My quest to build a push-pull stand-alone reverb

[MakerDP]

Short story... I wanted a tube reverb for my Trinity TC-15 kit I built. Didn't want to build a 6G15 just because I want to be different. Saw a nice simple design in Tonnes of Tone, saw the AX84.com FireFly by Doug H. So, I combined the two into my own design and built it. It sounds pretty good but I think it could be better. It has a 12AX7 (gain/isolation and recovery) and a 12AT7 driver (self-split design.)

I have a schematic for it as it sits today on my blog MakerDP

I did a lot of research on this and just the other day I discovered Hammond uses 12A_7 tubes in push-pull all over the place in their organ preamps and tone cabinets to drive percussion and reverb transformers. I have been studying those and they all use a Schmitt/LTPI to drive the push-pull tube. I am in the process of switching over from the self-split to this scheme to see how it turns out. Waiting on caps from AES so i should be able to do this fix on Tuesday or so.

Things I don't necessarily like about my design:

1) It has a grounding issue, which apparently is common for stand-alone reverbs. I have been reading all about that here and I think the solution is to build a transformer isolation box to sit between the reverb unit and the amp.

2) I have to push the dwell pretty hard to get a useable level of reverb, but it never really gets "boingy." I'll not regularly use a "boingy" setting but it would be nice to have just in case.

3) It is a little too "dark." I could use a little bit of a treble boost, or maybe somewhere in the circuit I am cutting too much out. The tone control has very little effect on the reverb.

4) I don't necessarily like the mix control. Specifically, how it really cuts the signal level when you dial-in the reverb. Will more gain at the recovery stage help with that so I don't have to turn the mix control quite so far to the "wet" side? How about just a standard "volume" control on the wet signal and then mix it back in with the un-attenuated dry signal?

Anyways, I'd like the pro's opinions here.

Thanks for looking!

[Chuck H]

There are a few things that may help:

There should be a grid stop on the input to prevent noise and instability.

The input and output jacks should have the sleeves isolated so that the host amps cable sleeve is the only ground for these circuits. This helps prevent ground loops.

The 330k/470k/220p coupling circuit between the input and the driver does not form a voltage divider and has very little HP filter effect. This could be why the amp sounds dark. You're reverberating too much low end. Moving the load will make it a HP filter.

The coupling cap from the input cathode to the wet/dry pot should be non polar because there is no steady 0V at the negative end. Not to mention that the signal here is tiny. There may be a tricky way to bump it with the suggested added gain stage.

You really need another gain stage for reverb recovery. The signal from the tank is typically under 7mV! With the tone circuit and loading I suppose less than 100mV of signal at the output.

Grounding for the pan depends on it's specific design. If the sleeves are connected to the pan chassis, etc.

[nickb]

Crucial question- What turns ratio and which reverb tank are you using?

[MakerDP]

Crucial question- What turns ratio and which reverb tank are you using?

Oops... yeah that would be helpful to know, wouldn't it!

OT running at 22.5k primary and 8-ohm secondary.

Tank is a MOD 4AB3C1C... 8-ohm input 2250-ohm output.

[MakerDP]

Thanks for your tips Chuck. To respond and ask some further questions...

There are a few things that may help:

There should be a grid stop on the input to prevent noise and instability.

The input and output jacks should have the sleeves isolated so that the host amps cable sleeve is the only ground for these circuits. This helps prevent ground loops.

The layout has a shielded input from jack to grid. Should be OK in most cases. Not really sure lack of a grid-stopper is an issue. Definitely worth looking into though. I do need to maybe isolate the input jack though. Currently everything except the power transformer grounds are grounded local-stars-to-buss-style to the input jack then to chassis. Well, except the mains are also grounded at the chassis of course. So, only those three ground points. But floating the input jack is something easy to try and I will definitely give that a shot.

The 330k/470k/220p coupling circuit between the input and the driver does not form a voltage divider and has very little HP filter effect. This could be why the amp sounds dark. You're reverberating too much low end. Moving the load will make it a HP filter.

This I am not so sure about. My first iteration has no grid-stopper here at all. I had a well-respected builder suggest I put one in. I did and I followed Merlin's designs in figures 4.20 and 4.27 (Designing Tube Preamps... 2nd ed.) pretty much verbatim. The reverb was WAY darker than it is now before I added the capacitor. Am I misinterpreting Merlin's ideas here?

