Winter is to be able to extricate it and do some research on differential drive to see if there is anything to it-see last part of this posting.
We revisit the 807 triode connection method. I had acquired a home brew modulator with a pair of triode connected 807s and an ART 13 mod transformer a 6BL7 audio driver using a UTC S 9 push pull plate to push pull grid driver transformer . This transformer has several sets of secondary taps for a number of turns ratios. I fixed up the lower level stages. Added bass and treble tone controls and decided to try it out. I did the audio output transformer hookup to the big monitor speaker on the wall. The FM tuner was patched in. The grunge distortion at low signal levels was apparent. I noticed a strange phenomenon . As the audio level is slow increased the 807 plate current went DOWN instead of up. This was when the distortion was most evident. As the levels were increased the distortion went down along with a greater plate current draw. Even at a good loud level from the speaker there was always an underlying distortion. After examining the  S 9 transformer. I took the secondary taps closest to the center tap and connected them to the grids of the 807s therefore eliminating the 20 K resistors. That did it ! Differential drive works well with the 807 in triode connection mode. I did the same thing many years later when I built the 50 watt SBE rig. I am using a pair of triode connected 807 s with 1000 volts on the plate and an ART 13 mod transformer. I used a UTC S 8 and a single 6V6 as the audio driver. The S 8 has a number of secondary taps like that of the S 9 . This arrangement also work very well.  Even though it has become standard procedure to just use screen drive I am ov the opinion that differential drive of big TV sweepies will yield more gain and perhaps better liearity. However recently Steve WA1QIX stuffed a pair of triode connected 6DQ5s in a Ranger owned by Mike N1HXA. A Stancor A 3893 60 watt mod transformer was used. A direct coupled FET audio driver was used. According to Steve he has researched the 6DQ5 s characteristics. In screen drive mode it is VERY linear. Being that there is no driver transformer used it is possible to wrap large amounts of negative feed back around the whole modulator to obtain superlative audio performance. While the use of a conventional type driver transformer circuit is very reliable and fool proof it can't quite measure up to a direct coupled driver scheme. The driver can also be hollow state based perhaps using a modification of the cathode follower screen modulator circuit used in the DX 60 and other transmitters of that type. Using two 6DE7 tubes with a phase inverter ahead of them to work them in push pull would probably make a very good driver.  A 6BL7 would also make a good push pull cathode follower with a push pull audio voltage amplifier ahead of it. I have not worked out the circuit details yet. A last observation: sometime back Terry W2PFY brought by a Sonar SRT 120 transmitter. A single 5894 with both sections in parallel modulated by a pair of 6L6s in triode connection like that of the 807. The audio driver consisted of triode section of one of the lower level stages into a small 1:3 audio transformer to a 12AU7 as a cathode follower to drive the screens of the 6L6s. The mod transformer is a big Triad unit. I did the following Used a much better 1:3 audio transformer and upgraded to 6CG7 dual triode as the cathode follower. A 12BH7 is a drop in replacement. I did not have one available at the time. The 6L6s are numb with just the screens driven. I went to a pair of 6Y6 s .Same pin out as the 6L6 but a 1.2 amp filament. This tube has similar peak emission to a 6146. The configuration worked quite well. Problem being that cathode follower was positive biasing the 6Y6s which were pulling too much current. I needed a lower resistance return to ground. I decided to use a Hammond push pull plate to 150/600 line output transformer. I used a resistor from the primary center tap to ground. There is bias voltage in the SRT 120. This worked very well. I later tied the secondaries in series I tied the center tap point to ground and each side of the total 600 ohm winding to each grid. The audio did not appear to have any grunge distortion. The power sensitivity improved. For the hell of it I reversed the lead to the grids to observe how it would work. The audio went to zip. To me this proves that I hit upon the correct level of differential drive to high gain tetrodes. As it is Terry hardly ever used the transmitter. He left it on at his basement work bench. He was greeted by the stench of burning power transformer. It stank up the house. His Wife insisted that he get rid of the transmitter. I wound up with it. It is in my collection. I replaced the power transformer with an old black and white TV power transformer. The transmitter makes about 25 watts .
In closing I hope this long old buzzard posting will shed light on the best ways of triode connected big TV sweepies and other tetrodes . All are a bit different. The 813 for example is utilized by tying the screen and grid together and driving it with audio. There are many possibilities with other tubes.
Christmas day-Tim WA1HnyLR

Tim,

Thanks for sharing some of your 50+ years of experience experimenting with modulator design. You are a font of knowledge on the subject and many have benefited by it in your on-air discussions.

When I first noticed your posting here, I worried that it might be a very, very, long one. Thankfully though, you kept it down to only one paragraph. ;D

From what I've read in the DIYaudio forum, where many hours of research have been done by top experts in the field of audio amplification using vacuum tubes, proportional drive of the grids of sweep tubes will yield linear amplification heretofore unknown in the realm of tube-type audio amplifiers. I intend to implement what I've learned there, in a modulator which I will briefly describe here.

The difference between screen drive and proportional 'crazy drive' is a resistor from G2 to G1 and a resistor from G1 to ground. For a push-pull amplifier that's only four additional resistors to accomplish the task.

However, I'm going to be fancy about it and use a dual potentiometer. One end of each pot will go to G2, the other ends going to ground. The wipers will go to each tubes G1. G2 will be directly driven by a source follower.

With this, I can tune the ratio of drive from zero (G1 at ground, just like conventional screen drive), to a point that causes the overall gm of the tube to be even more linear than with screen drive alone. At the same time, this should reduce the drive voltage requirement. The increased gm linearity will permit the idling current to be lowered as well, increasing power efficiency. This is the hope.

