'Zener Mod' on 12AX7 Gain Stage

[Ray Ivers]

I finally freed up some time to try this mod on the Marshall MV preamp circuit below:



I really liked the sound of this circuit - kind of a combination of the best characteristics of SS-diode and tube distortion.

Here are some observations:

1) The coupling cap (C23) value can be made as large as you want. I settled on the stock .022uF, but I tried values up to .1uF. Without the Zener - and with the gain dimed - chords using the .1uF value sounded like fog horns with your head wrapped in wax paper but with the Zener in, increasing the coupling cap value just added more low mids and 'thickness' on single notes while chords stayed defined. Without the Zener circuit, I had to reduce C23 to .001uF to maintain chord definition, at which point single notes sounded very thin and trebly, and a lot of sustain went out the window.

2) I liked symmetrical clipping the best. In this circuit, that happened with the 5.1V Zener value shown; smaller Zener voltages gave more of a smooth compressed diode-clipping sound, while higher Zener values gave a more dynamic (i.e., touch-sensitive) overdrive, unfortunately also accompanied by muddiness on chords, excessive 'envelope', and splattering on hard-picked notes.

3) The Zener only operates about maybe 8 on the gain control. This allows you to eliminate the diode-clipping simply by reducing the gain.

When this circuit is used on a push-pull output stage, Zener conduction occurs only after tube cutoff, so Zener-diode clipping is never passed on to the speaker. This is not the case in a preamp tube; I definitely heard some characteristics of diode clipping, but I thought it really added sharpness and good-edge to the sound (every note I played made me think of Van Halen I), rather than just producing a buzzy mush like some diode-clipping circuits do. Only half the waveform is diode clipped - the other half shows the rounded edge of attempted grid conduction - and since the (5.1V) Zener-clipped half does not drive the 12AX7 into cutoff, it is passed on to the next stage where it gets some 'grid-conduction tube flavor' added on.

This is not really a blues-overdrive circuit - more hard rock/metal - but for that usage I was pretty impressed, and it definitely has more gain than any other 3-12AX7-GS circuit I've played through.

Ray

[Smitty]

I finally freed up some time to try this mod on the Marshall MV preamp circuit below:

Thanks for sharing this stuff. 'preciate it. Nice touch modifying the ground leg of the volume pot in the schematic to fit the diode part numbers in.

2) I liked symmetrical clipping the best. In this circuit, that happened with the 5.1V Zener value shown...Only half the waveform is diode clipped - the other half shows the rounded edge of attempted grid conduction - and since the (5.1V) Zener-clipped half does not drive the 12AX7 into cutoff, it is passed on to the next stage where it gets some 'grid-conduction tube flavor' added on.

What is the bias voltage for V5B? 2.55 seems high.

[Ray Ivers]

Smitty,

No problem at all about sharing - hopefully people will start grabbing clip leads and handfuls of low-voltage Zeners, and checking it out. I didn't have a chance to try out every ballpark Zener value, different bias points, tube types, etc. so there's still a lot of avenues to explore.

What is the bias voltage for V5B? 2.55 seems high.

I think the idle bias voltage was around a volt or so. It seems the 'symmetrical-clipping' Zener value for an unbypassed cathode-bias stage is roughly 4 to 5 times the idling bias voltage, up from the 2 x Vbias requirement for fixed-bias stages - but I think I'd want to try this thing out a few more times (particularly with bypassed Rk's) before setting that in stone.

It can be difficult to attribute this or that sonic quality to a particular component or circuit action, but I really believe one of the major benefits to this circuit is the elimination of tube cutoff; the tubes are always passing current (and music!), and not acting as switch elements.

Ray

[soundmasterg]

This is really cool Ray!

I wonder if I could incorporate it somewhere in my brother's Silvertone to help get rid of a notch on the downslope in the negative half of the wave? It happens coming right out of the EF86 stage as the volume gets turned up and I've theorized (with help from others) that maybe the following CF is not able to keep up with the output of the EF86 and is doing some grid conduction or something along those lines. It sounds like a little mosquito on top of the distortion.

