AB763 Studio Mod

[SurvivorSound]

Hello everyone, I run a recording studio in Oakland CA, and while I used to do a ton of tube amp tweaking, most of my tech time these days is spent on outboard gear and console maintenance.

I came across a stock ‘75 Pro Reverb that seems to be the latest model without a Master Volume. A few extra ceramic caps here and there, but generally it will be able to get converted to AA165 pretty easily. Once that’s done and tested, I have an idea for a comprehensive preamp section mod for lower noise and significant reverb flexibility. I am not trying to change the fundamental tone of the Vibrato channel, nor significantly increase the gain above Blackface standards, but a little extra oomph is not a bad thing.

I’d love to know what you guys think of this. I believe it can apply to any of the latter Blackface models. It would be easier to follow if I redrew it in a linear fashion, but I recognize that people are very used to seeing the common way valves were laid out on the page in the Blackface and early Silverface era.

Yes, the V2B and V4A grids are tied together, and each does the same job downstream. My best shot at minimizing loading on the V2B plate.

V4B needs a 1M grid leak, I’ll include it in v0.3.

Also included are TSC curves for the modified Normal channel, now used for reverb recovery.

[stratomaster]

Seems like a long way around simply rerouting the reverb return RCA to the normal channel input via an adapter.  You can do some tone stack tweaks to the normal (now reverb) channel and internally route the reverb return to the grid of v1a via a grid stopper and just tape off the 1/4" inputs.

This way you can easily go back. 

[SurvivorSound]

Seems like a long way around simply rerouting the reverb return RCA to the normal channel input via an adapter.  You can do some tone stack tweaks to the normal (now reverb) channel and internally route the reverb return to the grid of v1a via a grid stopper and just tape off the 1/4" inputs.

This way you can easily go back.

Indeed, that is the genesis of this mod. I don’t need to go back.

[stratomaster]

In that case, I'd be suspect of the signal split at the vibrato channel volume wiper.  You might do better to take the signal off the plate of the following triode.  The mixing wet/dry circuit could work, but you already have independent volume control.  To minimize potential channel interaction you can forgo this entirely and just take the reverb eq recovery out to the unused PI input.  No mixing resistors or pots needed. Lower resistor noise.

I'm also a bit unsure of what the function of the line in is.  To potentially use the amp as an outboard spring reverb unit?

[SurvivorSound]

In that case, I'd be suspect of the signal split at the vibrato channel volume wiper.  You might do better to take the signal off the plate of the following triode.

I am curious what the problem is with tying the grids of V2B and V4A together. Others have expressed suspicion of that as well. I don't believe there's a loading issue, though the Volume curve might dip a hair. Here are some alternatives:

• If Dwell were a 1M-A with a 110K stop, preceded by a 500p cap (instead of 2n2), then I could feed said cap with V2B directly, and V4A could be eliminated altogether. My issue is noise. The resultant maximum source impedance into V3 would be something like 278KΩ, whereas V3 sees a maximum source impedance of something like 70KΩ as drawn.

• If I couple into the V4A grid off of the V2B plate (as I believe you're suggesting), I get excess gain into V3 via the Dwell control. But along those lines, I could set up V4A as a cathode follower, where the plate of V2B feeds the grid of V4A, V4A cathode is held 100K above ground, and V4A cathode feeds Dwell as drawn. It's still an extra stage in the chain, and a cathode follower doesn't quite align with the Fender design approach of the era. But it would work.

• If all three connections to V2B and V4B were swapped, I could run V4A and V4B in parallel (like V3) to do the job that V2B does now, and the resultant current could easily drive both the mix resistor load and the Dwell control as drawn, without undue level loss due to loading. Sort of a distribution amplifier approach.

The mixing wet/dry circuit could work, but you already have independent volume control.  To minimize potential channel interaction you can forgo this entirely and just take the reverb eq recovery out to the unused PI input.  No mixing resistors or pots needed. Lower resistor noise.

