Reverb Wet/Dry Mix Problem

[dchang0]

This actually feels like the right explanation, because of the way the volume on both wet and dry signals just rolls sharply off after a certain threshold. Would you mind explaining your reasoning in a bit more detail to help me understand? When you say plate impedance, do you mean of the triode-strapped-pentode (reverb driver) or of the triode (reverb recovery) or both? It's not likely the recovery stage, as that continues to grow louder as the reverb pot is turned up.

Or are you referring to the halves of the 12AX7?

I'll give your suggestions a try, but in a week or so after I get the replacement 6BM8. The microphonics are so extreme that I'm having a hard time not kicking off high-pitched feedback on certain notes--which of course are in the low frets and are frequently played.

Also, what pot value would you suggest if 500K is too high? 100K? 250K? Doesn't the reverb pot's lower half (below the wiper) end up serving as part of the collective grid leak resistance presented to the 2nd half of the 12AX7? If that is the case, reducing the pot would decrease gain in that stage, right (not necessarily a bad thing, being after the reverb)?

I obviously need some help understanding this whole "network of resistors" thing.

[dchang0]

Hey, guys--just for fun, here's a photo of the cramped confines of this amp.

[dchang0]

Yet another update--I think I may have proven Chuck and Enzo right. If I remove the 6BM8 tube completely and turn the reverb pot up, at 7.5 to 10, it starts into a low "farting" oscillation. (Unless of course this is supposed to happen if I remove the reverb driver and recovery tube.) Funny thing is that the oscillation is not apparent when the tube is in (and the 470K send leg resistor is in place) unless that low 60Hz hum IS the oscillation.

If layout is the issue, well, I may be doubly-screwed. Aside from having a teeny space to work in, I don't know anything about layout "theory," or what makes for a quiet, oscillation-free layout design.

[pdf64]

'When you say plate impedance, do you mean of the triode-strapped-pentode (reverb driver) or of the triode (reverb recovery) or both? It's not likely the recovery stage, as that continues to grow louder as the reverb pot is turned up.'

Yes, the recovery stage.
Are you familiar with Thevenin analysis? Basically, the plate-cathode of that triode looks like a ~40k resistor, as far as the rest of the amp is concerned. So as the reverb is turned up, the 500k pot track resistance gets turned into a 40k resistance, which acts as the bottom half of a potential divider, the top half of which is the 100k dry mixer/dropper resistor.

'it starts into a low "farting" oscillation'

That may be 'motorboating', caused by insufficient power supply decoupling - try increasing the value of the 8uF caps.
But from the schematic above, I can't see how the reverb control could affect anything without a recovery tube in place. Has anything changed, compared to the schematic?

[dchang0]

Wow, "motorboating"--what an awesome term to describe that sound! That's exactly what it is... I should note that it only happens if I pull just the 6BM8 and leave the 12AX7 in. With just the rectifier and 6V6, there's no microphonics, hum or motorboating even with the reverb cranked all the way up.

As for Thevenin analysis--probably did it years ago on super-simple textbook examples. But I get what you mean now that you explained it.

The current circuit is exactly as this schematic shows. The 470K resistor effectively cuts off the top "bad" range so that I can now turn the pot all the way up and get a pretty good mix, albeit with lots of 60Hz hum and microphonics from the tube.



Here's the new signal curve:



It's actually quite acceptable, sound and mix-wise. I'm sure the amp could run far better in terms of tone and efficiency than it is now. The only audible problem is the microphonics and 60Hz hum.

[dchang0]

Well, it might not really be motorboating, because it responds to the mix settings. With all the tubes in, if I turn both the volume and reverb pots, at some combinations, the low throbbing oscillation (not quite a putt-putt-putt) will come in and out. It most noticeably appears with the volume at 9/12 and the reverb pot at 10/10. The motorboating will not happen if the reverb is turned below 8/10.

I went back in and reduced the gain on the recovery stage by removing the cathode bypass cap. It cut out a lot of the annoying high frequency noise and microphonics (and also some of the high frequency response).

I am also contemplating reducing the driver side gain too by doing the same thing. One benefit of removing the relatively large cathode bypass capacitors is that I can dramatically shrink the size of the reverb daughterboard and move it to the right, away from the power section. I'll probably lose a lot of the reverb effect, though.