Channel 1 on the Silvertone looks like this:

EF86 >> direct coupled CF using 1/2 12AY7 >> tone stack >> volume control >> gain stage using 1/2 12AY7 >> PI using stock Silvertone 1484 PI >> power amp.

Voltages and parts values on the EF86 are as follows:

plate = 106v 220k resistor
screen = 96v 1M3 resistor
cathode = 2k2 resistor, 25uf cap, forgot the voltage at the moment
supply voltage about 324v

Supply on CF is 324v

Could I stick those diodes on the grid of the CF and use a coupling cap of some value instead of direct coupling to the CF? The notch I'm seeing on the scope is in about the middle of the negative half downslope, and goes down from the slope. It gets deeper as the volume gets turned up.

Greg

[Ray Ivers]

Greg,

It seems to me that notch-type waveform phenomena (other than crossover distortion) are usually caused by phase cancellations of some kind.

Since the EF86 and the CF always have the same signal-level relationship regardless of the volume setting, a notch appearing on the EF86 output only at higher volume-control settings probably is caused by some kind of feedback from downstream of the volume control, most likely through the power supply or ground - at least that's my take on it.

If the CF has 324V on the plate and 106V on the grid, it should be in little or no danger of drawing grid current except on the very highest positive signal peaks. I think maybe the first places I'd look are the tone-stack and volume-control ground points, and increasing decoupling/filtering on the B+ bus might be a good idea as well. Also, just for grins - you might consider deriving the EF86 screen voltage through a voltage divider consisting of medium-value resistances - 100K, 220K, that range; a 1.3M screen resistance on an input stage can be sort of a 'trouble magnet'.

I think the Zener mod would be a very good thing to try out in your amp, I'm just not sure it will affect the waveform notch. I would also definitely keep the direct-coupled CF, regardless of whatever other changes you might make; you could always adjust the CF bias point using different values of Rk if needed.

Ray

[soundmasterg]

Hi Ray!

You know I didn't think about it, but you're right regarding the signal level between the EF86 and the CF...so it must be affected by something downstream after the volume. Each stage in the preamp channel is seperately decoupled with a 20uf, and the grounds come to a star point that eventually grounds at the input jack. The tone stack and volume control grounds come to the star point also. The CF cathode resistor is a 75k. The heater CT is referenced to a +40v supply. The other areas in the amp aren't optimized for grounding as well as this channel since a can cap was used for channel 2, but channel 2 star grounds at the can cap, and the power amp, PI, and trem ground at a star point on the other end of the amp.

Regarding the 1M3 screen resistor on the EF86....the Vox AC30/4 circuit has a 220k plate and a 1M screen I believe, and I ended up going higher on the screen resistor to get the screen voltage lower than the plate voltage. With a 1M screen and a 220k plate, the screen was about 10v higher than the plate, and it didn't sound as good either. Are you suggesting to split the 1M3 in two and use two resistors to equal that value on the screen, or something else entirely?

I could probably try the zener trick on the grid of the 12AY7 gain stage as it is right after the volume control, so its a similar spot to when you used it in your amp.

Greg

[Ray Ivers]

Greg,

It might actually be better to connect each stage's grounds together (Rg, Rk, decoupling cap, etc. - I'd treat the pentode/CF combo as a single stage), and then run wires from each of these 'mini-stars' to the main ground point. I'd also make sure to use heavy wire - or preferably a ground plate - for any ground connections.

A 330K/100K voltage divider (330K to B+, 100K to ground, obviously) should get you about 98V from a 324V B+ supply, at a much lower resistance than a single large series screen resistor - and the screen voltage will stay a lot more stable during signal swings.

I agree, the volume control should be about the best point to try the Zener circuit out.

Ray

[soundmasterg]

Thanks Ray.