Well, that pot at the grid of V4B is a temporary trimmer. Its point is fine balance, then you adjust the 120K resistors to match whatever balance you like. I'm not sure I understand the suggestion -- are you talking about joining the plates of V1B and V2B directly? The reason for keeping the mix resistors and V4B in there is twofold: the heavy loading of joining those plates together, and the further loading of the Intensity pot. Even if I swap the 50K-RA for a 100K-L, there's a bunch of attenuation (and that's in the original circuit too). If V4B feeds just the Intensity pot in parallel with the P.I., it's a pretty reasonable load.

I'm also a bit unsure of what the function of the line in is.  To potentially use the amp as an outboard spring reverb unit?

You got it.

[SurvivorSound]

I wonder if it's the additional grid-to-plate and grid-to-cathode capacitance of placing the grids of V2B and V4A in parallel that is giving people pause. Do I have it right that a typical 1M volume control's maximum source impedance of 250K (or a little more with the tone stack) results in a -3dB point down in the 3300Hz range when facing the grid of a 12AX7 whose gain is approx. 60? And would the paralleling of these two result in a -3dB point down in the 1650Hz range? Granted, this is only significant when the control is somewhere around 8 on the dial.

[SurvivorSound]

Alright, here's a version that does not add capacitance after the Vibrato channel volume control, and uses the just-connect-the-two-plates mix method. I'm not 100% sure I'm doing the cathode follower right.

There is some loss by not running either channel through a makeup stage after they mix, but I'm making it up by letting the V2B cathode sit over 820Ω (while V1B gets its own 1500Ω), and probably dropping a 12AX7 into the phase inverter position instead of the stock 12AT7. Might actually be pretty close to the original design in terms of gain.

[PRR]

> doing the cathode follower right.

I think you were better with the previous plan.

Taking one signal to two grids "suggests" that a simpler solution exists. You might save a buck! No, because triodes come two to a bottle, and resistors are relatively free, especially since you will only build this once/twice.

[SurvivorSound]

Taking one signal to two grids "suggests" that a simpler solution exists. You might save a buck! No, because triodes come two to a bottle, and resistors are relatively free, especially since you will only build this once/twice.

I appreciate the point, but I don’t get it in this context. $ is no concern here.

[stratomaster]

The mixing wet/dry circuit could work, but you already have independent volume control.  To minimize potential channel interaction you can forgo this entirely and just take the reverb eq recovery out to the unused PI input.  No mixing resistors or pots needed. Lower resistor noise.

Well, that pot at the grid of V4B is a temporary trimmer. Its point is fine balance, then you adjust the 120K resistors to match whatever balance you like. I'm not sure I understand the suggestion -- are you talking about joining the plates of V1B and V2B directly? The reason for keeping the mix resistors and V4B in there is twofold: the heavy loading of joining those plates together, and the further loading of the Intensity pot. Even if I swap the 50K-RA for a 100K-L, there's a bunch of attenuation (and that's in the original circuit too). If V4B feeds just the Intensity pot in parallel with the P.I., it's a pretty reasonable load.

My recommendation was to eliminate the mixing resistors altogether and wire the output of the reverb eq stage to the lower PI input.  I think the Two Rock John Mayer amp does something like this.  They also attenuate the signal with a monstrous voltage divider.  Also lots of two channel 18 watt TMB amps also take advantage of both PI inputs instead of a mixing stage. 

You'd probably do well to leave the 3.3M/10pf in the circuit to act as the reverb tap point and additional loading you were seeking with the mixing stage.  It's a tried and true design and eliminates the parallel grids off the volume pot.   I don't think the cathode follower stage helps much with your goals. 

[SurvivorSound]

“ My recommendation was to eliminate the mixing resistors altogether and wire the output of the reverb eq stage to the lower PI input.”

Hell yeah. I get it now. That makes things really simple. Total channel separation and light loading means I can tap ‘verb send off where it normally comes from. I’ll post an update.

[EDIT: Suggestions on a bias Tremolo? Because now there’s no summing junction to tap the top of the Intensity pot in to.]

[EDIT 2: I’m assuming someone will suggest this: https://el34world.com/Hoffman/files/Hoffman_AB763_1.pdf]

[sluckey]

“ My recommendation was to eliminate the mixing resistors altogether and wire the output of the reverb eq stage to the lower PI input.”