The coupling cap from the input cathode to the wet/dry pot should be non polar because there is no steady 0V at the negative end. Not to mention that the signal here is tiny. There may be a tricky way to bump it with the suggested added gain stage.

OK this one has me stumped. The purpose of this is to isolate the "dry" input signal from the "wet" reverb output from the tank so they can be mixed properly at the "mix" or "blend" pot. In reading Merlin section 1.18, a value of about 22uF allows pretty much maximum gain (just a hair under 1.0x gain with this cathode isolator) across full bandwidth - fulfilling exactly the purpose of this, no? Lower values of this cap, especially in the non-polarized range would significantly attenuate lower frequencies of the dry signal. (This part of the circuit is also copied verbatim from O'Connor's design in Tonnes of Tone for his two-tube single-ended stand-alone reverb.) Again though, maybe I am missing something.

You really need another gain stage for reverb recovery. The signal from the tank is typically under 7mV! With the tone circuit and loading I suppose less than 100mV of signal at the output.

This may very well be the case and the sole reason why I don't get a "boingy" reverb. However, the original intent going in for my design was to keep it to two tubes. So, this may be something I just need to live with. Maybe I will try bypassing the tone circuit to see what that does.

Grounding for the pan depends on it's specific design. If the sleeves are connected to the pan chassis, etc.

OK, this I was wondering about and am flying totally blind on. I have no idea how to "properly" ground this tank. Both the tank's input and output jacks are grounded to the star-buss line and I am using a shielded cable sold by AES specifically for reverb tanks. Nothing special at the tank end, but I believe one of the tank's jacks is isolated from the tank itself if I read the tank's part number and the Accutronics document on what it means correctly. How that means I should wire the jacks on the chassis end is a mystery to me so I just wired it up as stated.

Your thoughts?

[Chuck H]

This I am not so sure about. My first iteration has no grid-stopper here at all. I had a well-respected builder suggest I put one in. I did and I followed Merlin's designs in figures 4.20 and 4.27 (Designing Tube Preamps... 2nd ed.) pretty much verbatim. The reverb was WAY darker than it is now before I added the capacitor. Am I misinterpreting Merlin's ideas here?

I'm sure. The impedance of the grid circuit is very high. With the load before the RC circuit there will be very little HP filter effect. If you move the load to the other side of the RC circuit the impedance drops to the load and the circuit effect will be much greater. I promise.

OK this one has me stumped. The purpose of this is to isolate the "dry" input signal from the "wet" reverb output from the tank so they can be mixed properly at the "mix" or "blend" pot. In reading Merlin section 1.18, a value of about 22uF allows pretty much maximum gain (just a hair under 1.0x gain with this cathode isolator) across full bandwidth - fulfilling exactly the purpose of this, no? Lower values of this cap, especially in the non-polarized range would significantly attenuate lower frequencies of the dry signal. (This part of the circuit is also copied verbatim from O'Connor's design in Tonnes of Tone for his two-tube single-ended stand-alone reverb.) Again though, maybe I am missing something.

I just looked at the 6g15 schem and it seems you DO have enough gain. I forget that you're not feeding a PI, but rather an amp input or effects loop return. And if merlin says that a 1.5k cathode will have a unity gain signal on it, well that's alright with me. But as to the 22uf cap... Any time the negative end of an electrolytic cap circuit isn't at a relatively steady 0V it's best to use a non polar cap. Non polar electrolytics in the 22uf value are readily available. The circuit may indeed be fine without a NP cap since the voltage will be very small. Still...

The boingy reverb is typically an effect of how hard the pan is driven. However, and I haven't researched this, with too much low end going to the pan it may not respond the way we expect it to.

The way your grounded seems fine. Try isolating the output jack sleeve from the circuit ground. Don't float the input jack or there will be no ground for the guitar cable if you plug straight into the reverb.

[MakerDP]

OK you have given me some good things to look at. I have some parts arriving today from AES. I'll try some stuff tonight.

[Mick Bailey]

1) I was pondering the 'floating' 22uf cap (C2). The gain and frequency response of the stage isn't optimised because the tail-end of the cap is presented with a combination of fixed and variable resistance to ground (adjusting the tone control affects the gain and response of the stage). It can also be affected by the loading on the output.

2) I'd go for a simple voltage divider after V1a, by inserting a resistor between C1 and R5 and removing C3, C12, R6 and R18. You can then experiment with the value of C1 to get the best low-end roll-off.

