How do I create a split rail?

[Sage]

A while back I built an effects loop for my amp using a TL072 to create a buffered and attenuated send, and a boosted return:



I powered it by hooking it up to a +18V line that was already present in the amp and being used to power the onboard reverb circuit.

Now I would like to update the buffer so that it's powered by +18V/-18V instead of just +18V, in order to get the maximum headroom out of the TL072.  After reading up on R.G.'s Spyder power supply for pedals (which I'm in the middle of building), I have a better grasp of how to use a transformer, bridge rectifier, and linear voltage regulator to get a specific voltage output.  This has helped me become more familiar with the inner workings of my amp as well.  But I still don't know how to get that -18V rail.

The transformer in the amp has three AC outputs: 225V, 12V, and 7.5V.  The 12V line is fed through an array of diodes which I believe act as a rectifier whose output reads 25V DC on my multimeter.  From there it goes through a 7818 linear voltage regulator which outputs +18V.  I want to piggyback my circuit off the power lines somewhere in this chain.

Does anyone have an example of how I can most easily get +18V/-18V out of this?  Do I need to use a charge pump to get to +36V and then convert that to +18V/-18V?  Or can I generate a -18V rail to complement the existing +18V one?  Are there any examples out there I can adapt to either solution?

[Unlikekurt]

1. so long as the transformer secondary winding that you are utilizing is spec'd for enough current for all the needs of the amplifier AND the circuit that you are adding.  This goes for both your usage of the current +18v rail and the forthcoming -18v rail.

2.  Depending HOW the AC is rectified in the circuit currently to provide the +18V is important.  That said, if a bridge rectifier is used and the negative side of the DC output is not referenced directly to ground, you could use a 7918 regulator on that side, invert the filter caps on that string and walla.

3.  After typing the above, it is clear that you should post the schematic for the device you are attempting to obtain the supply voltage from.  Looking at that would help someone be able to help you.

[Sage]

1. so long as the transformer secondary winding that you are utilizing is spec'd for enough current for all the needs of the amplifier AND the circuit that you are adding.  This goes for both your usage of the current +18v rail and the forthcoming -18v rail.

I've been able to power the +18V version of the circuit without any problems.  I *think* the secondary winding that's powering this part of the amp is rated at 500mA, but I could be mistaken, as the source of that info also claims it's a 10V output and I'm clearly reading just over 12V.  As far as I can tell it's only powering a couple of ICs, primarily the onboard digital reverb, and as you can see above, the circuit I'm adding is a pretty simple TL072 buffer, so it shouldn't be too draining.

2.  Depending HOW the AC is rectified in the circuit currently to provide the +18V is important.  That said, if a bridge rectifier is used and the negative side of the DC output is not referenced directly to ground, you could use a 7918 regulator on that side, invert the filter caps on that string and walla.

You're saying connect the input leg of the 7918 to the DC ground of the 7818, and connect the ground leg of the 7918 to the input side of the 7818 to reverse the voltage?

3.  After typing the above, it is clear that you should post the schematic for the device you are attempting to obtain the supply voltage from.  Looking at that would help someone be able to help you.

Here's the relevant section of the schematic:


CN15 and CN19 are the output wires from the transformer.  What confuses me is that instead of four diodes arrayed in a bridge rectifier, I see three diodes (D8, D9, D10) and a cap (C29) so I really don't know what's going on there.

[Sage]

Edit: accidentally duplicated my post, sorry

[Sage]

Is it possible that D11 is operating as part of the rectifier along with D8, D9 and D10 in the amp schematic I just posted?  I have a basic understanding of how a 4-diode bridge rectifier works but I'm not familiar with whatever is going on here.  Is this really as simple as reversing the ground and positive inputs of the 7818 and hooking them to a 7918, or do I need to start with the raw AC coming out of CN15/CN19 and build a second rectifier?  Is that even doable?  This is all new to me.

