Op amp based preamp supply. Charge pump, single vs symmetrical & ripple

[dazz]

I'm cloning the preamp of a Marshall Lead 12 3005 into a pedal. It's a two stage op amp job with a +16V -16V supply.
The plan is to use a standard 9V power supply to power it, so I'm considering two options:

1. A charge pump based Dual-Output Voltage Doubler (pic 1) based on an LT1054 to get a symmetrical supply of about +16V -16V
2. A charge pump voltage multiplier with two ICL7660 or LT1054 in series for a single supply of about 24V (according to my LTSpice simulation)

I'm leaning towards option 2, but I'm not sure if the reasons for that are justified. I believe that having the opamp's V- tied to ground will limit/halve the amount of ripple I'll get as opposed to having a positive and negative rails, besides, the PSRR or the V+ is much higher in the TL082 I plan to use, so that should further mitigate the effect of ripple in the power rails.

I know people are using 7660S chips with the boost pin to Vcc to increase the switching frequency from 10KHz to 30KHz, outside the hearing range with good results in symmetric supply configurations, but what if I was to use the cheaper 7660 at 10KHz in single supply mode? would the fact that there's just the positive rail ripple help make up for the lower frequency?

I tried crunching some numbers, but I have no idea what I'm doing. Here they go anyway: The PSRR of the TL802 in V+ is 80dB at 10KHz. The preamp's gain at 10KHz, according to LTSpice, is about 70dB. So I'm guessing the V+ ripple will appear at the preamp's output attenuated 10dB (assuming single supply charge pump with a 7660 at 10KHz). That's an attenuation of a factor of 3, give or take. Can I expect to have low enough ripple  in the charge pump output that it won't whine?

I've also run some simulations of the charge pump (pic 2) and with a 10mA load which I think is reasonable for a dual op amp circuit like this driving a 50K load (standard chipamp input impedance) and I get only 3-4mV ripple. That would translate to 1mV in the output. That doesn't sound like too much, does it? I mean if the voltage swing at the output is 24V, 1mV won't disturb it too much, would it?

Anything I may be missing please?

[Benta]

Quite frankly, the pain of running opamps on a single supply is not worth the effort if it can be avoided.
A dual supply opamp has a wonderful signal reference: ground. In a single supply design, you need to recreate this at some DC bias point. Not easy, and usually not noise free.

You have the option of running dual supplies, and I highly recommend this.
Your reflections on PSRR are correct, but can be alleviated with appropriate filtering.

[dazz]

Thanks Benta. Do you think the fact that symmetrical charge pumps usually have weaker V- rails (in my simulations I got something like +16V, -15V) could have unwanted effects? Wouldn't it cause the preamp to clip asymmetrically because of the imbalance of the rails, hence affecting its tone?

I definitely plan on testing all these things myself once the parts arrive, should be a good learning exercise

[Benta]

I expect your signals are at line level (~775 mV), but even if they are a bit higher or lower, supply voltage will have zero influence. Asymmetric supply voltage will also have zero impact.
You're overthinking this in my opinion.

[dazz]

I expect your signals are at line level (~775 mV), but even if they are a bit higher or lower, supply voltage will have zero influence. Asymmetric supply voltage will also have zero impact.
You're overthinking this in my opinion.

I should have mentioned it's a guitar preamp/effect. So the input is much lower than line level, perhaps 100mV peak to peak. Does that change anything?
And yes, I tend to overthink stuff, so it may very well be the case here, haha

[Benta]

Your signal level changes nothing, quite the contrary. At low signal levels, a good ground reference is even more important.

[Zero999]

Quite frankly, the pain of running opamps on a single supply is not worth the effort if it can be avoided.
A dual supply opamp has a wonderful signal reference: ground. In a single supply design, you need to recreate this at some DC bias point. Not easy, and usually not noise free.

You have the option of running dual supplies, and I highly recommend this.
Your reflections on PSRR are correct, but can be alleviated with appropriate filtering.

But converting a single supply to a bipolar supply is also a pain, because it introduces more noise.

I expect your signals are at line level (~775 mV), but even if they are a bit higher or lower, supply voltage will have zero influence. Asymmetric supply voltage will also have zero impact.
You're overthinking this in my opinion.

I should have mentioned it's a guitar preamp/effect. So the input is much lower than line level, perhaps 100mV peak to peak. Does that change anything?
And yes, I tend to overthink stuff, so it may very well be the case here, haha

If it's low levels, then the larger possible voltage swing a +/-16V supply could give is not needed. A lower voltage op-amp could be used with a single supply.