A few questions before I go to the expense and trouble of building a whole new eyelet board:

1) Would it hurt the reverb's tone if I replaced the gigantic 470nF Orange Drop coupling cap with the more reasonably physically-sized 22nF Orange Drop? As I understand it, both high frequencies and low frequencies don't sound good being driven through a reverb tank. I really want to cut the daughterboard to about half of its original length and get some distance between the circuitry and the power section, and this is only possible by stepping to physically smaller caps.

2) What is this sound if it isn't really motorboating but is a really low audible-frequency oscillation, and why would it respond to the particular mix?

3) If I'll be ripping out the eyelet board, I'll have a rare chance to change the reverb pot. Which do you recommend: 10K through 250K, and what taper? Also, I might as well aim for the right mix resistors, etc. What values are optimal? I'm guessing lower is better (100K).

4) Any tips on layout before I punch my new eyelet board?

[dchang0]

Removing the 50uF cathode bypass cap on the driver side reduced gain too much and cut out the highs. The surf tone was gone. I'll have to keep it.

[dchang0]

Okay. I went and looked at these two schematics and borrowed some elements of them.

g12.pdf

500-R2.pdf


The 500 R2 is a nearly identical usage of the 6BM8 triode-strapped-pentode. Unfortunately, the mix values differ greatly from what I need because they pick up the reverb right after the guitar input. They use a 50K reverb pot, no mix resistor, and a 47K send leg resistor.

So I went back to the Ampeg G-12, which has a very similar topology to mine. Obviously, their 500K reverb pot and 270K mix resistor with 150K send leg aren't working for me. (Mine is 500K pot, 1M mix, 470K send leg as of this moment.)

And I took your suggestions and made this new temporary schematic.

I have no idea whether the mix of this one is going to be right either, so I'm going to include small 1M trim pots to allow for tweaking the mix and send leg resistances. I cannot easily change the reverb pot once installed, so I'll go with a middle-of-the-road 100KA.

If you guys see any errors, please let me know. Thanks again for all your help!

[pdf64]

Yes 100k would be my choice, probably linear taper given that you like the surfy reverb levels.
I wouldn't go below 470nF for the coupling cap to the tank driver coil, but if 220n works for you then that's fine.
If your schematic is accurate to the amp, then I can't see how the motorboating can really be motorboating. Try increasing one/both of those 8uF B+ caps.

[dchang0]

Thanks--100K it is. I think I might detach one end of the reverb circuit, turn the amp on, and then measure its total operating resistance (empirical Thevenin analysis) to get an idea of what I'm dealing with.

Hmm. The 470nF cap is one thing I haven't been able to find any theory/explanation on. Why that value, of all values? Is it for coupling efficiency and/or tone shaping? All of Ampeg's designs use a 470nF AND also Merlin's one example on his website with the cap-coupled paralleled ECC82s, but there's no explanation why.

As for the 8uF caps, I can't change them without redoing the eyelet board for the main 5F1 circuit. Going from 8uF to even the 10uF in a Sprague Atom jumps from the skinny/long black can to the fat, light blue can form factor. Speaking of which, I'll order some extra blank eyelet board and a pair of 16uF caps (the 16uF is the largest physical size that can fit in the chassis). If I've got to go that route, then I might as well use three 16uFs.

Yeah, it doesn't seem like motorboating, because it actually behaves like feedback. If it turn the knobs to certain values, then hit certain notes, I can induce the LF oscillations, which fade in and out just like typical guitar feedback does. I could, if I wanted, actually "play" it somewhat, causing it intentionally. I figure it's because my current layout has the recovery side 0.022 coupling cap too close to the power amp. It is literally right above the 16uF and 8uF caps and is in parallel with them.

[dchang0]

More diagnostics:

I took my guitar and connected it directly to the reverb recovery circuit and found that it is heavily distorted. Then, I took the same guitar and hooked it up to the recovery section of my Princeton Reverb, and it is clean, with a teeny bit of soft clipping at the top.

So it looks like there is too much gain in my recovery circuit, introducing low-frequency and high-frequency feedback. I merely need to reduce the gain from that stage. Next, I'm going to try a lower grid leak than 1M to see if that helps. If it does, then I will redesign the recovery section to perform cleaner (reduce 180K plate resistor to 100K, etc.).