[slacker]

D8, D9, D10 and C29 make a voltage multiplier to generate the 25volt DC from the 12volt AC or whatever the AC voltage is. D11 and the components after it create a signal to control the 555 to reset whatever is attached to "reset".
I don't think you can get a negative voltage out of that set up, where you have a full wave rectifier( the four diodes) you can only get positive voltages, you can't also tap off a negative one.
The easiest way to get more headroom would be to run your circuit off the 25 volt supply, no need to regulate it but you might need to add extra power supply filtering.

[Sage]

Could it be done with a charge pump?  I've seen a +9V DC supply be converted to +/-9V that way.

[davent]

Charge pumps have pretty low maximum input voltages, don't know of any as high as 18, any out there?

[PRR]

That is a messy power supply. I would not try to get additional outputs from it.

Why do you think you need "more headroom"?? You probably only need 1V to smack your amp's return jack, and with 18V power you got 5V of possible signal..

[PRR]

And if you "do" need more output.... keep your existing build, but power it from the + end of C35. This point is surely higher than 21V, probably 24V, but can't be over 35V (or C35 would have already exploded). There is some issue taking load from a voltage tripler, but your TLO72 chip doesn't eat much.

[Sage]

Why do you think you need "more headroom"?? You probably only need 1V to smack your amp's return jack, and with 18V power you got 5V of possible signal..

If you look at the schematic for the effects loop itself in my original post, there's a -14db attenuator right on the input before the signal even hits the TL072.  You can see that the return signal is boosted by 14db via the feedback loop of the other side of the TL072.

The reason I did this wasn't just to accommodate outboard gear.  It's because, in order to insert this send/return after the gain stage of the amp for both the clean and overdrive channels, I also had to put it after the EQ section and volume controls.  The amp schematic's too big to post inline, but here's a link to it, and you can see for yourself where I inserted the fx loop, as exhibited by the red line up near the top middle of the schematic.

(All this just to insert a delay pedal after the distortion.  )

What I found (by testing the loop with the send connected directly to the return) was that without the attenuator (and the accompanying boost), the TL072 gets overloaded as I turn up the volume.  That's why the attenuator is placed before the TL072 buffer and not after.

The effects loop is on a DPDT switch configured for true bypass, so I can completely remove it from the signal path.  The send/return jacks are switched so that they pass a signal even if nothing's plugged into them.  At higher volumes, I've noticed a very slight change in tone when the effects loop is switched on that I can only describe as a slight loss in "girth," as though I was losing some of the lower end.  The difference is miniscule, but I wanted to see if I could improve on my initial design.

My assumption -- and *please* correct me if I'm wrong -- is that with more headroom for the TL072, I could significantly reduce the amount of attenuation, if not remove it entirely, hopefully retaining more of the integrity of the amp's tone.  I'm not completely certain how this would affect the gear that's actually in the effects loop, but I'm using an Empress Tape Delay, which has an adjustable input gain down to -12db, and a T.C. Electronic Flashback X4 in looper mode, which supposedly has "plenty of headroom" according to the vague marketing copy.

And if you "do" need more output.... keep your existing build, but power it from the + end of C35. This point is surely higher than 21V, probably 24V, but can't be over 35V (or C35 would have already exploded). There is some issue taking load from a voltage tripler, but your TLO72 chip doesn't eat much.

I think that's actually a voltage doubler... the transformer's output is supposedly rated for 10V but I'm reading about 12.3V, and you're right, the + end of C35 is at about 24-25V.  D9-10 and C29/35 look to be a textbook voltage doubler.  I still don't know what D8 is doing there, but wouldn't there need to be another capacitor involved to make it a tripler?

A thought: Couldn't I just build an inverted voltage multiplier from the AC source to drive a 7918 and get that -18V rail?

[Sage]

Also, correct me if I'm mistaken, but the TL072 can't take a +24V power supply, can it?

Or am I being stupid?  Is the +18V upper limit relative to Vref?  I could've sworn someone told me otherwise, but looking at the data sheet I'm starting to think I don't need a negative rail at all.