[T3sl4co1l]

If you must, consider heavy filtering (or investing in opamps with high PSRR at higher frequencies).  The likes of ICL7660 are quite noisy indeed, and at moderate frequencies (~10kHz), which is all the more annoying.  An LDO may be worthwhile to take the crud out (but again, mind the PSRR, as that's also the frequency range where most LDOs start to really suck).

But there should be no need for any more than a resistor divider (or a boosted one, i.e. using a spare opamp section for a rail splitter) in an application like that.

Tim

[Zero999]

But there should be no need for any more than a resistor divider (or a boosted one, i.e. using a spare opamp section for a rail splitter) in an application like that.

Tim

Agreed. Let's also not forget the TLE2426.
http://www.ti.com/lit/ds/symlink/tle2426.pdf

If the power supply is only powering that circuit and isn't connected to anything else, then no AC coupling capacitors are needed: the 0V from the rail splitter can be treated as the common rail for the whole circuit.

[spec]

Crossed Posts

Just so we can see whats involved here is the schematic for the Marshall Lead12-12W: https://www.thetubestore.com/lib/thetubestore/schematics/Marshall/Marshall-Lead12-12W-3005-5005-Schematic.pdf

My first thoughts are to use better RRIO opamps, so that the supply rails could be lowered and get the same or better dynamic range and open loop gain. That would possibly enable a simpler circuit with a better performance, including a lower noise, not that noise is an issue here.

A 12V, 15V, 18V, or 24V wall wart should do nicely for the PSU.

[b_force]

I wouldn't go for these charge pumps anymore these days. They can be noisy, regulation is poor or limited and the output current is small.
There are quite some good converters that can do split rail (TI has a few)
Also you can use pretty much every buck converter to make a negative voltage rail.
If efficiency is not the main priority there are also of the shelf converters available.

Look for something that's at least 100kHz, but the higher the better plus some post filtering.

[bd139]

Another option - use an AC wall wart and just derive two rails from it. This works pretty well, gives you no compromise rails and keeps your circuits dead simple. Typical example of deriving three voltages cheaply (I would use regulators myself and be done with the zeners):

[dazz]

Thanks everyone for your input. Lots of suggestions to explore.

If it's low levels, then the larger possible voltage swing a +/-16V supply could give is not needed. A lower voltage op-amp could be used with a single supply.

Sorry mate, I don't understand this. The input is low level, but the output at maximum gain is a square wave 30V peak to peak. I think I need to have that +/-16V supply to replicate the way the circuit distorts. Let me know if this is wrong, but the rationale is that the gain pot (attached schematic) changes the frequency response of the amp, at low gain it boosts higher frequencies more (around 10KHz) but as the gain is increased, the higher gain frequency shifts to about 1KHz. So if the original circuit with the +/-16V supply starts clipping at 10KHz with the gain knob at 50%, with a 9V supply it will start clipping very early at maybe 10% gain, and at all frequencies, making it almost unusable. I don't know if the circuit can be modified to behave the same way as the original one but with a single 9V supply. I can attenuate the gain slightly by increasing resistor R8 in between the op amps, but not enough to compensate for the lack of voltage swing.

If you must, consider heavy filtering (or investing in opamps with high PSRR at higher frequencies).

Well, I'm not sure I can just swap op amps like that. People will complain that all the mojo in the original LM1458 or the TL072/082 is lost. I personally have no idea if the op amp of choice plays any roll in how the thing clips and sounds. The simulations seem to suggest that the clip differently, but who knows.

But there should be no need for any more than a resistor divider (or a boosted one, i.e. using a spare opamp section for a rail splitter) in an application like that.

Can you please elaborate on this? What rail would the resistor divider split and what would the opamp do there please?

[dazz]

But there should be no need for any more than a resistor divider (or a boosted one, i.e. using a spare opamp section for a rail splitter) in an application like that.

Tim

Agreed. Let's also not forget the TLE2426.
http://www.ti.com/lit/ds/symlink/tle2426.pdf

If the power supply is only powering that circuit and isn't connected to anything else, then no AC coupling capacitors are needed: the 0V from the rail splitter can be treated as the common rail for the whole circuit.

Thanks Hero, googling that chip now. And yes, the only thing that the power supply needs to power is that dual opamp circuit.

Crossed Posts

Just so we can see whats involved here is the schematic for the Marshall Lead12-12W: https://www.thetubestore.com/lib/thetubestore/schematics/Marshall/Marshall-Lead12-12W-3005-5005-Schematic.pdf

My first thoughts are to use better RRIO opamps, so that the supply rails could be lowered and get the same or better dynamic range and open loop gain. That would possibly enable a simpler circuit with a better performance, including a lower noise, not that noise is an issue here.

A 12V, 15V, 18V, or 24V wall wart should do nicely for the PSU.

