• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Proposed screen-drive amp design... comments, please?

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've been thinking, lately, about trying to build a screen-drive amp, after reading some of the results that tubelab and others have gotten, from stuff like 6AV5s delivering up to 80w per push-pull pair...

I went looking around in my stash, and I actually had 5 12AV5s... but while searching, I also discovered a stash of NINETEEN NOS 6JB6s. Hmmm...

So, off I went, trying to figure out how to use them. My first stop, was looking at the schematic for the Berning EA230... kind of a standard starting-point for many screen-drive designs. My thought was, though- instead of a Williamson front-end (voltage amp, cathodyne phase splitter, intermediate gain amp, feeding a cathode follower which drives the output tubes), why not try to incorporate a Mulllard-style (input gain amp, LTP, cathode follower, output)? The potential ability to remove not one, but TWO sets of coupling caps between stages, was pretty appealing.

Through the help of folks in a discussion about this over at AudioKarma, I've gone through several design iterations... getting rid of bone-head errors. :D , working to improve the phase splitter symmetry, and trying to get enough output tube drive, without going to solid-state components (call me a hopeless traditionalists, but I don't want any silicon in the "direct drive path" of the amp, unless it absolutely can't be dispensed with).

I originally thought of using a pentode front end (and I may still, if it looks to be significantly advantageous), but as of now, I'm thinking of using a 12AT7 in the front end... just much easier to tune (no confusion trying to work with screen grids, that kind of thing).

I've been looking for parts for this, that would be feasible for construction- so far, I've found an Edcor output transformer ( the CXPP100-MS-5K) that should work- I'd think I could at least get 60w output from the amp, without problems, and maybe as much as 80w, with this iron. For main B+, I'm thinking Hammond 278CX (465ma/400-0-400v), one per monoblock. I've also located some cheap small power transformers that would work for the lower-voltage B+ and B- rails (+ and - 250V), and some chokes... it's going to be a choke 'n' transformer farm, but fortunately, none of them are that expensive.

My main questions now are:
-How would I go about calculating the proper values of the resistors and caps in both the inner and outer feedback loops?

-What would be reasonable tunings for the bandwidths of the first two stages (input amp/LTP) for maximum stability (now, the input stage has about a 600Khz bandwidth feeding from a 10K output-impedance preamp driving the input, and even more bandwidth feeding from a lower impedance preamp- it drops to 200K or so driven by a 30K impedance- is that enough, for one example?)?

I would also like to hear from anyone, about any other potential problems I might be inadvertently setting myself up for here, or any ways I could simplify the circuit while improving it... any comments welcome!

BTW: All resistor values in the main circuit are "good estimates", gotten through modeling each stage using TubeCad, to set bias and operating points. I realize that many of these values will change, somewhat, in practice... I would also like comments from anyone, if they think any of these operating points could be better-set, of course!

Thanks in advance!



  • screen amp 3-10-08.jpg
    screen amp 3-10-08.jpg
    32.9 KB · Views: 1,230
The 6JB6 is a good candidate. What you'll need attention to is gain and linearity in the input and driver stage- as it's set up now, I would guess that both are inadequate; figure you'll need an open loop gain of at least 400 (52dB) out of the driver. You're running the 12AT7s with a very low plate load which means low gain, high distortion.

For the 12AT7 LTP, 200-200 V. on the plate and IB = 3 mA. works WELL. 570 V. of B+ would allow the use of 120 KOhm load resistors in the LTP.

A 6GK5 as the voltage gain triode will provide more oomph and less distortion than a 'T7 section. Both mu and gm are higher in the 6GK5 than in the 12AT7, while RP is lower. ;)

IMO, "sand" can be a very good "friend" in this project. Cascoded DN2540s in the LTP tail CCS allow the use of a much "shorter" negative rail, while enhancing performance. IRFBC20s in the DC coupled driver positions along the lines provided in MOSFET Follies will do very nicely too.
Joined 2004
You need a high impedance in the tail for better coupling, also lower capacitance for better coupling at high frequency....

I agree a pentode would be an improvement but you could get a much higher impedance in the tail from a transistor cascode CCS than from any tube. Even if it made no audible difference, the practical advantages alone of an SS CCS would be worth it.

As Eli said, you could use a better tube that a 12AT7 for the first stage. A pentode would give you a useful open-loop gain - maybe EF86 or 6AU6. You might find it necessary, also, with that global NFB loop, to control HF stability by adding a zobel network across the plate load of the first stage.
SY said:
The 6JB6 is a good candidate. What you'll need attention to is gain and linearity in the input and driver stage- as it's set up now, I would guess that both are inadequate; figure you'll need an open loop gain of at least 400 (52dB) out of the driver. You're running the 12AT7s with a very low plate load which means low gain, high distortion.