Hell yeah. I get it now. That makes things really simple. Total channel separation and light loading means I can tap ‘verb send off where it normally comes from. I’ll post an update.

The lower PI input is not available. It is being used by the NFB circuit.

[SurvivorSound]

The lower PI input is not available. It is being used by the NFB circuit.

Can I leave the 820R / 22K / 100R NFB node alone, and run a second NFB resistor off the trafo secondary in a particular ratio to the bottom P.I. input? NFB is about 10%. I’m picturing an 820K coming back to the (new) joint between the 22n and 100n, with 100K to ground. The 100n could shrink in value, or the 22n could. A bit of loading on V2B, but not terrible. Or is that just an invitation for noise? [EDIT: Both options attached]

Alternatively could we cathode follow V2B and drop a second NFB value into a split cathode resistor there? [EDIT: Third sketch - I forgot to change the top 100K back to 82K.]

And, what happens if no NFB is provided at the bottom PI input, just at the 820R/22K/100R node? A little additional differential gain, or something more problematic?

I just really like the dual input PI direction as it solves several challenges in one fell swoop.

[LAST EDIT: Flipping the script in the fourth sketch, Normal channel feeds the Dwell control and Vibrato channel does reverb recovery and the aforementioned cathode NFB scenario. Just pretend I changed that top 100K back to 82K.]

[SurvivorSound]

I think I got it.

[SurvivorSound]

Sorry for the barrage above. Here's the full picture.

Typo on the 100Rs on the heater supply, whoops.

[d95err]

Are you sure you want to run the reverb return signal through a Fender-style tonestack?  It has a huge mid dip, unless treble and bass are at zero.

I would imagine taking the mids out of the reverb return signal would basically leave mostly hiss and rumble. I haven’t tried it though.

I would try something like a Baxandall tonestack instead.

[kagliostro]

Just asking because curious

the Middle Pot you added (in series with a 1.8K resistor) is 50 K-A as you wrote or is to be considered it is a 5K-A pot ?

Franco

[SurvivorSound]

Are you sure you want to run the reverb return signal through a Fender-style tonestack? 

Look at the values on the fka Vibrato channel stack. Same topology, radically different response. Does the EQ basics that we do in the studio all the time. TSC curves upthread (first post IIRC)

Anyone have thoughts on whether my NFB solution for the second P.I. input will work?

[SurvivorSound]

the Middle Pot you added (in series with a 1.8K resistor) is 50 K-A as you wrote or is to be considered it is a 5K-A pot ?

50K proper 10% log taper. Can go on the back if need be. 1K8 + 5K = 6K8. The best knob for that is one from a Fender PA100 EQ (-5/0/+5). The control becomes an overall gain above Fender standard 10K, before veering into Marshall and finally nearly “raw mod” territory. As one would expect, treble and bass controls become less effective as gain increases exponentially. It’s really just a way of having a regular Fender mid control and a tone stack bypass switch consolidated into one control.

[kagliostro]

Thanks for the reply SurvivorSoundand

interesting info about the effect of the Fender PA100 EQ, I didn't know about it

Franco

[SurvivorSound]

Okay here's v0.6, wherein I've removed the NFB tie to the tail of the P.I. (a "Leo error" which I was reminded provides PFB to cancel a wee bit of NFB) and compensated with a little less NFB to the node in question. Still haven't seen this on the 'net anywhere but I figure 39K off the V2B plate vs 390K off the OP trafo secondary is about the right ratio. Unsure how that ratio might be affected by the 1M grid leak, but probably not by much. Comments?

And yeah, I got plans for V4. Probably a White cathode follower. 12BH7, ECC99, et al. Haven't gotten there yet. If anyone has done this in an instrument amp, please post link or schematics. Mostly I see OTL headphone amp designs, which are pretty easy to adapt, but I need the actual output impedance so I can match it on the cold for an impedance-balanced output. And of course, the fewer components, the better. Will probably want to set that up for 300Ω-600Ω and terminate it 600Ω-1k2Ω so that the WCF doesn't care if the line receiver is 600Ω or 10KΩ or 100KΩ.

