Shunt NFB and gain Q

[Chuck H]

I'm working on a reverb coupling circuit. I'll be using a shunt feedback "virtual earth" type mixer to reduce the typically high series resistance in reverb mix circuits. I've done this type of mixer before with good results. But this time I'll be imbalancing the input resistors to achieve some amplification of the reverb signal while keeping the dry signal at unity gain. I'm trying to arrive at an aplification of about 25 for the reverb. So I'm wondering what shunt resistor/dry signal input resistor value will get me there if I don't use a resistor for the reverb at that input.

I'm not strong on formulas though my math skills are fine. I just didn't stay in skool long enough to learn squiggles, dots, slashes, brackets and such. So explained math or simple analogies would be great. Typing slowly and loudly also helps.

TIA

[pdf64]

Are you looking to use a reverb recovery triode, as per regular BF Fender?
A gain of 25 may be over ambitious.
Given 1M feedback and dry input resistors (would need 40k reverb input resistor, but 12AX7 output impedance is about that).
A sketch of 'building block' signal flow would be handy (where's the reverb intensity control?).
Pete

[Chuck H]

Thanks Pete. I thought a gain of 25 was a lot to ask. It won't break my heart if it's not quite as deep as a BF type reverb though. Who ever uses one of those on ten?!? Here's a MSpaint sketch of my idea. The dual pot is a linear taper. Arranged like it is I expect the circuit to behave more like an audio taper. This circuit should keep the impedance constant at the mixer input... I think. I'm just wondering how low I can get the series resistance (with a corresponding change in reverb pot value). As in, a pair pf 220k resistors (dry feed and shunt) and a dual 250k pot, or a pair of 100k resistors and a dual 100k pot, etc. Of course I can just gear up with the parts and experiment. But since I sort of know my gain goal I'd like to minimize tweaking by knowing roughly how to get there in the first place.

[Wombaticus]

Thanks Pete. I thought a gain of 25 was a lot to ask. It won't break my heart if it's not quite as deep as a BF type reverb though. Who ever uses one of those on ten?!? Here's a MSpaint sketch of my idea. The dual pot is a linear taper. Arranged like it is I expect the circuit to behave more like an audio taper. This circuit should keep the impedance constant at the mixer input... I think. I'm just wondering how low I can get the series resistance (with a corresponding change in reverb pot value). As in, a pair pf 220k resistors (dry feed and shunt) and a dual 250k pot, or a pair of 100k resistors and a dual 100k pot, etc. Of course I can just gear up with the parts and experiment. But since I sort of know my gain goal I'd like to minimize tweaking by knowing roughly how to get there in the first place.

Chuck,

I've used exactly this type of mixer in my Wombat Amps "Grail" design, and it works very well, although it's not as dripping wet as a BF...

See the linked schematic for the Grail.. note though that the feedback resistor was upped to 470K, and the feedback cap was increased as well, I think to 47nF or 100nF, can't recall at the moment):

Wombat_Grail_revH.pdf

You'll note that there is no mix resistor at all for the reverb, as I needed this gain to be maximized, and it is relying entirely on the output impedance of the tube for the series resistance into the virtual earth. This reverb comes late in the amp where signals are already quite big, so I ended up having to use a surprisingly large mix resistor on the dry to balance it appropriately with the wet. 705K was more than I would have liked, but it works well in practice and still is a lot better than the 3.3Meg seen on some Fenders! This could be a lot smaller if the reverb return was done a little bit earlier in the chain where the signals were smaller.

The other fun thing here is the use of a footswitched vactrol to mute the reverb. I used the VTL5C1 with a dark resistance of 5Meg, but the 50Meg of the VTL5C3 would be even better. Incidentally, the driver is Merlin's transformerless circuit, and makes use of a high impedance F-series tank.

