Rail bootstrapping design guide and theory?

[Fancy Lime]

Hi everyone,

and Merlin in particular, if your reading this.

I've had some trouble with peak clipping on a (very) resonant autowah circuit, leading me on a hunt for maximum headroom. I experimented with different charge pump setups: +18V <> 0V and +9V <> -9V, each from a ICL7660 and a TC1044S. All configurations were extremely noisy, absolutely unusable. This was all done on the breadboard, so that may contribute a little. Anyhow, I got frustrated and decided to put further experimentation into charge pumps on hold and explore alternatives first. I found this thread:
http://www.diystompboxes.com/smfforum/index.php?topic=91629.msg782225#msg782225

And that does seem like a much more elegant solution. Kudos to Merlin for yet another brilliant contribution. The concept is easy enough to understand, it works basically like a charge pump that uses the music signal instead of an independent constant frequency, which should make it inherently less prone to additional noise, while also adapting the rail pumping to the volume of the input signal. To me this seems so incredibly brilliant in it's relative simplicity and so amazingly useful that I'm really scratching my head how come we're not using this all the time instead of charge pumps. Sure, there are some things a charge pump can do, that bootstrapping cannot do easily (like putting put several different voltages with reverse polarity etc.) and bootstrapping of transistor circuits is not exactly news. But I had never seen it done this way and I'm wondering, by bootstrapping the rails, could one increase headroom for more complex circuit, which would remove the need for a charge pump on highly dynamic circuits.

So my questions are:
- Is there a good reason that I'm unaware of, why this is not done more often?
- Are there any potential pitfalls when trying to pump the rails of an entire complex circuit with several op amps and transistors like this?
- Can someone point me toward a good resource for design guidelines, especially in terms of scaling the bootstrap circuit to meet higher current requirements?
- How is it even possible I have never seen this done to a Klon Centaur clone? I though everything that could even remotely be considered a "boutique mod" had been done to those. And this thing here sure sounds like one of the boutiquest mods I can think of, right after building all the op amps from discrete hand selected Ge transistors in evacuated glass housings.

Thanks,
Andy

[Fancy Lime]

Really guys, no-one? Is this such an obscure topic or is it too pedestrian and I'm not getting it?

Cheers,
Andy

[EBK]

I've looked into this in the past, but I'll have to dig a bit to see if I can find the resources that seemed pretty good before.  I was researching with the aim of figuring out how to misuse bootstrapping for some modulated distortion.   

(Having two small children means I can't always answer when I want to...  )

[R.G.]

It's not too obscure or too pedestrian. It's just a complication with issues of making the power rails agile in an environment where many people think of a one-transistor rangemaster as complicated, and we have long discussions about how to do a mute switch with only one transistor.  It's also a complication in an environment where the circuits being fed the bootstrapped rail may or may not have any significant power supply rejection ratio, and so any noise from the bootstrapping may feed directly into the audio lines. There are also timing issues. Ideally, you'd like the power supplies to rise just before the signal peaks rise. Since we don't have any negative time diodes, you have to make the power rail rise faster than the signal does, or delay the signal by some time to raise the power rail.

Not a problem, not too complex, but the usual issues of appropriate and practical complexity arise.

[Transmogrifox]

So my questions are:
- Is there a good reason that I'm unaware of, why this is not done more often?

More parts?  If you're doing anything interesting with each op amp stage, you pretty much need one IC per stage (think a bunch of single op amps and transistors to accompany each).  Changing frequency response between stages means the timing and amplitudes of the ups and downs changes, and then they are no longer synchronized for bootstrapping to be of any real benefit...well you might be able to generate some interesting intermodulation distortion like this.

With careful design you could bootstrap multiple devices as long as you knew one stage would never need more than the bootstrapping stage gave it.

- Are there any potential pitfalls when trying to pump the rails of an entire complex circuit with several op amps and transistors like this?

See above.  Changing frequency response between stages such that the timing of ups and downs no longer correlates.