[slacker]

The TL072 can take up +-18 Volts Ie: +18 Volts on its positive supply pin (pin 8 ) and -18 Volts on its negative supply pin (pin 4) or 36 Volts in total across the two pins. It's shown as +-18 Volts because that's a typical way to use them but it doesn't care what the actual voltages are so long as there isn't more than 36 in total so you can use +36 and 0 or practically any voltages so long as the difference isn't more than 36 Volts. So it will work fine on +24 Volts.

Like Paul said that should be more than enough, 24 Volts will give the loop more headroom than 9 or 18 Volt pedals, so there shouldn't be a problem.

[Sage]

Okay, thanks for clearing that up... for some reason I had thought that the +/- 18V limits were relative to ground, not to Vref.  Being able to use a single rail supply at a higher voltage will make this so much easier!

[Keppy]

If you look at the schematic for the effects loop itself in my original post, there's a -14db attenuator right on the input before the signal even hits the TL072.  You can see that the return signal is boosted by 14db via the feedback loop of the other side of the TL072.

It's a small point, but your schematic actually attenuates by a factor of 5 before boosting by a factor of 6.

What I found (by testing the loop with the send connected directly to the return) was that without the attenuator (and the accompanying boost), the TL072 gets overloaded as I turn up the volume.  That's why the attenuator is placed before the TL072 buffer and not after.

If the TL072 is capable of boosting the signal back to where it started without clipping, then you already have enough output headroom with the power supply you have. Instead of redesigning your supply, why not try an opamp with rail-to-rail inputs? If supply headroom is really your problem, then the output stage would clip even with the attenuation/boost.

[PRR]

> Couldn't I just build an inverted voltage multiplier from the AC source

No. The +12V supply's bridge rectifier is grounded at one corner, so both sides of the AC winding "float" off-ground. The upper doubler/tripler is hacked to work with this condition (may be why it has an odd number of parts). I'm not saying you *can't* also get a negative voltage, but most such schemes will have a hidden through-diodes short across the winding. This could be a very expensive experiment.

The power amp V3B V4A needs only 2V CN13B to get its full output.

The tonestack driver V2B V3A and the tonestack can only deliver about 6V to the RL2B relay. (Less if you use OD VOL.) Yes, you had to cut this down to fit through a 12V opamp and a typical effect, and boost-back afterward. I do think a 24V supply would equal or exceed the signals possible with the stock signal path.

One thing I will suggest. Break-out at CN24B/CN13B instead of after RL2B. The signal levels are somewhat less. The drawback is that the signal here is already reverb-erated. However if you are inserting "a delay", maybe you don't also want reverb on it?

Note that the designer took RL2B signal directly to the IC3A opamp with 18V power (albeit at unity-gain). This is a hint that the designer did not expect signal levels above what would fit in 18V. (However slight clipping on a reverb drive is not a great sin.)

> why not try an opamp with rail-to-rail inputs?

Hardly worth the effort. With 18V supply the TL072 can do 14V p-p. A "R2R" opamp on 18V can do 17V maybe 17.5V p-p. 25% more. "Usually", if you are "short" on audio, another 25% does not ensure total happiness. Maybe; it's marginal.

You only need 5mA of TL072 power, so you "could" drop down from the 350V in large resistor or MOSFET. Power dissipation near 1.5 Watts, maybe 2 Watts for a simple Zener regulator (the Zener needs to be fed). So it is not neat nor small, and has dangers.

[Sage]

It's a small point, but your schematic actually attenuates by a factor of 5 before boosting by a factor of 6.

Are you sure?  I did all the math back when I first put it together and I'm pretty sure it's both attenuating and boosting by a factor of 5... when I a/b it via the true-bypass switch, I hear no difference in volume.

If the TL072 is capable of boosting the signal back to where it started without clipping, then you already have enough output headroom with the power supply you have. Instead of redesigning your supply, why not try an opamp with rail-to-rail inputs? If supply headroom is really your problem, then the output stage would clip even with the attenuation/boost.