EDIT:

Hmm. Well, I reduced the grid leak on the recovery stage from 1M to 220K, and it helped some. It cleaned up the distorted reverb, reduced the overall reverb effect, and slightly reduced the reverb in the mix. So then I put the 1M back and reduced the 180K plate resistor to 100K to reduce the gain, and instead, it got dirtier. Interestingly, it made the LF feedback much stronger, so perhaps the gain wasn't really reduced. I'll probably have to change all three resistors (plate, cathode, grid leak) to get a clean tone out of the recovery section.

UPDATE:

LOL--well it dawned on me that the overdrive must be coming from the 6V6 and the second half of the 12AX7. By running the guitar directly into the reverb recovery, I had effectively copied the 5F1 circuit, but with a 6BM8 triode for the first stage instead of the 12AX7 triode. It does mean that the reverb pot, if turned all the way up, will NEVER produce clean Fender surf tones, because the power section just isn't clean enough to pull it off.

One good thing that came out of this is that I have definitely proven that the noise and hum are coming from the reverb recovery circuit and nothing else. I also proved that the low-frequency feedback is limited to that section, since changing its gain/bias points directly affected the feedback.

I suppose that I need to reduce the gain in the second 12AX7 stage to clean it up and then raise the curve on the 6BM8 triode to clean that up, leaving the first 12AX7 stage and the final 6V6 to do all the 5F1 overdrive.

[Chuck H]

Your actually not running the recovery stage ideal to the tube data. You might do well to lower the plate resistor to 100k and perhaps raise the cathode resistor to a 2.7k.

And, FWIW, the circuit can't be putting more than about 4V or 5V of reverb into the 6V6 grid. So I think any distortion in the reverb may be due to the bias condition of the recovery stage.

To tame the system you could also RAISE gain in the recovery stage and then dumb it down more. As in, add a cathode bypass cap to the recovery stage and use a larger series resistor and maybe even a 50k pot.

[dchang0]

Hi, Chuck--

Yeah, I did the math and figured out that a 100K plate resistor matches with a 4.7K cathode resistor for a slightly colder than center bias point of -3V. That should result in a clean, lower-gain recovery stage.

However, I did discover that nearly all of the overdrive is coming from the 2nd half of the 12AX7 and the 6V6, so without an oscilloscope, I have no way of determining how clean the reverb recovery is on its own. Based on my calculations, it should already be pretty clean even as it is with the 180K plate resistor and 1.5K cathode resistor.

What you say about taming the system actually proved true. When I put the 50uF cathode bypass cap back in on the recovery stage, all the low frequency feedback disappeared. So my schematic is going back towards the original theory-based design, with only two problems outstanding:

1) I don't know how to calculate the mix network, so I'm just guessing. I've ordered a 250KA pot and a 100KA pot to play with. I'll also stop by the local electronics store to pick up some 1M trim pots for tweaking purposes.

2) More and more, things are pointing towards the layout as the cause of the weird feedback. I think I put the 0.022uF coupling cap too close to the 16uf and 8uF filter caps (and in parallel). Looking at the 5F1 layout, it occurred to me that the coupling caps are separated from the filter caps by the cathode circuitry of the three valves. So, I'm drawing up a new layout that purposefully makes the grid wires as short as possible and puts the coupling caps roughly above the lower circuit board's coupling caps, about 1-2" away from the filter caps.

Well, at least I get to sleep while waiting on parts to arrive. I think this may be my third all-nighter on this amp, LOL. Thanks again to everyone who helped!

[Chuck H]

Well, get some sleep. I know how it is when the bug bites you.

Something else to consider is that often times motor boating can be fixed by changing the filter value and/or the filter ground location for individual stages. With as much going on in there as you have now you should probably have two ground points or a very careful buss ground terminated near the input, not near the power supply as many do. You might also try re routing or even removing the NFB lead. At least lift the NFB from the jack end and see if it helps. It could be a clue as to how you might stabilize the amp overall.

[dchang0]

Will do. Sleeeeeeeeeeep...

Anyway, I am very sure there is a ground problem, as there has always been a low 60Hz hum throughout all this testing. It is very likely a loop between the reverb ground (serviced by (B) B+) and the 12AX7 ground (serviced by (C) B+). The reverb ground is currently located at the (C) ground instead of further down the bus, at the (B) ground. I did that because I literally could not reach the bus wire there with my soldering iron after I had installed the filament wiring, pilot light, and virtual center tap.

Thanks again, and I'll be back in a week or so when all the goodies get here.