I went and compared my CURRENT gain structure to that of the Berning EA230... and it's actually 6 dB MORE gain than the original amp... I actually went and reduced the gain a bit on the first stage, as it was MUCH higher than this (which was compromising open-input bandwidth, of high-impedance source). The input 12AT7 now has about a 13K anode resistor and a 600 ohm cathode resistor, in round terms...

The input stage of the original Berning design (which also used 12AT7s, just under lower current- 1.2ma as opposed to about 4ma for mine) had a gain of about 11, while mine is about 11.9... about .75 dB more gain. I originally started out with a gain of nearly 30... which turned out to be FAR higher than the original (29 dB as opposed to about 21 dB for the original). I couldn't get an open-input bandwidth matching the original (about 90KHz, with open input, the only input loading being a 100K resistor to ground paralleled with the input jack), until I reduced the gain to that point.

The phase splitter/voltage amp of the Berning had a gain of about 24 (about 27.6 dB, all in the voltage amp between the concertina and the cathode follower- the concerina has no gain of its own), while my LTP with CCS has a gain of nearly 45 (33dB)... about a 5.4 dB increase.

Comparing the cathode follower sections... Berning's 6SN7 and my 6BL7 both have about -.75dB gain... about 3/4 dB below unity. No real difference there, other than the ability of mine to handle MUCH more current supply to the screens of the output tubes (just over 1/3 the output impedance of the original cathode followers).

So, for my input stages, if you add the gains (21.5 dB input stage, 33 dB phase splitter, -.75dB cathode folllower), you get a gain of about 53 dB, at the grid of the output tube, open loop, referenced to the input.

As for low gain, high distortion... I'm a bit confused there. I have been using TubeCad to set the DC bias points of these tubes- and it keeps insisting that I'm actually using TOO LITTLE plate current... and that it wants the plate resistors to be even SMALLER numeric value (and even LESS gain). I'm already running the tubes at an average of THREE TIMES the plate current of the originals... but I'm still well below the max current for the 12AT7 (6 ma for the input stage, about 4ma for the LTP main tubes and 8ma for the CCS- both below the max normal rating of the 12AT7, which is either 10ma or 15ma, depending on which data sheet you look at...

AFAIK, the 6JB6 and 6JN6 have about the same screen gain... they both have the same maximum plate voltage and screen voltage ratings. and the graphs look almost identical for the plate curves.

So... even considering that I'm trying to squeeze about 3 dB more output out of the thing, that leaves me with 3 dB MORE gain to use in the feedback circuit, than the original design. I don't know if people thought that the Berning had inadequate feedback originally, but at least my design has a LITTLE more...

I appreciate the comments so far... please keep 'em coming. I think I'm finally STARTING to wrap my head around some of the more esoteric aspects of the design (as in how to predict current draw and loading of the PARALLELED cathode loads of the cathode resistor and the grid of the 6JB6 output tube, on the 6BL7- how to roughly draw a load-line for a "stepped load" depending on voltage, for one thing). But, there's still several things that I'll probably be asking about, in the next couple weeks!!

Eli Duttman said:

For the 12AT7 LTP, 200-200 V. on the plate and IB = 3 mA. works WELL. 570 V. of B+ would allow the use of 120 KOhm load resistors in the LTP.

A 6GK5 as the voltage gain triode will provide more oomph and less distortion than a 'T7 section. Both mu and gm are higher in the 6GK5 than in the 12AT7, while RP is lower. ;)

I tried using larger plate resistors on the LTP... but kept running into bandwidth problems. On the original Berning design, they had about 350K bandwidth on the phase splitter/driver amp stage (concertina plus grounded-cathode driver amps); I managed to get up to around 360KHz out of the setup I have now.

As for the current through the 12AT7s... it's currently at about 3.8ma on each of the LTP main sections, and double that (7.5ma) through the CCS...

As it is now, this is the ONLY alignment between the input amp and LTP phase splitter that I could reasonably find, that gave gain similar to the original (6 dB more, in fact), sufficiently higher voltage swing (about 40% more than the original, commensurate with me wanting 60w output as opposed to 30w for the original Berning design) at the output of the LTP, and similar bandwidths to the original stages as well, while at the same time giving the ability to "voltage-match" the DC plate voltage of the input amp to the needed grid (bias) voltage of the LTP top tube (in order to run it in direct-coupled "Mullard-style")...

I would still like to hear any comments anyone might have, about methodologies of calculating good ballpark ranges for the feedback loop resistors and caps. I would also like to hear anyone's thoughts on potential values for a input-amp Zobel around the plate resistor as well... I had thought of that kind of thing, but didn't know exactly which way I wanted to go...


Gordon, thanks for all the pertinent comments. I've never cloned an EA230, so I watch your project with great interest.

