Op amps: rail to rail vs higher supply voltage?

[blackieNYC]

I think I'm missing something here. In an op amp-only circuit, why should one bother with higher supply voltages instead of simply using a rail-to-rail op amp?  If one feels the need for more headroom?  The r to r will cost more, but so will the parts to give yourself a higher supply voltage, like a max1044 or a 15v wall wart.  Thanks.

[karbomusic]

I would think R2R is to maximize the headroom you should be getting from whichever voltage you are using. I use an R2R  (TLC2272)  for that reason but still spec one of my ODs to run anywhere from 9-18V and the R2R is beneficial across that entire voltage range.

[anotherjim]

Rail to rail will only get you a bit more headroom. Also, if you look at the specs, you often find they only manage  full rail with high output load impedance >2k. Most audio circuits will be higher impedance, but not all. As output current increases, maximum swing moves away from the rails. Not only that, but it's by different amounts for source or sink so the clipping becomes asymmetrical.

[blackieNYC]

Thanks for the response.  Not an answer I expected.  You are talking about having a big waveform to clip after the amplification.  For arguments sake, why would you need a rail to rail op amp AND an 18v supply?  That's a huge swing.  I would think that downstream you have clipping diodes - for the peak levels coming out of your guitar, is there much difference between clipping a 16v p-p signal down to 1.5vp-p (silicon diodes) and clipping an 8vp-p signal down to the same?  or maybe you just have a high level from the previous pedal that you don't want to clip the OD.
I guess what I'm primarily talking about is say a delay or phaser that one might put after a pedal with some gain - an overdrive set up to boost a bit and push your amp as well, but with a delay you're currently designing in between the two. You don't want the delay op amp to clip, so maybe you decide your TL072 at 9v isn't enough headroom.  I would think an R to R op amp OR a 15-18v supply would be enough.  At that point, if you had to choose between the two, what are the pros and cons? The headroom is fairly similar either way, more or less.

[blackieNYC]

Another Jim - but if you keep output impedance above 2k, you won't face this reduction?  And wouldn't the lower output impedance also affect an op amp running on 18v?

[anotherjim]

As far as I've seen, ordinary op-amps have fairly "stiff" outputs which are also more symmetrical. R2R amps have different internals.

Of course, if you're going to increase peak to peak output volts, you need to look at ohms law with regard to output current increase, whatever the amp. So, some diode to ground hard clippers may have 1k series resistance to limit the current. You may have to look at increasing that.

The following is copied text from the AD822 R2R amp data. Format is lost and I've cleaned it up just to show worse case figures, but you can see how the drift from the rails increases with current and how it differs for source or sink...

All at +/-15v supply

VOL − VEE ISINK = 20 μA  7 mV

VCC − VOH ISOURCE = 20 μA  14 mV

VOL − VEE ISINK = 2 mA  55 mV

VCC − VOH ISOURCE = 2 mA  110 mV

VOL − VEE ISINK = 15 mA 500 mV

VCC − VOH ISOURCE = 15 mA 1500 mV

Even as low as 20uA, source and sink differ. At 15mA sink, it's 0.5V away from the rail and source is 1.5V off!
Up to 2mA, I doubt the asymmetry would be audible.

Here's the LM833, which is not classed as R2R. You don't get much data, but to be fair to the AD822, I've also only included worst case...

All at +/-15v supply

RL = 10 kΩ ±12 V (I make it 1.35mA source and sink)

RL = 2 kΩ ±12 V (I make this 6ma source & sink)

...that's all it has to say. It may be 3v off  top & bottom (typical value happens to be 1.6v), but it is symmetrical.

If I've got the point of the original question right, changing from 9v to 18v with the original "ordinary" amp, is always going to beat keeping 9v and changing to an R2R amp. Cost implications and what it sounds like are whole other things

All that said, double supply voltage to 18v isn't going to get you better than 6dB extra headroom, although that could be enough to make the difference you want. I can't be bothered with the maths, but the extra you might get using R2R @9v is what? 1 or 2 dB?

[ggedamed]

Could be some cases when you only need those 3Vpp of additional headroom gained by simply changing an op amp. Which is somewhat equivalent to increasing the supply voltage from 9V to 12V.
Also, if the signal is clipped, you should not care too much if it is clipped symmetrically or not.

Note: most of the the rail-to-rail op amps have lower maximum supply voltage (like 16V for TLC2262/2272).

[amptramp]

You should check whether the rail-to-rail spec applies to the input or the output or both.  Some amplifiers can handle rail-to-rail input but have limited output swing.  Others can swing to the rails but have limited input common mode range.  Still others can accept inputs going to one rail but not the other (e.g. the LM324 can go to the negative rail but not the positive on the input).  Check to make sure what you need before you specify the op amp.

