Op-Amp Supply Voltage and Slew Rate

[redbagy]

Hi! I bet that this question has been asked before but I was reading on running circuits at 18V - specifically this article (https://www.linkedin.com/pulse/can-i-run-my-guitar-pedal-18v-jon-cusack/) and whilst I agree that 9V might be enough for some drive circuits especially when clipping is used, sometimes a higher supply voltage does give a slightly better presence (to my ears at least and also depending on the type of circuit). Thus I am trying to understand where this effect (higher presence/highs) comes from. Electrosmash mentions the slew-rate (https://www.electrosmash.com/klon-centaur-analysis#big_power_supply) however I'm curious as to whether there are other factors apart from this?

[ElectricDruid]

Given that the Slew rate of an op-amp is stated in V/µs without reference to the supply voltage, it ought to get *worse* at higher supply voltages, since larger signals would mean it has further to travel, so a slew rate limitation would become progressively worse. That would *limit* top end and cause distortion at higher signal levels. Of course, a little distortion on higher frequencies *might* be heard as "extra presence" so perhaps that's why. But if that were so, it would be because the op-amp is *failing* to do a good job, not because it's doing a better job at 18V.

I've never heard of Slew rate improving with supply voltage, and the datasheets don't mention this, unless I'm missing something (wouldn't be the first time... ) so I don't agree with Electrosmash on this.

[Rob Strand]

It's hard to generalize but the source of slew-rate is a current source driving a base-collector capacitance in parallel with a compensation capacitance.   Typically the current source is constant but if anything it would increase a bit with supply voltage.   The capacitances tend to decrease with voltage.  So both those would point to an increase in slew-rate with supply voltage.



Just to be clear.  If the slew-rate was constant a higher supply voltage would increase the *time* to transition from one supply to the other but the slew-rate is the slope.   The slew-rate changing with supply voltage means the slope is changing (it of course affects the time as well).    The time and the slope are completely different things.

[FSFX]

In most cases the Miller capacitor that people use in their circuits completely dominates the performance of the circuit overriding any contribution slew rate makes to performance. This is particularly relevant in the Klon where the 820p feedback capacitor on the summing amp stage and the tone control negate any changes to the slew rate of a TL072 device.     

[antonis]

IMHO, for 18V (+/-9V) supply, even the humble 741 could manage the audio work..

[ElectricDruid]

IMHO, for 18V (+/-9V) supply, even the humble 741 could manage the audio work..

Well, we know that the 741 can do audio, because it *did* for most of the 70s and 80s!

But out of interest, how well? The TI datasheet says 0.5V/us for the slew rate.

https://www.ti.com/product/LM741

That's 36us to get from -9V to +9V, and another 36us to get back again. That means the fastest signal you could get out of it at full amplitude would be a triangle wave of period 73us, which is 13.8KHz.
In practice, it can't get as far as the rails anyway, so we could probably look at -7V to +7V or so as a bit more realistic. That'd push that triangle wave up to 17.8KHz.
So, yeah, it can reproduce audio, just about. Unless we're doing tests, we're unlikely to want to listen to full volume signals in the high treble anyway, so real-world audio will be an easier case.

[ElectricDruid]

It's hard to generalize but the source of slew-rate is a current source driving a base-collector capacitance in parallel with a compensation capacitance.   Typically the current source is constant but if anything it would increase a bit with supply voltage.   The capacitances tend to decrease with voltage.  So both those would point to an increase in slew-rate with supply voltage.

So the datasheet figure is probably quoted at a typical +/-15V supply or some such, right? (LM741 says it is, for example) How much do you think it varies from the given figure across the supply range, Rob? For some op-amps, that could be pretty wide - from +/-3V up to +/-18V or more. Say a five or six-fold range. Do you think the slew rate is proportional?
It'd be handy if it was, since that would mean the performance didn't change across the supply range.

[Rob Strand]

So the datasheet figure is probably quoted at a typical +/-15V supply or some such, right? (LM741 says it is, for example) How much do you think it varies from the given figure across the supply range,

For the LM741 it's going to be +/-15V; the conditions written at the top of the tabulated data.   For other opamps you might see different conditions.

Rob? For some op-amps, that could be pretty wide - from +/-3V up to +/-18V or more. Say a five or six-fold range. Do you think the slew rate is proportional?
It'd be handy if it was, since that would mean the performance didn't change across the supply range.