As for linearity questions, there are flashing lights on your screen which read, "Be suspicious of this sim!" Take a real 12AT7 and set it up as a voltage amp. Take into account the effect of the feedback look on the local feedback in that stage (usually, it overwhelms the local feedback) when choosing biasing. Measure the output spectra with sine wave excitation. You'll find pretty quickly that you need to run them with much higher plate loads to maximize linearity.

If you find something different than that, please post!

Finally, do not be afraid of using MOSFETs to drive the screens. They work significantly better than cathode followers in that position (source impedance). That's what I've done for my last two amps and I'll never look back.
I've taken some time to re-adjust some of these values... some before reading the comments here, some afterwards. I will definitely keep Sy's admonition in mind as far as loading is concerned... but, it could wind up being a PITA to increase the plate resistors, while keeping the voltage arrangement between the first stage plate and the LTP grid in the right range, to bias the LTP correctly...

However, those problems CAN be solved... it's just a matter of learning to artfully juggle... :D

Here's my current take on the schematic... I'd appreciate any advice on how to set a preliminary Zobel value for the plate of the input tube, as well as any suggestions for starting values for feedback loop resistors and caps...




  • screen amp 3-17-08.jpg
    screen amp 3-17-08.jpg
    49.2 KB · Views: 937
Keepin' on pluggin'...

I decided to make life easier on everyone... I found a set of tube schematic stencils, and made the schematic MUCH neater (see below!).

I also have changed the input stage, for higher-value anode and cathode resistors... Sy, is this more along the lines of what you were thinking?

I managed to preserve my voltage ratios, while reducing the loading... the input 12AT7 stage is now idling at about 2.1ma (up from 1.2 in the original), while the LTP is about 3.8ma per section. Both the input stage and the LTP have about 205-210 volts across the tube at idle...




  • screen amp 3-19-08-redraw.jpg
    screen amp 3-19-08-redraw.jpg
    78.2 KB · Views: 940
25k is mighty low for an anode resistor. Very high second harmonic.

I went back and looked at the EA230 schematic. A few things struck me. First, the feedback from the output tube plates to the driver cathodes is miniscule. I think it helps bias that stage while keeping the cathode resistor low, but the AC feedback seems almost nonexistent.

The overall feedback of that amp also seems extremely low. I haven't seen one in years, but I'll bet the distortion figures are not impressive.

In any event, you'll need a gain of about 100 from input to the output tube screens just to get to max power without feedback. Your first stage looks like about 35, the second stage about 20. So you've got 10dB or so to burn. Unfortunately, the output tube has a gain of something like 6-8dB! And if you burn that, the driver will have to swing several hundred volts. So you really need to include the driver in that loop.

If it were me, I'd linearize the input stage so it doesn't need feedback (for a 12AT7, figure 2mA current, 150k plate load, and a correspondingly bigger B+ rail), then run symmetric feedback paths from the transformer secondary to the driver tube cathodes.
You know, I have been puzzling through the EA230 schematic some more, and I'm more and more thinking that the loop from the output tube anode to the cathode of the driver tube, is actually supplying POSITIVE feedback.

The driver inverts polarity, the cathode follower preserves it, and the output inverts it again. So, since the cathode of the driver tube is in SAME POLARITY as the grid, and with the two inversions, the plate of the output is in the SAME POLARITY as the grid and cathode of the driver tube... putting drive from the output tube anode there should INCREASE the gain, unless I'm missing something. Almost sounds like a Stuart Hegeman type trick... my HK A300 has that sort of thing going on, to increase the gain of the driver stage...

I'm almost tempted to go back to a similar arrangement as the original Berning input stage (47-68K anode resistor, about 2.2-2.7K cathode resistor, or as close as I can get and still make the voltage work to bias the LTP properly) and see what happens... or, maybe even better- if I can linearize the input stage as Sy suggested... even if it means I need to use a negative voltage instead of ground on the bottom (I have a negative rail available, might as well use it if it's beneficial), to get the stage linear on its own... maybe that might be a way to go...

I'm wondering why the 12AT7, with three times the transconductance of the 12AX7 (about 5000, the way I'm using it, as opposed to about 1600 for the 12AX7) and ONE FOURTH the output impedance of the 12AX7 (about 15K as opposed to about 65K or so for the 12AX7, IIRC), would need such a high-resistance anode resistor. 12AX7 circuits work just fine with 220K and even 160K anode resistors... I'm wondering why a proportionally smaller anode resistor (to the transconductance ratio and/or impedance ratio) of something like 39-68K shouldn't work out?

No, it's negative. It's returned to the cathode of the driver tube, so the signal is not inverted at the driver plate. That means it's not inverted at the output grid, hence applied in a plate-to-grid (inverting) sense. But there's no more than a dB or so at AC.