Thanks for posting the schematic, should have done it myself in the 1st post.
I'm still not sure how I could use lower power rails without significantly impacting the way the amp will perform. The thing has 80dB of voltage gain (with the gain pot at max) and the gain pot is part of a low pass filter so it changes the frequency response of the amp. So if the original one clips at 50% and mine clips at 10%, they won't sound the same. At least I think so.

Also, a 9V supply is pretty much a requirement, since this will be a guitar pedal and most pedal power supplies are 9V 

[b_force]

I know 9V is pretty standard, but it's a little unfortunate.
Otherwise it would be pretty easy to use something like a Meanwell MDD01N-15 or similar.
The minimum voltage for those is either 10.8V or maximum voltage for 5.5V (for the 5V version).

The second best choice would be a Meanwell DPAN02A-15 or DPAN02E-15
they also switch at 100kHz, so that can easily be filtered out.
How much opamps/current do you need?

This is the +/- 15V version, which also gives you plenty of headroom.
There are also 12V and 5V version available.

[dazz]

I know 9V is pretty standard, but it's a little unfortunate.
Otherwise it would be pretty easy to use something like a Meanwell MDD01N-15 or similar.
The minimum voltage for those is either 10.8V or maximum voltage for 5.5V (for the 5V version).

The second best choice would be a Meanwell DPAN02A-15 or DPAN02E-15
they also switch at 100kHz, so that can easily be filtered out.
How much opamps/current do you need?

This is the +/- 15V version, which also gives you plenty of headroom.
There are also 12V and 5V version available.

It's just a dual opamp (TL082 or LM1458). Maybe 10mA? 20mA tops I think
Off to google those chips too, thanks for the suggestion

[b_force]

I know 9V is pretty standard, but it's a little unfortunate.
Otherwise it would be pretty easy to use something like a Meanwell MDD01N-15 or similar.
The minimum voltage for those is either 10.8V or maximum voltage for 5.5V (for the 5V version).

The second best choice would be a Meanwell DPAN02A-15 or DPAN02E-15
they also switch at 100kHz, so that can easily be filtered out.
How much opamps/current do you need?

This is the +/- 15V version, which also gives you plenty of headroom.
There are also 12V and 5V version available.

It's just a dual opamp (TL082 or LM1458). Maybe 10mA? 20mA tops I think
Off to google those chips too, thanks for the suggestion

It might be tricky to get these things stable (since some converters have a minimum load needed).

Obviously there are ways to build a circuit from scratch.
I don't know how comfortable you are with soldering SMD stuff?

[dazz]

It might be tricky to get these things stable (since some converters have a minimum load needed).

They're also pretty expensive at $10 or so. And would make the pcb a lot larger too. I'm trying to fit this thing in a 1590BB enclosure

Obviously there are ways to build a circuit from scratch.
I don't know how comfortable you are with soldering SMD stuff?

I've only soldered SOT23 jfets, LOL. But I don't mind trying something else if needed be.
The thing is that this will be a DIY thing that I will publish, and most people prefer through hole since it's easier to assemble.

But I'm very interested in how the circuit could be rebuilt to perform the same with a lower power supply. So if you could give me some pointers that would be very much appreciated

[b_force]

It might be tricky to get these things stable (since some converters have a minimum load needed).

They're also pretty expensive at $10 or so. And would make the pcb a lot larger too. I'm trying to fit this thing in a 1590BB enclosure

Obviously there are ways to build a circuit from scratch.
I don't know how comfortable you are with soldering SMD stuff?

I've only solder SOT23 jfets, LOL. But I don't mind trying something else if needed be.
The thing is that this will be a DIY thing that I will publish, and most people prefer through hole since it's easier to assemble.

But I'm very interested in how the circuit could be rebuilt to perform the same with a lower power supply

10 bucks? Well, that's a pretty bad deal, I normally buy them for around $4-6
Which isn't that much more expensive compared to developing your own stuff.

Another idea i to get a basic 9-12 or 9-15 boost converter and use a flyback or 1:1 coupled inductor to form the negative voltage.

[dazz]

Another idea i to get a basic 9-12 or 9-15 boost converter and use a flyback or 1:1 coupled inductor to form the negative voltage.

Yeah, I already have one of those boards I got from ebay. Maybe I could add a boost converter to my board. Something based on a LM2577-adj perhaps? Ideally something that can be got for cheap off ebay or the likes

[b_force]

Another idea i to get a basic 9-12 or 9-15 boost converter and use a flyback or 1:1 coupled inductor to form the negative voltage.