For example, a state variable filter would probably need 3 single op-amps, each bootstrapped...or you would get rails changing out-of-phase with the signal on some of the devices.

In the same vein, using 2 op amps in the same package to invert phase between stages.

- Can someone point me toward a good resource for design guidelines, especially in terms of scaling the bootstrap circuit to meet higher current requirements?

I don't know of any resources, but P=VI is a good place to start.

- How is it even possible I have never seen this done to a Klon Centaur clone?

I don't know.  You can be the first.

[PRR]

> peak clipping on a (very) resonant autowah

Reduce gain. (Peak gain should probably be unity.)

Add a limiter.

If you use a real strong magnifier, any little thing is going to be too big.

[Fancy Lime]

@ PRR
Hi Paul. That is in fact exactly what I ended up doing for this particular circuit. Schematic is here:
http://www.diystompboxes.com/smfforum/index.php?topic=118788.0
By reducing the gain systematically between all stages and checking when the clipping disappeared, I was able to put the blame on the OTA. So I lowered the level before it by a factor of 0.5 and raised it by 2.4 after the OTA. That killed almost all peak clipping. Adding the LEDs before the JFET input buffer to care of the rest. It seems to act as a primitive peak limiter, which clips in itself of course but way less unpleasant than the OTA. To be quite honest I don't really understand how the LED to ground thing works in this position but it seems like it does. Runoffgroove use this a lot.
So for this project, the rail bootstrapping is not really relevant. Just asking for future developments. The reduction and make up of gain of course carries with it a noise penalty. This isn't a problem with an envelope controlled low pass, because this thing is its own noise gate (neat!). But for clean preamps, EQs and the like it may be nice to have another way of increasing headroom. I had also considered a compandor and even bought a NE571 but ended up not trying it because the static gain adaption worked so nicely.

@ Transmogrifox and R.G.
Well the phase inversion thing does not seem like a big obstacle. One could have two bootstrapping circuits, one for the op amps which handle inverted and one for those that handle non-inverted signal. Would make thing complicated and take some particular design choices though. Might not be practical in all that many situation. Phase delay seems like a much more series issue, as long as we are talking about "normal" bootstrapping. I guess my original question should have included "is it possible to keep the rails pumped throughout the signal swing?".

My train of though was this: If we use the signal to shuffle charge around from some large capacitors to the rails (as the rail bootstrapping does), could we not by clever use of another capacitor and diode(s) for each rail see to it, that rails stay pumped until the next pulse comes in (like in a peak detector or envelope follower)? Then the whole thing would work like a "normal" charge pump, not so much like a classical dynamic bootstrap. That would of course require that the pump can deliver more current on each cycle than the op amp consumes until the next one. And I'm more than a bit fuzzy on how to calculate the charge that the pump can deliver, to put it mildly, hence my asking for design guides. If we had very high frequencies, like the 20-50kHz used in most charge pumps, current draw is much less of an issue because new charge comes in so often. So maybe a better way of asking the question would be "why does this not work". Because if it did work like I think it might, someone would be using it.

Cheers,
Andy

[Transmogrifox]

Now you're not really talking about bootstrapping any more.  What you are describing sounds more like this:
http://www.gyraf.dk/schematics/Voltage_multipliers_with_CMOS_gates.pdf

I have played around with using the audio signal to pump up rails. 
Lessons:
1) You can't use an op amp to pump up its own rail. The bootstrapping works because it pushes the negative rail and positive rail up at the same time -- sort of the same idea as moving the ground reference around to shift with the signal, but boot strapping doesn't make VDD-VEE > VDDin - VEEin.
2) You can use an op amp to pump up the rail of another op amp.  This is a charge pump.
3) You can use an audio signal as the "charge pump oscillator".  This responds slowly to the incoming signal so you will always clip the onset of transients.
4) Charge pumps can be quiet if properly applied.  Good power supply filtering and decoupling and good layout (not a breadboard).

You might find a hex-package CMOS inverter to be a cheap alternative to a charge pump if you want to do it quick & dirty.