All the grounds for channel one ground together at that star point. Its a simple matter to ground from there to somewhere else instead of the input jack as the star point is isolated from the chassis, or I could make the star point the chassis ground easily enough too. I'll have to experiment a little.

Channel two grounds at the cap can, and theres no way to change that ground point easily while using the can. Would it be better to ground channel 1's grounds to this point? Should I ground all the amp's grounds to this point or should I keep the heavy current stuff on the other end of the amp?

If I go with your suggested 330k/100k combination, do I need a cap in parallel with the 100k to ground? Currently theres a .1uf film cap from the screen to the cathode just like the Svetlana EF86 notes recommended.

Greg

[Ray Ivers]

Greg,

Having all the grounds for each channel connected together to a single point may be at least part of the problem. Unless the other channel's controls affect the notch, I wouldn't really be too concerned about it just yet - in fact, you might want to remove its tube to take it completely out of the picture for now, if that's feasible. I'd use the separate star-grounding for each gain stage that I mentioned in my last post, with all the separate gain-stage stars connected together to the existing star grounding point near the input jack.

Yes, it would be a good idea to bypass the 100K with a plastic cap. A .1uF will bypass down to about 21Hz, but you can use a larger cap to go down lower or a smaller one for tonal shaping, if you'd like.

Ray

[Matt T.]

Very interesting Ray. Couple of questions:
How is the transition between no diode-clipping into diode-clipping? Smooth, abrupt?

Why did you choose the second stage instead of the third stage? I would have guessed that most of the distortion was from the third gain stage but on the other hand, there has been more than one occasion where reducing gain has yielded a more crunchy tone...I assumed it was because the preamp tubes weren't going into cutoff as long/much. The way you have it now (i.e. zener-mod on second stage), does the diode-clipping/no-cutoff affect the following stage in any way?

I think you deserve credit for thinking outside of the 'all tube' box. I've always said that I didn't care if there were a hamster spinning a wheel inside, if it sounds good it is good.

fog horns with your head wrapped in wax paper

If I had a nickel for every time I built a high gain circuit that sounded like this (a few have been tin foil instead of waxpaper)...I'd have enough to spring for two diodes.

[Ray Ivers]

Matt,

How is the transition between no diode-clipping into diode-clipping? Smooth, abrupt?

I'd say maybe in-between; the Zener-clipped part of the waveform is ruler flat, but the leading/trailing edges are slightly rounded. The top half (that's clipped by the GS grid conduction) has both rounded edges and and a slight 'hump' along the top.

Why did you choose the second stage instead of the third stage?

It's always seemed to me that the 1st/2nd-GS interface is the most critical one in cascaded-GS overdrives; if the sound gets thin or muddy here, it can't be recovered later on. If I had put the Zener circuit only between the 2nd & 3rd stages, I would have had to greatly reduce the 1st coupling cap value, and the results wouldn't have been as good IMO. I also tried an additional Zener circuit between the 2nd and 3rd stage; initial results didn't really warrant leaving it in, but I'm going to try it again before I button it up.

The way you have it now (i.e. zener-mod on second stage), does the diode-clipping/no-cutoff affect the following stage in any way?

Yes, I think it does; by Zener-limiting the input, you prevent the following stage (the one the Zener circuit is installed on the grid of, #2 in this circuit) from going into cutoff, and it also has some effect on reducing the positive signal swing being fed to the next GS (#3, in this circuit) as well.

IMO it's so much better to stray 'outside the box' than to 'box oneself in' by eliminating from consideration 90% of the available technology with a tubes-only mindset.

Ray

[Matt T.]

Well I finally got around to ordering my 'handfuls of zeners'...I got all the Mouser values from the high 3s to the low 7s, figuring I can add two together to make other values. So I'll have 4.8v and 5.4v (I think those are the values 'surrounding' 5.1v) among others. They are in the mail.