Yeah, I already have one of those boards I got from ebay. Maybe I could add a boost converter to my board. Something based on a LM2577-adj perhaps? Ideally something that can be got for cheap off ebay or the likes

Well, the problem is to get a good efficiency with these low currents.
Most converters are optimized for something around 100-500mA or thereabouts.
The efficiency at low currents can be as bad as 50-60%. Not really ideal for a 9V battery.
(I am personally getting rid of all my 9V pedals or rebuild them to accept two AA etc)

What I would suggest is otherwise use something like a LT1615 (or a similar TI parts) look at page 8, or LT3460, LT1316 etc (probably better because it has a constant switch Freq)
With LTSpice you can simulate this circuit and optimize it for 9V

[Zero999]

Thanks everyone for your input. Lots of suggestions to explore.

If it's low levels, then the larger possible voltage swing a +/-16V supply could give is not needed. A lower voltage op-amp could be used with a single supply.

Sorry mate, I don't understand this. The input is low level, but the output at maximum gain is a square wave 30V peak to peak. I think I need to have that +/-16V supply to replicate the way the circuit distorts. Let me know if this is wrong, but the rationale is that the gain pot (attached schematic) changes the frequency response of the amp, at low gain it boosts higher frequencies more (around 10KHz) but as the gain is increased, the higher gain frequency shifts to about 1KHz. So if the original circuit with the +/-16V supply starts clipping at 10KHz with the gain knob at 50%, with a 9V supply it will start clipping very early at maybe 10% gain, and at all frequencies, making it almost unusable. I don't know if the circuit can be modified to behave the same way as the original one but with a single 9V supply. I can attenuate the gain slightly by increasing resistor R8 in between the op amps, but not enough to compensate for the lack of voltage swing.

Then I misunderstood. If you need 30V peak to peak, then you need a higher total supply voltage than 30V. A +/-16V supply won't do for the TL082, which needs a couple of volts of headroom on either supply rail, to avoid clipping, unless that's desired for the distortion effect.

The OPA2196 will be able to give 30V peak to peak from a +/-16V supply. If does have a lower slew rate, than the TL082, but it will be the lower audio frequency components which generate the largest voltage swings, so this is a non-issue.
http://www.ti.com/lit/ds/sbos869/sbos869.pdf

[dazz]

Then I misunderstood. If you need 30V peak to peak, then you need a higher total supply voltage than 30V. A +/-16V supply won't do for the TL082, which needs a couple of volts of headroom on either supply rail, to avoid clipping, unless that's desired for the distortion effect

Yes, it's built to distort, for rock'n'roll purposes! 

More precisely, it's supposed to do both clean and distorted tones. That's what the gain pot is for

The OPA2196 will be able to give 30V peak to peak from a +/-16V supply. If does have a lower slew rate, than the TL082, but it will be the lower audio frequency components which generate the largest voltage swings, so this is a non-issue.
http://www.ti.com/lit/ds/sbos869/sbos869.pdf

I'm afraid it really needs to be an LM1458 as in the original circuit, or one of those TL082 with jfet inputs (which are supposed to clip more softly). Well, not really those 2, just one op amp with lots of mojo regardless of whether other types make more sense from an engeneering perspective. That's how guitar players roll 

[Zero999]

Then I misunderstood. If you need 30V peak to peak, then you need a higher total supply voltage than 30V. A +/-16V supply won't do for the TL082, which needs a couple of volts of headroom on either supply rail, to avoid clipping, unless that's desired for the distortion effect

Yes, it's built to distort, for rock'n'roll purposes! 

More precisely, it's supposed to do both clean and distorted tones. That's what the gain pot is for

The OPA2196 will be able to give 30V peak to peak from a +/-16V supply. If does have a lower slew rate, than the TL082, but it will be the lower audio frequency components which generate the largest voltage swings, so this is a non-issue.
http://www.ti.com/lit/ds/sbos869/sbos869.pdf

I'm afraid it really needs to be an LM1458 as in the original circuit, or one of those TL082 with jfet inputs (which are supposed to clip more softly). Well, not really those 2, just one op amp with lots of mojo regardless of whether other types make more sense from an engeneering perspective. That's how guitar players roll 

I admit I'm more used to avoiding distortion when building an amplifier.

Do you have a specific type of distortion i.e. transfer function in mind?

I think most op-amps will give fairly harsh clipping, as once the output goes beyond the limit, there will be no negative feedback. I suspect the clipping level may vary between op-amps with the same part number, because it's not a tightly controlled characteristic. By operating an op-amp at its clipping level, it's being driven past its deign specification.

[schmitt trigger]

Another thing with using charge pumps, is that your negative output won't be completely symmetrical.

As the application circuit shows, there will be the diode's forward losses, plus any losses in the converter itself.

And indeed if not filtered correctly, the plain 7660 will generate an audible whine when used with sensitive audio circuits.