Rather than scratch-build a transmitter and modulator, I've opted to modify a Heathkit DX-100. I've replaced the 1625 modulator tubes with a pair if 12GC6 sweep tubes which will be driven by a pair of STP7N95K3 MOSFET source followers. This is similar to a portion of the schematic that Steve, WA1QIX has posted, elsewhere in this forum, of a modulator that he designed and built for a Johnson Ranger transmitter - sweep tube modulators driven by 'FET source followers. The similarity ends there, as I will explore the proportional drive of the sweep tubes and use different line amp stages.

I have been moving forward on this project as time and funding become available to me. I'm following the learn-as-I-go building method on this project rather than amass a pile of paper notes that, in the end, never get built.

I greatly appreciate the members of this forum who have taken the time to follow and post here. It's my hope that what we find here together will be of benefit to those, like yourself, who enjoy building their own AM equipment.

Thanks again for your contribution!


Don

Here's the latest drawing... for those of you who are following along. ;)

Don

Edit: I think I finally figured out how to get the image correct.  ::)

That looks great including the dual pot.
You could always label the knob for the dual pot on the grids "Linearity" or "English" or something.
It needs a name - you ought be the one to name it if you make it work.
Or use an audiophool mystery name like "Differential Mu Compensator"..  ;D
Illudium Q-36 Modulator?
20K to GND -may dissipate some peak power 660mW? in the pots. Might pay to try to use 2W parts if they can be had.
I Like what you did.

Phil,

I had ordered a dual 2 watt pot from Hong Kong - still waiting for delivery. It didn't even occur to me to "name" it. You proffer some interesting suggestions. :)
It's just a standard "Manually Operated Dual Carbon Rotating Armature Potentiometer." or "MODCRAP". With that in mind, I'll update the drawing with the label "Set Linearity Under Test". ;D ;D ;D ;D

Don

Don,

   A few talking points:

In your schematic, the 12AY7 phase inverter needs to have the right side tube's grid at an AC ground, NO?

In some if the DIY Audio forum notes on cray drive, one recommendation kind of stuck out. That was to bias G1 negative, and G2 positive. The idea was to set the resistive divider ratio such that G1 peak voltage hits zero volts when G2 is at the positive peak. This keeps grid current from flowing into G1, and therefore the resistive divider ratio stays constant with drive level. Might need something like -30v on G1 with +60v on G2 (or thereabouts) to establish the desired idle current (maybe 10ma cathode current per tube?).

In other DIY audio discussions about G2 drive, it is mentioned that some sweep tubes can get the plate voltage minimum as low as 40v when the G2 voltage is at the Maximum peak (maybe about +150v peak). The screen then becomes the plate and therefore the screen current soars. There are mentions of the need to limit the screen current to protect the sweep tubes from a catastrophic destruction of the screen grid. This could be as simple as a drain resistor added on the FET cathode follower. The caution exists if excess drive overload happens too often, or real quickly if the overload is sustained (i.e square waves).

Jim
Wd5JKO

Jim,

Yes it does. I totally missed that - added it to my next drawing revision. Thanks.

Re: Biasing G1 negative and G2 positive.
I hadn't seen any mention of that in my readings. Not sure how to impliment it, or if it's going to be needed. I just don't know. I'll look at this when experimenting.

Re: Limiting screen current.
I'll add the drain resistor to offer some protection. Fortunately, the tubes I'm using, 12GC6, are inexpensive and readily available so I can make a few mistakes. ::)


Obviously, you've been reading the audio forums on crazy drive. That's a good thing! I'd be lost going this alone. Thanks for your assistance.

And you've been helpful to me in ways you don't realize. While searching this forum last night as to why my DX-100 can't make more than 80% modulation, this posting you made, almost 6 years ago, proved to be very informative.
http://amfone.net/Amforum/index.php?topic=33195.msg258713#msg258713

Thanks again!

Don

At this point...

I have the 12GC6 sweep tubes in the DX-100 modulator. With G1 grounded and driving G2 with a transformer, and -7.7V bias, the cathode resting current is about 20mA.

Driving it with an external amplifier, the audio sounds pretty good through this lash-up but positive modulation won't go above about 80%. Negative modulation can easily white-line the baseline. It's got plenty of power! Swapped input phase - same thing. Swapped plate cap leads - same thing. At 100% negative it's only 80% positive. Increasing the drive level, I can make the negative clip hard at the baseline but the positive just sits at 80%. It's weird looking on the old Eico 460 'scope.

It seems like the problem is at the RFPA screens. I have the original 20k screen dropping resistor in there. I'm going to increase that to 30k (as is often recommended to lower the screen voltage) and see what happens.

On the DX-100 there's a 500/2300 ohm tap on the 2800 ohm modulation transformer's secondary. As a lark, I may tap-off it with the 30k dropping resistor. That gives me the option of 18% or 82% modulation for the screens (relative to 100% on the plates) depending on how I wire it. This might prove interesting.

Either way, I need to get the rig modulating properly before I attempt any crazy drive setup. It should make for a fun weekend!

Don

Well...

After spending a few hours going around in circles trying to figure out why the DX-100's modulation envelope displayed on my old Eico 460 'scope looked so weird, I schlepped the HP 'scope up from the basement to get a second opinion. Looks much better with the HP - making 100% positive modulation now. The Eico's fate is sealed.

Next step will be pulling out the amplifier tubes from the Eico and going straight to the deflection plates with RF and audio for a trapezoid pattern modulation monitor.  It's probably the best way to get an accurate measurement.

You know, I think the old girl would look good sitting next to the DX-100... and so will the Eico. ;D

Don