Peach with two band tone control

[soggybag]

I was looking at the Fran Tone Peach Fuzz and noticed it used only a single op-amp. I thought I'd try a build with a dual op-amp, add a James tone stack, and use the extra op-amp to make up the gain at the output.

Check my work! I feel that this looks correct, but I'm still learning. I added everything after the second LM386. I also added a 1M from VR to the + input at each of the op-amps.

[antonis]

(after a rough glance..)

You don't need Zobel networks (R8/C7 & R10/C11) for LM386s..
You dont need R16 (1M bias resistor) for IC3A.. (but you DO need it for IC3B - unless you configure it as inverting amplifier, to result into no phase reversal between Input and Output..)
IC3B gain should be close to 8X (18dB) for Tonestack recovery..

P.S.
There are some weird item values (like R18/C21 subsonic HPF) but I let you decide for them..

[ElectricDruid]

You dont need R16 (1M bias resistor) for IC3A..

Yeah, what's going on with IC3A? It looks like an op-amp that can't make it's mind up whether it's an inverting amplifier or an non-inverting amplifier. Which is it *supposed* to be?

[antonis]

Yeah, what's going on with IC3A? It looks like an op-amp that can't make it's mind up whether it's an inverting amplifier or an non-inverting amplifier. Which is it *supposed* to be?

Just an inverting amp with 100/133 input attenuation and about 25k gain resistor (Rg in 1 + Rf/Rg gain formula..)

[ElectricDruid]

Just an inverting amp with 100/133 input attenuation and about 25k gain resistor (Rg in 1 + Rf/Rg gain formula..)

Ok, follow-up question - what's the point of having an attenuator followed by gain in this situation? Why not drop the attenuator and reduce the gain proportionally?

As it stands we've got a 33K/100K divider coming in, so we only see 3/4 of the input signal. The 33K and 100K act in parallel to give us the gain, which is roughly 560K/25K = x22. Instead, we could drop the 100K and get rid of the attenuator. If we do that, the "25K" input resistor suddenly becomes...33K! And ho and behold the overall gain remains the same. So what is the point of that attenuator? I don't get it.

[Rob Strand]

IC3A is wired as an *inverting* amplifier.

Ok, follow-up question - what's the point of having an attenuator followed by gain in this situation? Why not drop the attenuator and reduce the gain proportionally?

As it stands we've got a 33K/100K divider coming in, so we only see 3/4 of the input signal. The 33K and 100K act in parallel to give us the gain, which is roughly 560K/25K = x22. Instead, we could drop the 100K and get rid of the attenuator. If we do that, the "25K" input resistor suddenly becomes...33K! And ho and behold the overall gain remains the same. So what is the point of that attenuator? I don't get it.

Your calcs are entirely correct.   Both methods are *expected* to give the same gain.   The 100k (R2) factors out because it connects to a virtual ground point, *if* we treat C1 (100n) as a short.

Ignoring R2 gives the simplest method to get the gain.   What you don't get though is the correct cut-off frequency with C1.  The more complicated method which ends up with the 3/4 divider and the 25k impedance will give the correct cut-off.  (R2 is *before* the cap so it's not really a virtual ground point at all frequencies.)

The resistive divider will load the pickup differently compared to a straight 25k input resistor.

[antonis]

After all, I think Mitchell just followed the original circuit..
(so don't ask him about that "weird" configuration..)

https://www.flickr.com/photos/cow4prez/3314934163

[soggybag]

After all, I think Mitchell just followed the original circuit..
(so don't ask him about that "weird" configuration..)

https://www.flickr.com/photos/cow4prez/3314934163

Exactly, I wanted to build this for fun. I used the schematics from the internet. I used a different schematic, same circuit, it looks the the author of that schematic added zobel network following the app notes for the 386.

I saw the design and realized there would be a used op-amp and thought I should make use of it by adding a better tone control. The original tone control looks like the big muff type. I thought I'd use the James because I had used it before.

I open to suggestions. For an alternative active tone control that makes use of the extra op-amp.

I'm updating my schematic to fix the issues mentioned with the output op-amp.

[soggybag]

I looked zobel network. It looks like this is used to balance the impedance following changes with frequency?

I'm guessing me don't need this between the two 386s because the impedance is constant?

What about at the output, does the tone stack require a zobel before, or not because the impedance is constant?

[Rob Strand]

I looked zobel network. It looks like this is used to balance the impedance following changes with frequency?

I'm guessing me don't need this between the two 386s because the impedance is constant?

What about at the output, does the tone stack require a zobel before, or not because the impedance is constant?

The Zobel network is so the LM386 doesn't oscillate.  You can think if it as being part of the LM386. 

[soggybag]

Here is an updated version of the schematic. I tried to address the comments here.

[soggybag]

I looked zobel network. It looks like this is used to balance the impedance following changes with frequency?

I'm guessing me don't need this between the two 386s because the impedance is constant?

What about at the output, does the tone stack require a zobel before, or not because the impedance is constant?

The Zobel network is so the LM386 doesn't oscillate.  You can think if it as being part of the LM386.

I had the impression you would use this if driving a speaker. Sounds like you're saying that it would be required?

[Rob Strand]

I had the impression you would use this if driving a speaker. Sounds like you're saying that it would be required?

That's often the argument but it's not the whole picture.  Some amps can go unstable, or are less stable, with no load and the Zobel network ensures there is a resisitive load at high frequencies.  The datasheet is unclear if the LM386 is stable into *no load*.  My argument is if a circuit needs a Zobel network with with a speaker load then there's a good chance it needs it with no load.