[Chuck H]

Indeed. As you can see by the "to PI" indication on the diagram this circuit will be the last thing before the PI. I may even find that I need to do a little dry signal padding. That would make things even easier. I'm sure my hope for a tweak-less scenario is pie in the sky. Still... If the approx amplification factor of a 12ax7 w/100k plate (Vp 180 to 200), 1.5k fully bypassed cathode is roughly 60, what happens to that amplification factor if I wrap a 470k resistor around it from plate output back to grid? I know that if I feed the grid through another 470k that the result is unity gain. And I know that the load at the grid makes less difference because the FB voltage division is offset by the output. But how can I know what the amplification factor will be for this stage when NOT feeding the grid through a resistor??? I have a JL Hood book with a good write up on feedback that I must revisit. But it went terribly over my head the last time I tried to digest it.

[Wombaticus]

Indeed. As you can see by the "to PI" indication on the diagram this circuit will be the last thing before the PI. I may even find that I need to do a little dry signal padding. That would make things even easier. I'm sure my hope for a tweak-less scenario is pie in the sky. Still... If the approx amplification factor of a 12ax7 w/100k plate (Vp 180 to 200), 1.5k fully bypassed cathode is roughly 60, what happens to that amplification factor if I wrap a 470k resistor around it from plate output back to grid? I know that if I feed the grid through another 470k that the result is unity gain. And I know that the load at the grid makes less difference because the FB voltage division is offset by the output. But how can I know what the amplification factor will be for this stage when NOT feeding the grid through a resistor??? I have a JL Hood book with a good write up on feedback that I must revisit. But it went terribly over my head the last time I tried to digest it.

Randall Aiken has a really great tutorial on the design of this type of inverting feedback amplifier.

The key formula though is: Acl = (Ra + A*Rf) / (Ri + Rf + Ra - Ri*A)

Where Acl = closed loop gain, Ra = plate impedance, A = open loop gain, Rf = feedback resistance, Ri = input resistance

In the case where there is no mix resistor, just use the output impedance of the prior stage as Ri.

Things are slightly more complex when you are using multiple inputs, and I find it's easiest just to sim these things.

[Chuck H]

I find it's easiest just to sim these things.

Yep. Once I get my head around a concept I find that ball park is usually damn close enough for guitar amps. That is, you can sort of guess at the approximate contribution of peripherals, use the main contributing factors and end up really close on paper to how things actually value out. I don't have any SPICE programs but I do rely on a number of on line calculators.

EDIT: So, in my formula impaired state, I've calculated a gain decrease of approx 1.89dB for the reverb. That would still achieve more amplification than I think I'll need. I think for the actual design it's going to be by ear anyway. So I'll probably temporarily install pots to act as the shunt/series resistor for the dry signal (dual ganged) and the load for the reverb. Then I can dial in the reverb balance I like and guarantee the lowest possible series resistance. I had hoped to add a tool to my cranium but I may not quite be there yet.

[J M Fahey]

Just curious as to why you use "Op Amp" style configurations in Guitar Tube circuits.
"Defined_by_NFB_net" gain blocks work very well and exactly as intended when used with Op Amps that by definition are supposed to have infinite gain, zero output impedance and infinite input impedance.
Plus having inverting and non inverting inputs .
Only parameter Tubes comply with is infinite input impedance, but are severely lacking (that's an understatement) in the other 3.

[Wombaticus]

Just curious as to why you use "Op Amp" style configurations in Guitar Tube circuits.
"Defined_by_NFB_net" gain blocks work very well and exactly as intended when used with Op Amps that by definition are supposed to have infinite gain, zero output impedance and infinite input impedance.
Plus having inverting and non inverting inputs .
Only parameter Tubes comply with is infinite input impedance, but are severely lacking (that's an understatement) in the other 3.

Why not? It's true that the virtual earth that one obtains with a tube isn't nearly so good as what one might get with an op amp -- but it still adequate to accomplish mixing with negligible bleed between channels, while allowing the flexibility to get a bit of gain (or attenuation), if that's what you need. The inverting local feedback stage also works well as wet/dry mix for an FX loop...it adds little colouration and has good bandwidth. It's a handy thing to have in the bag of tricks.