[SurvivorSound]

Thanks for the reply SurvivorSoundand

interesting info about the effect of the Fender PA100 EQ

just the knob itself! i have one but haven’t checked the schematic - the 0 in the middle is just a good reminder that that is the stock value. easier than saying “not a normal Fender mid” to a visiting player.

[stratomaster]

Okay here's v0.6, wherein I've removed the NFB tie to the tail of the P.I. (a "Leo error" which I was reminded provides PFB to cancel a wee bit of NFB) and compensated with a little less NFB to the node in question. Still haven't seen this on the 'net anywhere but I figure 39K off the V2B plate vs 390K off the OP trafo secondary is about the right ratio. Unsure how that ratio might be affected by the 1M grid leak, but probably not by much. Comments?

And yeah, I got plans for V4. Probably a White cathode follower. 12BH7, ECC99, et al. Haven't gotten there yet. If anyone has done this in an instrument amp, please post link or schematics. Mostly I see OTL headphone amp designs, which are pretty easy to adapt, but I need the actual output impedance so I can match it on the cold for an impedance-balanced output. And of course, the fewer components, the better. Will probably want to set that up for 300Ω-600Ω and terminate it 600Ω-1k2Ω so that the WCF doesn't care if the line receiver is 600Ω or 10KΩ or 100KΩ.

I like where this is going and can't wait to hear about how it works/sounds. 

If you're after a sinusoidal tremolo and up for reducing complexity you can try Sluckey's Trem-O-Nator circuit.  This will also free up a triode if you wanted to add the 3rd stage back into the signal path and reuse the bulk of the components that already exist in the tremolo circuit.  If you're dead set on bias trem then disregard.  Just know there are some inherent drawbacks with bias trem and 6L6 output stages--but they can be mostly overcome if you're willing to use diodes, LEDs, alternate resistor values, and a reduced overall bias setting. 

[SurvivorSound]

I like where this is going and can't wait to hear about how it works/sounds. 

Appreciate that. I don’t think the amp is gonna sound any different, fundamentally. Reverb should sound exactly the same with its Treble and Bass controls at 5 and Dwell at 10. The only difference is level. The Normal channel has 6dB less attenuation than stock, and it’s loaded lighter than stock. So call it a smidge louder than twice as loud as stock. Probably about the same as playing dry through the stock Vibrato channel.

If you're after a sinusoidal tremolo and up for reducing complexity you can try Sluckey's Trem-O-Nator circuit. 

I have seen the Trem-O-Nator and I like it. It’s the simplest thing out there AFAICT, and I am not attached to bias trem.

However, if I use the two-channel P.I. mix method, there is no opportunity to use a tremolo that relies on loading the output of a single triode stage. If there is a way to adapt the Trem-o-Nator to this v0.6 architecture, I’m all ears. I’ve never dug into trem in any mod previously so I’m no expert.

There are master volume solutions that simply apply a resistance across the signal lines at the 100n / 220K / 1K5 nodes. Is there a simple trem design that places the Vactrol’s variable resistance there instead? Or a roach’s Cds cell? If a swing between 1M and several tens of kilohm is typical, that would be an optimal differential loading opportunity.

Sluckey commented a coupla days ago, inspiring my unusual NFB solution, so….any more insight on how this is progressing? I don’t know how to @ a user on this forum.

[kagliostro]

May be you can try to simply double the VTL5C1 and use the two "exit" one for one "channel" and one for the other ?


Franco

[SurvivorSound]

May be you can try to simply double the VTL5C1

Now THAT would change the sound of the amp.

There is no way to match them perfectly, so the signal gets lopsided.

It’s a differential signal. You have to deal with both sides with one control. By bringing the two sides closer together, you force a perfectly symmetrical phase cancellation, and you get a matched plate loading effect.

A circuit that has both speed and intensity controls modulating the optical element could work.

Both ends of the Cds would be at the same DC potential so no worries there. I dunno about current capacity, but it would have to be able to handle the current sunk across its lowest resistance.

[kagliostro]

You have to deal with both sides with one control.