Set up a 12AT7 in a test jig and play around with plate loads. You'll find pretty quickly that it needs to be much larger than you're proposing. A hint of this can be seen in the datasheet curves for plate resistance versus current- the closer you get to constant current, the more linear. 10x rp is not unreasonable.
Ah, yes. I had to start thinking of the Berning voltage amp in terms of how the FEEDBACK loop drives it- it's basically a grounded grid stage, when you think of it that way (driven from the cathode). And grounded grid stages are, naturally, NON-inverting...

I guess it might be time to start thinking about how to apply feedback to the cathodes of the phase splitter... though, I can't get my head around solving the apparent problem of what happens when the feedback interferes with the drive for the negative side of the phase splitter... seems like it would interfere with the balance of the stage. Though, with the adjust-ability I have with the anode resistors... that might not be a problem, maybe?

This stuff is a bit esoteric sometimes... takes more thought than would first seem!!

OK. I think I may have found a bug in TubeCad.

I have been modeling my LTP with tube-CCS, compared to the SAME LTP without CCS... and TubeCad keeps telling me, that with the SAME anode resistors and same current, that the CCS version has almost TWICE the gain of the non-CCS version... 40 vs about 21, with around 36K anode resistors.

Is there anything so "magical" about the CCS that would cause this sort of gain increase? I know it's more linear than a resistor... but if anything, shouldn't an 'infinite resistance' like a current source actually DECREASE gain a bit as a cathode "resistor replacement" (since gain is generally set by the ratio between anode and cathode resistance)?

Sorry if this is a bonehead question, but I'm questioning EVERYTHING at this point! :confused:

SY said:
Nope, that isn't right. What CCS loading of the LTP cathodes will do is equalize the gains on each side. The side with lower gain will increase slightly, but nothing like 21->40, more like about 20->21.

That's what I thought... it's just basic common sense, in essence.

Thanks for the confirmation... I was briefly being sold the Brooklyn Bridge by the sim program, I guess. :D

On that tack- I'm thinking about re-visiting the 6GH8A as an input/CCS tube. Gain on the triode part is 46 (I think I should be able to make that work in place of the 12AT7 section, in the CCS for the phase splitter), and the pentode can probably give me enough gain on the first stage while maintaining sufficient bandwidth to keep good stability. On that subject- anyone got a favorite high-gain (at least 40 or so) grounded-cathode alignment for the pentode part of the 6GH8A? Preferably, with about 200v across the tube at idle, if possible... that would make my phase splitter happy... :D

I definitely need to get more practice at doing these curve/load-line fittings for pentodes... just haven't gotten it into my brain completely yet, I guess, cause it just seems to take forever, as of now...

Thanks for any help!!

I'm keeping on trying...

I decided I HAD to do something about the front-end gain. I was having to ask way too much gain from the LTP, to get enough good voltage swing. Also, this was impinging on the bandwidth of the whole amp... the phase splitter was the "bottleneck", there, as well.

So, the question became: HOW could I get a LOT more gain from the first stage, without throwing everything off? I did a lot of modeling of various topologies, with various tubes.

Also, in the phase splitter, I wanted to find more voltage swing... so, I was off looking for any ways to get that, too. After a lot of research, I found a neat, somewhat obscure but easily available, tube, that gave me quite a bit more voltage, with the same B+, without having to run outrageously large or small current levels.

So, here's what I've got, now:

Input stage: 6N30P, in cascode. Almost INFINITE bandwidth (over a megahertz in open-loop!), with a gain of nearly 90!

Phase splitter: the smaller triode section of the 6CS7. 500v plate voltage... not as much gain, but with the input stage as it is, it didn't need much gain here!

Cathode follower/output driver: Larger triode section of the 6CS7. This tube was made for the same duty as the 6BL7 I had spec'd before, and is capable of enough current (60ma continuous, IIRC) to do ANYTHING the output tubes might demand. Gain is pretty good- about .9 (a tad lower than the original .95, but good enough to work fine).

Output tubes: 6JB6- same as before. I was checking curves and the original Berning EA230 design, and it looks like they're getting a gain of around 3 to 4 out of these... which should work fine, with the rest of my gain structure.

With this setup, and mimicking the Berning input requirements (about 1.4v for full output), I get an open-loop gain of between 1900 and 2500, and a needed closed-loop gain of about 283. That gives about a factor of 6.7 to 8.9, for feedback... between 16-19 dB or so. I'd think that should work OK...but I'd like to hear any comments or suggestions, thereof...

I really think this is a much more 'sensible' design than the others... it requires much less power supply complexity (the only non-normal thing is a negative rail of about 200v, which shouldn't have to carry more than a few watts of power consumption) than before... I will probably still use a secondary transformer for the 350V and the -200V, but the other rails will ALL run off of the SAME 550V main B+...

As I said, I'd really appreciate any continuing comments... thanks in advance!!



  • screen amp 3-23-08.jpg
    screen amp 3-23-08.jpg
    76.2 KB · Views: 637
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.