3) I've also had lots of problems with hum when prototyping standalone reverb units, but I've never had to resort to an iso transformer as a fix. Tube quality, pan orientation, cable screening, PT location, lead dress and overall grounding scheme have been issues. Sometimes the breadboarded design works fine, but the final build is noisy. Fender now elevates the signal grounding on the reissue units and this can have a drastic impact on noise, but is dependent upon what amp you use.

4) Getting enough signal drive into the pan at higher frequencies can be challenging. A good reverb should 'track' the guitar neck. Accutronics has a really interesting spec document for the signal drive characteristics and I've used this as a basis for designing the drive. I notice that the 6G15 can drive the pan really hard (compared with the drive spec for the pan) and this contributes to the character of the reverb at higher dwell settings.

[MakerDP]

2) I'd go for a simple voltage divider after V1a, by inserting a resistor between C1 and R5 and removing C3, C12, R6 and R18. You can then experiment with the value of C1 to get the best low-end roll-off.

I've actually thought about trying almost exactly this after looking at some preamp designs last night. My thoughts were to replace the Dwell's 250K pot with a 1M pot and eliminating everything after that up to the next triode. I wouldn't need the added resistor between C1 and R5 that you suggest then, would I? If I left in the 250K pot, what would you suggest for a value of that added resistor? I was also thinking of trying a "bright cap" around the dwell pot to allow some highs to pass through as well.

4) Getting enough signal drive into the pan at higher frequencies can be challenging. A good reverb should 'track' the guitar neck. Accutronics has a really interesting spec document for the signal drive characteristics and I've used this as a basis for designing the drive. I notice that the 6G15 can drive the pan really hard (compared with the drive spec for the pan) and this contributes to the character of the reverb at higher dwell settings.

I'll go and look for this document. Thanks for the tip!

The tone control does effect the dry signal. Is this avoidable by tweaking/eliminating C2 or is this just a reality of the design of V1a serving double-duty as gain and signal isolator?

As far as hum... I am thinking my issue is layout and lead dress related. I currently have this in a VERY small chassis so things are pretty tight inside and I did not experiment with power/output transformer placement before I put it all together. I probably should move this to a bigger chassis and pay more attention to transformer orientation.

This may get sidelined for a couple weeks... I am going on vacation tomorrow and while I am gone my Edcor transformers for my next build will arrive. It's basically a simplified Fender Twin 6V6 amp. Keep the ideas coming though because I am definitely wanting to get this reverb unit working satisfactorily.

[Mick Bailey]

You may not need a resistor before the dwell pot - depends if you're getting too much drive or clipping when the dwell is turned up. A 1M pot + bright cap would be my choice.

Taking the dry signal off via C2 is compromised. Ideally the cap would be grounded to give a fixed-gain stage and the dry signal taken off elsewhere. I'd be tempted to ground the cap, bleed some signal off C1 and mix that in with the wet signal. The nature of reverb means that phase cancellation isn't an issue, and plenty of FX pedals invert anyhow. Whether this is an issue depends on where you place the unit in your signal chain.

I'd also temporarily bypass the tone circuit as you originally suggested. Get the wet/dry mixing as you'd like it and the drive to the pan optimised. Then add in the tone control. If the bottom end of the signal going into the pan is just right, then I've found that a simple treble roll-off to manage the high-end is all that's needed. You'd get a little less insertion loss and keep as much wet signal strength as possible. You could even experiment with locating the treble control on the input to the pan.

I like what you're doing with this and breaking away from established designs is a good thing. It takes a lot of work to get a simple circuit working well - more so than a complex one, but when you nail it the satisfaction is immense.

[MakerDP]

You may not need a resistor before the dwell pot - depends if you're getting too much drive or clipping when the dwell is turned up. A 1M pot + bright cap would be my choice.

Taking the dry signal off via C2 is compromised. Ideally the cap would be grounded to give a fixed-gain stage and the dry signal taken off elsewhere. I'd be tempted to ground the cap, bleed some signal off C1 and mix that in with the wet signal. The nature of reverb means that phase cancellation isn't an issue, and plenty of FX pedals invert anyhow. Whether this is an issue depends on where you place the unit in your signal chain.

I'd also temporarily bypass the tone circuit as you originally suggested. Get the wet/dry mixing as you'd like it and the drive to the pan optimised. Then add in the tone control. If the bottom end of the signal going into the pan is just right, then I've found that a simple treble roll-off to manage the high-end is all that's needed. You'd get a little less insertion loss and keep as much wet signal strength as possible. You could even experiment with locating the treble control on the input to the pan.