[antonis]

But Rob, in OP's circuit, neither IC1 nor IC2 could work under no load..
(for IC1, pin3 of IC2 is connected to GND via a 50k resistor..)

[Rob Strand]

But Rob, in OP's circuit, neither IC1 nor IC2 could work under no load..
(for IC1, pin3 of IC2 is connected to GND via a 50k resistor..)

It doesn't have to be "no load" as such just something significantly higher than 4 to 32 ohm (what the LM386 is designed for, more or less).   (And capacitive loads can make things worse.)

The datasheet doesn't give any clear statements about not having a Zobel.   You might expect operating the LM386 at a gain of 200 might be less likely to oscillate than at a gain of 20 but even then there's nothing in the datasheet.   If you do anything outside of the datasheet it comes at your own risk.   Today's batch could be OK and tomorrows batch not.

[Rob Strand]

We had a power outage today so I had time to read over some books (yes, ancient paper technology).

I stumbled on  National Semiconductor Applications Note AN69, LM380.

https://www.ti.com/lit/an/snaa086/snaa086.pdf

It makes some very specific comments about the Zobel network on the LM380,

Page 1: "Note 4: If oscillation exists under some load conditions, add a 2.7 ohm resistor and 0.1 uF series network from Pin 8 to ground"

Page 3:, section "Oscillations" and Figure 7
"The (Zobel components) Rc and Cc shown as dotted line components on Figure 7 and throughout this paper suppresses a 5 to 10 MHz small amplitude oscillation which can occur during the negative swing into a load which draws high current. The oscillation is of course at too high of a frequency to pass through a speaker, but it should be guarded against when operating in an RF sensitive environment."

Page 7: Figure 7
(Zobel Network Required...) *For Stability With High Current Loads

You can see Applications note is very specific about what it does.  Seemingly regardless of whether the load is a speaker or not.  Also the problem occurs at high current, an only on the negative cycle.  No doubt this is something caused by the slow PNP transistor on the output stage.  It is unclear if the oscillations are due to loop stability or parasitic oscillation of the output stage - check out the schematic in the datasheet.

What is clear is this is a very specific problem.   The applications note implies " If oscillation exists under some load conditions".  While it does state the very specific conditions it doesn't state if there are other conditions.

When we look at the datasheet,
https://pdf1.alldatasheet.com/datasheet-pdf/view/8880/NSC/LM380.html

We see there parts are not optional, and the intricate reasoning in the applications not is not present.  After all it is good safe practice to include it.

So what's this got to do with the LM386?
- well the LM386 and LM380 come from the same family of amplifiers; there's five or so amps in the series
- it shows the datasheets don't always rationalize the details
- the reasons for including the Zobel on the LM386 could well be different
- it's good practice from feedback stability point of view if you don't want any unexpected surprises.
  It would be hard for a datasheet to pin down all the cases where it oscillated without the Zobel Network.

[antonis]

Good work, Rob..

Personally speaking, the only case using LM386 for purpose other than speaker driving was for Vref formation (both inputs tied to GND, internally forcing output to mid supply rails) with absolutelly no issues..
(no Zobel network or anything other oscillation preventive configuration)

[Rob Strand]

Good work, Rob..

but a 100% fluke to stumble on it!

Personally speaking, the only case using LM386 for purpose other than speaker driving was for Vref formation (both inputs tied to GND, internally forcing output to mid supply rails) with absolutelly no issues..
(no Zobel network or anything other oscillation preventive configuration)

Yes, it's very true.  Most audio power amp data sheets only imply the intended use is with a speaker and occasional use with for Vref/2 supply splitters or a "power oscillator".   More often than not:
- very little in terms of op amp type specs
- less than half even give slew rate
- they give distortion figures (op amps usually don't)

I think the supply splitter idea was stolen from the LM380 applications.   No Zobel but 10uF caps hanging off both rails! - something which would be unwise for op amps, and probably for many power amp IC's.

IIRC, the old National Semiconductors Audio and Radio Handbook / Audio handbook had some stuff on the LM380, LM386 family of amplifier - not sure what year book had moreinfo.   There was another family of amplifiers, LM378 etc, which were a different design and appeared in those books.  I think they came before the LM380 family.   Those books might have more info.


EDIT:
National Semiconductors Audio Handbook 1977, chapter 4, section 4.7.6, p 4-31.   It states the Zobel network on the LM386 is to prevent negative swing oscillations when drawing "high current" on the negative swing.  It then references section 4.5.5 which makes the same comments about the reason for the LM380 Zobel network as the AN69 document I posted before.  The LM386 needs a Zobel for the same reason as the LM380.

We could summarize we might be able to get away with no Zobel with light loads on the LM386.   However, in chapter 4 the general theme for all the amplifiers is if you have oscillation problems: try a bypass cap across the rails, then if that doesn't work add a Zobel network.   It never implies adding the Zobel network causes oscillations or does anything bad.

[soggybag]

Thanks for the analysis, most of it is over my head, I like to think I pickup something along the way. I added the zobel cap and resistors again. They can be omitted on a build or included.

[PRR]

> it is good safe practice to include it.

But... but... buttt... that adds 30 cents to the cost of my build!!! You are going to bankrupt me! (Like the Silicon Valley Bank run today.)

I gave-up preaching Applications Note AN69, LM380, which is on my bedside shelf. Nobody believes stuff written on paper anymore. (Also for some years AN69 was missing from the WWW, so existed only in the so-called memory of old geeks.)