[Gingertube]

J M - The virtual earth mixer works well within its limitations, so I guess you were really asking "how bad are its limitations".
The only real limitation I,ve experienced are to make sure you choose the input resistor values such that you don't overdrive the triode (else you have no gain and no virtual earth anymore either). Otherwise the only other limit is on channel separation. I don't recommend trying to mix more than 3 signal sources. In any event the virtual earth mixer will always give you far superior separation than a passive mix scheme.
Cheers,
Ian

[Chuck H]

Truth be known I am building this amp with particular aesthetics in mind. It is to be an all tube circuit. However, the charms of what tubes can do and the charms of what SS can do are not lost on me. Since all my personal amps are pretty basic, none having trem and/or reverb, I am planning to build one of this particular design for myself. I guess I'm broadening my horizons at this time. This amp will switch between Vox TB and Fender BF type preamps with a corresponding switch between NFB (Fender) and no NFB (Vox) with the addition of the "high cut" circuit in "Vox" mode. Add a bias mod trem and on board reverb. It's intended for mostly clean to only slightly distorted tones. And it sounded like a really fun amp to me. The one I build for myself will use SS devices for reverb recovery, the mixing stage and the trem coupling circuit. And it will sound just as good as the amp I'm building now. So... For the moment let's just ignore the fact that I'm not using any SS devices on this particular build. I really like building with tubes anyway and it makes another criteria for this build (eyelet board) more plausable. Ever try to install an eight pin op amp on an eyelet board?!? Anyway, let's just roll with it. Peace?

[J M Fahey]

Peace *is* the idea.
It was a friendly comment, and addressed specific strenghts and weakness of different options.
That said, you'll be amazed, but I started building amps in 1969, tubed and on eyelet board.
In 1972 I switched to SS ... and continued with the technique I had mastered for years: eyelet boards !!!
I already posted here a (still working) 1972 vintage 200W into 2 ohms amplifier ... on eyelet board.
Too sleepy now to find its picture, but it'll be here tomorrow.
Now the best part: I started building discrete transistor preamps, "inspired" in Ampegs, straight out of Jack Darr's book ... but later started building my own designs , with then new and advanced uA741.
Still on eyelet boards.
How come?
I used round, metallic case , 8 long legs uA741.
They were a novelty and cost as much as a 12AX7, go figure.
I made Guitar amps, Bass amps and 6 channel PA mixers with them.

[Chuck H]

So... I was one year old when you started building amps.!.

I just put a pair of eight leg cans (ca3080a's) into that Kustom amp I posted about before. I actually don't think it would have been any tougher mounting them on eyelets! The Kustom board had two sets of holes for these can/chips. Get the leg in the wrong hole and you were poking at the mounted side of the solder pad! I had to be very careful and spent much time making minute bends in the legs.

I was surprised that, as you say they were novelties then, they are novelties now! People seem to love them so I paid extra for the privilege of owning two. That and a 14pin dual op amp chip that cost twice as much. What's next? Sites that specialize in "NOS" SS devices? With the same sort of "Manufactured by Texas Instruments for US Telecom." descriptions like you see supporting NOS tubes.

[J M Fahey]

Well, we are seeing some of that.
Starting with crappy germanium transistors !!!
If you wnt to repair "original" your transformer driven, mid 60's Vox amp, or some early Rhodes amps, you'll have to pay through the nose.
Same with some obsolete chips.
Try to find the ¿TL604? (or something like that) CMOS switches favored by many in GK, some Peaveys and even some Kustom (and any other amp made in that age).
Even CA3080 are getting hard to find.
Or the once very popular uA4739, the "first" dual Op Amp designed *specifically* for audio.
With external compensation caps and pinout incompatible with any modern Op Amp.
They were the IC of choice in Craig Anderton's excellent book.
Here's my eyelet board built 1972 amplifier.


You gave me an idea, I'll design and post an eyelet mounted SS power amp, for PCB challenged "Toob builders".
100W SS amp, loud and clean, *EASY* to build and repair, what do you think?

[Wombaticus]

Hey JM,

I forgot to mention earlier that there is another place where you not uncommonly see this type of inverting local feedback stage in a tube circuit, albeit in a slightly disguised form. No-one ever seems to mention it explicitly, but the floating paraphrase inverter essentially uses local feedback to set the gain of V2.