Thanks for bringing this up; I was actually wondering if that was the case.  I apologize if this sounds rude, but can anyone else confirm this?

[Sage]

> Couldn't I just build an inverted voltage multiplier from the AC source

No. The +12V supply's bridge rectifier is grounded at one corner, so both sides of the AC winding "float" off-ground. The upper doubler/tripler is hacked to work with this condition (may be why it has an odd number of parts). I'm not saying you *can't* also get a negative voltage, but most such schemes will have a hidden through-diodes short across the winding. This could be a very expensive experiment.

Whoa, nice catch, thanks for the warning.  I see now why you said it was messy.  Unfortunately, this sounds like it rules out any attempt at hanging another voltage multiplier off that AC input, which is where I was headed once I found out I could just use more voltage with my existing buffer design. 

The power amp V3B V4A needs only 2V CN13B to get its full output.

The tonestack driver V2B V3A and the tonestack can only deliver about 6V to the RL2B relay. (Less if you use OD VOL.) Yes, you had to cut this down to fit through a 12V opamp and a typical effect, and boost-back afterward. I do think a 24V supply would equal or exceed the signals possible with the stock signal path.

Thanks.  So I could just take the DC voltage coming right out of the multiplier and use that, essentially bypassing the 18V regulator... what would be the disadvantages of doing this?  Inconsistency of some sort?

One thing I will suggest. Break-out at CN24B/CN13B instead of after RL2B. The signal levels are somewhat less. The drawback is that the signal here is already reverb-erated. However if you are inserting "a delay", maybe you don't also want reverb on it?

I do want reverb after the delay, which is why I broke the circuit before the reverb instead of after.  I suppose the exact order of those effects may be largely moot, though.

Note that the designer took RL2B signal directly to the IC3A opamp with 18V power (albeit at unity-gain). This is a hint that the designer did not expect signal levels above what would fit in 18V. (However slight clipping on a reverb drive is not a great sin.)

Interesting.  The reverb is essentially sitting in its own parallel effects loop, and if the opamp is only handling the wet signal, then maybe it doesn't need to get quite as loud as the dry signal?

[Keppy]

It's a small point, but your schematic actually attenuates by a factor of 5 before boosting by a factor of 6.

Are you sure?  I did all the math back when I first put it together and I'm pretty sure it's both attenuating and boosting by a factor of 5... when I a/b it via the true-bypass switch, I hear no difference in volume.

A non-inverting opamp gain stage like the one you're using has gain equal to the ratio of the resistors plus one. As I said though, it's a small difference. Total gain of the stages combined is 1.2 for about a 1.5db boost. If you wanted the the overall gain to be exactly unity, you could tweak the values, or just change both stages to inverting configuration. As you don't even hear a difference in volume, though, it's probably not worth the trouble.

[PRR]

> reverb .... if the opamp is only handling the wet signal

Reverb is 3 parts. Driver, tank, and (because tank is lossy) a recovery make-up amp.

The driver only carries "dry" signal (wet happens in the tank).

However-- if a signal is clipped then run through a spring reverb tank, the clipping artifacts come out of the tank all different times and phases. It isn't straight clipping. In fact many classic tube reverb drivers run awful close to or into clipping at high levels. This may be the "splashy" sound of hard-worked reverb. Which some players like; it adds emphasis they can't get by reed or vocal-cord overblowing strain. So I'm not convinced the level in the reverb driver is "sure to be clean". But by other estimates I can't see how it could be very overloaded.

> take the DC voltage coming right out of the multiplier... disadvantages of doing this?

That point may have 120Hz ripple. The TL072 is pretty good at rejecting it. Your Vref splitter needs filtering but it appears ample. Worst-case, there is a buzz. Add 150-220 ohms and 470uFd R-C filter between the 18V and the opamp power pin, that will knock the ripple down without much drop of voltage.