I like what you're doing with this and breaking away from established designs is a good thing. It takes a lot of work to get a simple circuit working well - more so than a complex one, but when you nail it the satisfaction is immense.

Thanks for the encouragement!

I changed the dwell to 1M, added a 220pF bright cap (didn't have anything smaller), eliminated all the garbage beyond it, and bypassed the tone control. That was a disaster... no sound coming out at all. Then it occurred to me that I removed the grid leak resistor 330K R6 on V2a so I put that back in. Better, but still something wrong... so I added the 22nF coupling cap C8 back in and presto... the reverb came alive. WAY more wet level than before. We are getting closer, but.... I introduced two MAJOR issues by doing that:

First, I have some pretty serious clipping with the dwell cranked-up. So... I think I need that resistor you were talking about in front of the dwell pot. Have a recommendation on a value to start at? I have no idea where to even begin. I guess I could try 10K and go up or down from there, but 10K is pretty insignificant compared to a 1M pot, so maybe I'll try 100k and go from there.

Second, there is a really bad oscillation (loud low-frequency hum) that starts when I set the mix level too far off of dry. It starts when I set the pot to about 2 o'clock (12 o'clock being 50%) so still mostly dry but still pretty wet. Here is what that part of the circuit looks like: out from R9/V1 annode junction through 22nF coupling cap into one end of mix pot. The unit did not work at all without the 22nF cap. Any ideas?

[MakerDP]

Added a 100K resistor before the dwell pot. That killed the clipping. I'll probably try around 50K and then narrow it down to lowest value I can get away with.

I still have that real nasty low frequency hum. It sounds to me like it's feedback... got to do some reading on that.

Thanks to the suggestions here, I've got the reverb sounding the best it has sounded so far. When I kill that feedback and find the right frequency to roll-off on the lows then kill the ground hum this will be a nice little unit. Part of me wants to get it sorted out before I start my next build because I keep going back and forth on whether or not I want to add reverb to it or not. But it would be nice to not have to lug around an extra box when I play out...

[Mick Bailey]

I'm puzzled why the circuit needed the 330k grid leak adding back in plus the additional coupling cap C8. The pot itself ought to act as the grid leak and the DC isolated off V1a by C1. No harm in leaving it like it is, just one of those things that gets my curiosity.

Taking the dry feed off V1a anode to the mix pot has to be via a DC blocking cap to pass just the AC audio signal. You can take the feed off the top of the dwell pot just the same as adding a dedicated cap.

The hum may be LF oscillation being fed through the pan; either disconnect the input to see if it disappears, or gently damp the springs with your finger to see if it stops. You could try swapping over the driver transformer primary leads to change the phasing, or you may need to reduce the LF going into the pan by reducing the reintroduced C8. Many reverbs use a very small cap into the driver stage - 560pf or so with a standard grid leak of 1M.

[MakerDP]

Back from vacation...

yeah having to add the grid-leak confused me too... you'd think the Dwell pot would take care of it.

I'll try your suggestions for the LF hum.

My grounding-issue hum is definately tank or later. If I plug the reverb send into a speaker instead of going to the tank, I get a nice hum-free signal with only slight distortion at the extreme of the dwell knob.

I'm puzzled why the circuit needed the 330k grid leak adding back in plus the additional coupling cap C8. The pot itself ought to act as the grid leak and the DC isolated off V1a by C1. No harm in leaving it like it is, just one of those things that gets my curiosity.

Taking the dry feed off V1a anode to the mix pot has to be via a DC blocking cap to pass just the AC audio signal. You can take the feed off the top of the dwell pot just the same as adding a dedicated cap.

The hum may be LF oscillation being fed through the pan; either disconnect the input to see if it disappears, or gently damp the springs with your finger to see if it stops. You could try swapping over the driver transformer primary leads to change the phasing, or you may need to reduce the LF going into the pan by reducing the reintroduced C8. Many reverbs use a very small cap into the driver stage - 560pf or so with a standard grid leak of 1M.

[nisios]

You should try floating the transformer secondary ground. Grounding the reverb pan only at the output?

If you put a capacitor across like 200n or so (experiment) from Reverb output to ground you get a nice reactive treble boost.

Hope this helps.