Actually, the idea of dynamically adjusting power supply rails for headroom reminds me of "Class T" amplifiers.  They gain a little bit higher efficiency because they use the incoming audio to dynamically adjust the switch-mode power supply regulation set point.  This keeps total voltage drop across the output transistors to a minimum.

So your idea of using the audio signal as a charge-pump might have some value for low-power design...but this is a fundamentally different thing from bootstrapping like Merlin implemented on the Glass Blower.  The Glass Blower is a really good example of the kind of circuit that can make effective (and practical) use of power supply bootstrapping.

[Fancy Lime]

Hi Transmogrifox,

aha! That's exactly the info I was after. So my gut feeling that it cannot possibly be a simple as I first thought was right. I think I'll try and apply the normal rail bootstrapping to a Rat or TS type overdrive and see what happens. I have done some experimentation with MOSFETS as clipping diodes. Their threshold is quite high before they go into full clipping and with a normal 9V supply it is not easy getting them to do that there without clipping the op amp too. Maybe that is where bootstrapping can help.

So in summary, this whole topic is more something for specialized stages rather than whole circuits. Still worth keeping in mind though.

Thanks,
Andy

[merlinb]

If you're doing anything interesting with each op amp stage, you pretty much need one IC per stage (think a bunch of single op amps and transistors to accompany each). 

This is the main reason you don't see it used, IMO. I have done it with a dual package, but it can be squirrly. Some opamps won't like it.
Having said that, a resonant circuit is surely one of the best applications for this technique. When you don't want to attenuate the rest of the audio band before/within the resonant circuit you can bootstrap its rails, then attenuate after the signal has emerged from the opamp (so not everything in the pedal needs to be bootstrapped, just the resonant opamp).

[anotherjim]

Have thought of it to get N-channel JFET gates negative without a fixed negative supply, but not had to use it yet. Have since seen designs that do just this. Getting a noisy delay to shut up when there's nothing for it to delay being a case in point.

True "parasitic powering", that is using signal energy itself, and not just as a charge pump oscillator, is fairly common in other fields. Radar transponders can be entirely worked in the presence of the correct radar beam, powering themselves from the antenna signal enough to transmit their message.

You could rob power from speaker outputs and run pedals, but you'd need a re-chargable battery or super caps to make it reliable enough.

[Fancy Lime]

If you're doing anything interesting with each op amp stage, you pretty much need one IC per stage (think a bunch of single op amps and transistors to accompany each). 

This is the main reason you don't see it used, IMO. I have done it with a dual package, but it can be squirrly. Some opamps won't like it.
Having said that, a resonant circuit is surely one of the best applications for this technique. When you don't want to attenuate the rest of the audio band before/within the resonant circuit you can bootstrap its rails, then attenuate after the signal has emerged from the opamp (so not everything in the pedal needs to be bootstrapped, just the resonant opamp).

Aye aye, sir! Bootstrapped single op amp stage resonant filter section it is. Have to find an appropriate circuit but thats going to be another thread. Can you recommend or warn of specific op amps? I have a bunch of TL072's and NE5532AP's lying around. I guess that using only one half and leave the other unused and stabilized would work for this?

EDIT: By the way, the bootstrapping technique does not care if the op amp it straps is used inverting or non-inverting, right?

Thanks,
Andy

[merlinb]

Can you recommend or warn of specific op amps? I have a bunch of TL072's and NE5532AP's lying around. I guess that using only one half and leave the other unused and stabilized would work for this?

Yes I think it will be OK with most any opamp if the other half is unused, otherwise it depends on the overall topology, i.e what signal the other half is handling at the time.

[Transmogrifox]

EDIT: By the way, the bootstrapping technique does not care if the op amp it straps is used inverting or non-inverting, right?

Doesn't matter unless you were using a dual-package with one inverting stage followed by a noninverting stage (or vice-versa).

You don't want one of your outputs going negative as the negative rail is coming up to meet it.

A single inverting stage is ok because the output pushes the rail in the direction it wants to go.