I think the idle bias voltage was around a volt or so. It seems the 'symmetrical-clipping' Zener value for an unbypassed cathode-bias stage is roughly 4 to 5 times the idling bias voltage, up from the 2 x Vbias requirement for fixed-bias stages - but I think I'd want to try this thing out a few more times (particularly with bypassed Rk's) before setting that in stone.

I didn't really catch this before but the cathode is bypassed in the schematic in the op but you still used the 5x value...is that correct? I guess my question is: what difference does the cathode resistor being bypassed make here?

Another question: would putting a resistor in series with the zener mod soften it's effect...in a good way?

I'll report any findings I might have.

[Ray Ivers]

Matt,

I got all the Mouser values from the high 3s to the low 7s, figuring I can add two together to make other values. So I'll have 4.8v and 5.4v (I think those are the values 'surrounding' 5.1v) among others.

IMO that's the way to do it - otherwise Murphy's Law will screw you with the missing values.

the cathode is bypassed in the schematic in the op but you still used the 5x value...is that correct? I guess my question is: what difference does the cathode resistor being bypassed make here?

Well... with an unbypassed cathode, in this circuit a 5.1V Zener would basically create symmetrical waveform clipping at all guitar-range frequencies. With the circuit values shown, though, the 5.1V Zener would have less effect - and would allow progressively less-symmetrical clipping - as frequency increased above maybe 300Hz (or perhaps I should say 'more and more effect as frequency decreased below @ 300Hz', which better illustrates the mud-reduction effect of the Zener). I use a 400Hz test signal, though, so it could well be that I was in a transitional zone for this circuit, with some asymmetry occurring both above and below that frequency... hmmm...).

Another question: would putting a resistor in series with the zener mod soften it's effect...in a good way?

By all means, try it out and see! I did, and wasn't really happy with the dilution of the Zener effect it caused - I preferred the effect of using a higher-voltage Zener - but I only tried a few different combinations and could well have missed some really good-sounding 'sweet spots'. Some particular combination of coupling cap/Zener value/Rk/Ck might be absolutely perfect for the tone you're going for, and there may be all sorts of milder/bluesy overdrives available in addition to the "'You Really Got Me' intro" ones.

I'll report any findings I might have.

Cool!

Ray

[Matt T.]

Well... with an unbypassed cathode, in this circuit a 5.1V Zener would basically create symmetrical waveform clipping at all guitar-range frequencies. With the circuit values shown, though, the 5.1V Zener would have less effect - and would allow progressively less-symmetrical clipping - as frequency increased above maybe 300Hz (or perhaps I should say 'more and more effect as frequency decreased below @ 300Hz', which better illustrates the mud-reduction effect of the Zener). I use a 400Hz test signal, though, so it could well be that I was in a transitional zone for this circuit, with some asymmetry occurring both above and below that frequency... hmmm...).

That's so cool that it naturally tends to 'target' exactly what we want it to with a bypassed cathode (if I'm understanding you that is). I have (and have heard others) a homebrew Marshall smallbox and that amp just seems to get muddy and bassy when you crank it (with the accompanying higher, harsh harmonics). Even typical 'Superlead' values are still muddy and when I crank it I must run the bass on zero. Reducing cap values later in the circuit just make it sound thinner but don't really correct the muddy distortion problem. Why one amp does this and another seemingly 'almost identical' amp doesn't is beyond me but it looks like this might be just the ticket.

[Ray Ivers]

Matt,

That's so cool that it naturally tends to 'target' exactly what we want it to with a bypassed cathode (if I'm understanding you that is).

Yes, IMO that's exactly what it does. Rk bypassing with a smaller-value cap (to accentuate mids and highs) has a very limited effect on mud created by the grid circuit; increasing the grid-overload voltage to maybe 2 or 3 volts at lower frequencies doesn't mean much when you're hitting the grid with >20V peak-peak. In the Zener circuit, though, you're changing the activation frequency of the Zener, which will have a noticeable mud-reduction effect once it kicks in fully. You have to tweak a little to get it right for your own purposes, but it's worth the time spent.

Ray