I was thinking to something like this, each VLT with his output control



Franco

[SurvivorSound]

I was thinking to something like this

Oh, I see, you're proposing to load the outputs of V1B and V2B equally. That would work, but would really attenuate levels heading into the P.I., which I'm trying to avoid altogether. Also in my architecture it would affect the reverb send level.

Here's what I'm saying -- Differential Trem-O-Nator attached. Very little attenuation with the Intensity pot at CCW. This is an automatic variable U-pad, using the plate impedances of V6 as the input values. I suspect the plate load resistors and bias feed resistors may be in parallel to the plate impedances in this case. I am guessing that it would be of benefit to use matched 100K plate load resistors and matched 220K resistors. Matching the plate loads might be okay with my alternative NFB method, since the second P.I. input is not grounded. I dunno. Waiting for an expert to weigh in.

[EDIT: I was thinking 12AX7 instead of 12AT7 in the attached sketch, and hadn’t calc’d the pad. So. Rlim should be low, maybe 1K for starters. Probably looking at a lower value for the intensity pot as well, 250K-RA could work. 100K-RA would work better, but it’s a rather heavy load when Intensity is CCW.]

[kagliostro]

One other option will be to use "my" previous proposal connecting the VLT as PPIMV, something like the LarMar or as substitution of the 220K resistors instead of differential PPIMV ?

Franco

[SurvivorSound]

One other option will be to use "my" previous proposal connecting the VLT as PPIMV, something like the LarMar or as substitution of the 220K resistors instead of differential PPIMV ?

Franco

I’m not attuned to the reason why the 220Ks are that value. Bias current I assume. Are there options there? Also can you post the LarMar reference please.

[Willabe]

You should look through Rob Robinette's amp web site, if you haven't yet.

Lots of Fender mods. Probably find what you need that's already a proven working mod.

Amp Mods

[SurvivorSound]

You should look through Rob Robinette's amp web site,

I did, thanks

[kagliostro]

Some examples of PPIMV









About the 220K value is a value often used, sometimes you find there 470K and few times very high values in the order of 800K

I don't think the value is joined to the bias path as you can see it also in Cathode Biased amps

Franco

[SurvivorSound]

Some examples of PPIMV

As a Vactrol is not a potentiometer, I fail to see how two of them can be used as a PPIMV. I can see how a dual Vactrol could be used as a load in parallel with the 220K resistors. I like the markh technique of adding a 1u pair after any control there - though 1% matched 220Ks would be almost as good. The treble cut control in the Rocket schemo is along the lines of what I drew, except without the cap.

I compared the plate loads of AB763 (well, AA165 Pro) and the AA769 Twin. It seems to me that a 100K-RA across the hot and cold lines, after the 100nF caps, would present the same loading to V6A and V6B as they are presented with in that Twin.

So the question is if there is any issue with simply treating the 12AT7 P.I. outputs in the exact same manner as we treat any 12AX7 triode that we are applying a 50K-RA shunt tremolo to. 100K differential is the same as 50K single ended. And the follow-on question is if the phase inverter has any particular aversion to having its output shunted almost entirely together at full CW Intensity (which, by the way, is the same as turning the PPIMV almost all the way CCW in that Rocket schemo).

[kagliostro]

As a Vactrol is not a potentiometer, I fail to see how two of them can be used as a PPIMV

Don't you think you can connect the fotoresistor of the Vactrol as to shunt to ground the signal like when you move the wiper of the potentiometer of a PPIMV near the ground connection ?

Instead to rotate the shaft of the potentiometer you have the LED that lights according to the Oscillator frequency, didn't that give a similar effect ?



Franco

[sluckey]

As a Vactrol is not a potentiometer, I fail to see how two of them can be used as a PPIMV.

He's not suggesting using a vactrol as a PPIMV.