[PRR]

> simply using a rail-to-rail op amp?

Dumb sloppy old TL0 chips lose about 1.5V each side. "R2R" chips with practical loads may lose only 0.3V each side.

So do math:

r2r with 9V supply == 8.4Vpp
TL0 with 9V supply == 6.0Vpp
TLO with 12V supply == 9Vpp
TLO with 30V supply = 26Vpp

TL0-to-R2R may give 3dB headroom. Usually when you want "more", that will just barely cover it, or maybe not at all.

With "higher" supplies you can get MUCH more even with the sloppy chips. (TL072 can stand 36V forever, 38V for years, if you must push it to the max.)

Note that instead of gaining-up SO high, you can reduce gain (or add loss), and work at more-sane signal levels. What you do inside the box is your business. If the interface to the next box needs a HOT signal, only the last stage needs to be heroic.

[bool]

PRR's math dosent mention that the only practically usable use for a RR chip is when you have an adjustable-gain stage - so you are under all circumstances able to hone in and exploit this 2-some volts narrow window of peak output signal ability. IOW you have to be careful with "dynamic" signals. There is a different scenario with pre-clipped waveforms though.

A dual-battery in combination with a suitably low-power consumption non-RR chip will give you proportionally better I/O ability every time - and also a practical, hands-on durability because you will be able to fully exploit the "all the way down to 7-volts" time/voltage window of the 9V battery. In such case, from cca. 19V to 14V.

[MrStab]

Where was this thread when i was trying to figure it out a few months back?!

A few notes from memory - may overlap with what's already been said, may be irrelevant altogether:

• R2R opamps seem to have a lower slew rate on average. Probably not an issue for most things - that's a niche topic of its own.
• R2R's seem to be limited to about 16V, or +/-8V. So no ultra-mega-headroom pedals by combining one with a charge pump. boo.
• afaik, CMOS opamps have similar pros and cons regarding impedance as FETs do against BJT opamps.

i chose rail-to-rail for my parametric EQ project because there's 1 board per band already, so it was just impractical to squeeze more circuitry in there to up the voltage. i figure it'll appease situations where there's a really hot input, like some of those steroid-infused pickups around these days, but if the input's so high that a R2R can't handle it, then there's probably some misuse going on.

...that said, with the limited supply voltage of R2R, pedals that use em are probably more at risk from the "let's try a kettle-plug"-type people. But the vast majority of users will stick to 9V, and that was key to me choosing TLC2274s. As mentioned in greater depth a few posts back, i concluded that if a higher supply voltage is available to begin with, even if R2R's could handle more Vcc, the whole debate would be kinda moot for guitar work after a point.

[PRR]

> if the input's so high that a R2R can't handle it, then there's probably some misuse going on.

Words to remember.

[SISKO]

I thinks R2R are like bridged amps. They are usefull when you have limited power supply voltages. There are better ways to increase headroom/power if you dont have a voltage limit

[blackieNYC]

This has been very helpful. So, it could be helpful to swap TL0s in existing (built) circuits with R2Rs if you are getting occasional clipping, and the input maximum voltage should be looked into - I understand there are a variety of "R2Rs".  Maybe my mixer or parametric EQ. They might be found downstream of some accumulated gain devices.  And that substitution will only get you so much of course, based on the change in max. swing.  Attenuation with an input resistor Would probably serve you better, and isn't impossible to add to a built pedal.  But design-wise, a higher supply yields superior headroom.  Any objections to this somewhat oversimplified takeaway?

[PeterPan]

My first use of an R-R op amp was the dual OP amp LME49721MA. Up until the project where I needed them, I'd be content to create a split supply and use one of the TL0 series, so a 9 or 12V supply could easily be split to +/- 4.5 or +/-6, and as you observed, unless you really need to have a wider swing, that's perfectly fine. I'm not sure my use of the LME49721MA required any less or more parts. Its supply has to be 5V, so its not a super low voltage R-R amp, but the rest of my circuit required a regulated 5V supply, and the circuit also never had to deal with anything beyond line level audio. So I sacrificed one OP-AMP to make a virtual 2.5V reference for my signal 0V, which is pretty similar to what I'd do with a non R-R op amp.

Now from what I've researched, a R-R op amp has to have a specialized internal cross over point, causing different components to activate as you get closer to the rail voltage. Apparently this can cause some distortion, which is harder to control in the lower voltage versions. But as I read, it seemed to me that these limitations have been addressed, at least to the point where any distortion would be small enough to be insignificant for my project. And in fact, it worked out far better than I'd have expected. But it is a consideration. As far as the price goes, yes they do cost a bit more, and so that is a consideration too. Because unless you are talking about a true split supply (not a developed virtual ground), I'm not sure i see where the non R-R op amp supply is going to cost any more.