Yes it's a large range to cover.  Just the same I suspect on the whole SR pretty constant since the designs would try to keep the front-end bias currents constant and the BC capacitance would be swamped by the compensation cap.   Too many things varying will upset bandwidth and stability.   The trend I mentioned is likely but the effect is small.   

There's quite elaborate schemes to keep the bias current constant.  The LM741 makes some effort but it only uses a Widar current source, good enough but you can do better. 

I pulled out a transistor based LM741 model and got these results.  It won't simulate any voltage dependency of the compensation cap but it will account for bias current variations and transistor Cbc.

Model LM741_QS2F - more convincing slew-rate waveform
Buffer (as per datasheet)  Rising edge:
+/-15V: input swing -10V to 10V: 0.650V/us
+/-6V: input swing -2V to 2V: 0.477V/us
+/-3V: input swing -1V to 1V: 0.375V/us

So you can see it struggles a bit at the +/-3V end,  SR drops by a factor of 1.7.
Not an insignificant change at the extreme.

As far as representing reality I don't know.   I haven't checked out how much the bias current changed.
It's likely to show something simple like SR proportional to bias current.


FWIW, the lowest voltage for the LM741 isn't 100% clear.  The specs only go down to +/-5V but that is already below a 9V single ended supply.   The Onsemi/Motorola specs imply 5V single ended (but it's not stated clearly).  An old Motorola datasheet shows the open loop gain dropping about 18dB with +/-2V rails.




This is more like what you would expect from modern parts, but still interesting a 9V single ended supply is on the fringe of the datasheet,



The circuit starts to deviate at low voltages because the constant current circuit hasn't got enough voltage to function at 100% - not really dropping out but losing regulation.

[PRR]

> the 741 can do audio, because it *did* for most of the 70s and 80s! ....the fastest signal you could get out of it at full amplitude would be a triangle wave of period 73us, which is 13.8KHz.

That's how it got the bad rap. Disco was young. If you run disco drums at clipping, the important cymbal sizzles come out as 9+KHz triangle waves. With enuff coke, we didn't care (the last of the acid-heads could hear cymbals in silence). But a few folks noted that a well-worked '741 would smear cymbals; in extreme case clarinet and violin too. (The old spectrum graphs seem to have been very mellow players.) Yes you can turn it down and the TEAC 10 mixer was supposed to work all -10dBu internally ('5532 at the end to deliver "studio level"). The one and only Gately mixer Kit was '301 with 4.7pFd comp caps, so was 8X faster than '741 (sadly very hissy).

As to why 18 sounds different than 9: some of it is the 2X gain-structure change needed for it to make a difference (you overload at a different part of the signal chain), and some of it is in our head. After all, these 9/18V signals sometimes go to 300V tube preamps, and ultimately to 5V ADCs to cut your WAV and MP3.

[redbagy]

Thank you all for your input about this and especially for refreshing my memory about slew rate. Not to jump to another topic (and I still haven't read the Electrosmash analysis) but does the Klon need a 27V headroom? If the TL072 op-amp slew rate is 13V/us that would give approximately a maximum period of 4.15us which would result in 240kHz for a triangle wave (correct me if I'm wrong). So unless a very large amplitude signal is needed to span across the entire 27V, are there other reasons as to why the Klon uses this supply range?

Regarding as to why 18V sounds "better" I tend to agree with @PRR and @ElectricDruid i.e. it's probably the overloading of a different part in the chain or a little distortion in the higher frequencies.

[FSFX]

FWIW - I built a slew rate test circuit some time ago to check out some LM308s and compare with OP07s and other op amps.
Here are my results and a comparison with the TL071 type of op amp.

[antonis]

Well, we know that the 741 can do audio, because it *did* for most of the 70s and 80s!
But out of interest, how well? The TI datasheet says 0.5V/us for the slew rate.
https://www.ti.com/product/LM741
That's 36us to get from -9V to +9V, and another 36us to get back again. That means the fastest signal you could get out of it at full amplitude would be a triangle wave of period 73us, which is 13.8KHz.
In practice, it can't get as far as the rails anyway, so we could probably look at -7V to +7V or so as a bit more realistic. That'd push that triangle wave up to 17.8KHz.
So, yeah, it can reproduce audio, just about. Unless we're doing tests, we're unlikely to want to listen to full volume signals in the high treble anyway, so real-world audio will be an easier case.

Tom, you know that full-level signals at 20kHz (or so) simply don't occur in reality..
(the energy on the high frequency end of audio spectrum is much lower than that of the low frequency end..)

After all, internally compensated for unity gain stability op-amps have lower slew rate than if they were compensated for higher gain..