[kagliostro]

Ciao Steve

According to an old memory about something I've read on a Merlin's book ... if you use a variable voltage to supply the LED I think also a PPIMV can be realized, obviously I wasn't suggesting that on my previous post

(Merlin showed the use on a Long Tail PI as "Power Control")

Franco

[stratomaster]

Ciao Steve

According to an old memory about something I've read on a Merlin's book ... if you use a variable voltage to supply the LED I think also a PPIMV can be realized, obviously I wasn't suggesting that on my previous post

(Merlin showed the use on a Long Tail PI as "Power Control")

Franco

I think you're referring to a variable fixed voltage to establish the LED at a certain brightness which-in parallel with the Intensity pot as proposed on the differential Trem-O-Nator sketch--would act as a cross line master.  I like the idea.  Switching would be tricky and likely need relays unless there's a good way to freeze the oscillator at a certain point on its swing. 

[SurvivorSound]

Strike the ground tie and the lower vactrol and connect the bottom of the upper vactrol to the cold line. There is no need for two Vactrols, and the inability to match them (even within a dual package) causes more issues than it solves.

Then redesign the Trem-O-Nator to control Intensity in advance of the LED. Then we’re talking. Maybe Steve has an idea. Maybe 2deaf has an idea. I was trying to avoid messing with the TON design.

Putting a 100K-RA in there to roll in the Cds in parallel to itself should work without reinventing the TON.

But hey, since we’re here: What about replacing the 10K Rk of the TON with a 10K-L pot, and feeding the vactrol’s LED off the wiper? Stupid simple. No audio there, just DC, no scratchy worries…

[kagliostro]

@ stratomaster

Ahem .... no, I was only saying that despite the fact I wasn't suggesting to use a Vactrol as PPIMV if wanted it will be possible to realize it

@ SurvivorSound

I'm not sure that the crossline MV itself is a good choiche, there are people that aren't satisfied by it, don't know exactly the reason, may be because they have a PI that isn't at a perfect balanced level ? Don't Know ....

This is the reason for which I suggested the LarMar or similar PPIMV

Franco

[SurvivorSound]

@ SurvivorSound

This is the reason for which I suggested the LarMar or similar PPIMV

the first three you posted are good PPIMV designs, yes. the Rocket misses the mark.

and now i understand what you were saying about optical control of master volume.

[SurvivorSound]

Light load differential trem-o-nator, with 2K max depth trim, 10K intensity, and a slug option for faster startup off of CCW (if needed).

The slug could be a 20K/25K trimmer. I am unsure how critical the 10K value of Rk is, and whether the change in total value at lower settings is enough to effect functions of the circuit other than Intensity alone.

Yea/nay?

[SurvivorSound]

Edging ever closer to answering my own question here, with the pot repurposed as a VCS…surely some Rp experimentation will be required.

[stratomaster]

The ToN is rather sensitive to the VTL5C1s used.  I found Rp had a big impact on tremolo depth on the traditional circuit, but also limited the resistance during the dark swing to a point well below the expected max resistance.  I suspected the LED was not going full dark and basically confirmed it with slow motion footage.  Since the LED sets the bias of the triode I was reluctant to put any resistance in series with it at the cathode but did not experiment much beyond the plate resistor and parallel cathode resistance.

This doesn't answer your questions, but are relevant learnings.  Build your proposed circuit with an LED instead of the vactrol and study it visually and with current/voltage measurements. 

[kagliostro]

A small step back, just to add a bit to the PPIMV info

...... (which, by the way, is the same as turning the PPIMV almost all the way CCW in that Rocket schemo)

..... the first three you posted are good PPIMV designs, yes. the Rocket misses the mark.

I remembered I've a bit of (Rob Robinette) documentation about that way to implement a PPIMV



One other way to do a similar PPIMV (using pot in place of the 220K Grid Leak resistors) is this



Franco

[SurvivorSound]

The ToN is rather sensitive to the VTL5C1s used.  I found Rp had a big impact on tremolo depth on the traditional circuit, but also limited the resistance during the dark swing to a point well below the expected max resistance.  I suspected the LED was not going full dark and basically confirmed it with slow motion footage.  Since the LED sets the bias of the triode I was reluctant to put any resistance in series with it at the cathode but did not experiment much beyond the plate resistor and parallel cathode resistance.

This doesn't answer your questions, but are relevant learnings.  Build your proposed circuit with an LED instead of the vactrol and study it visually and with current/voltage measurements.

Very relevant, cheers. Here’s an updated starting point.