The slew rate you need is always at least 15-20% more than the slew rate of the op-amp you have.
The recommended value for ruling out any slewing induced distortion is about 10 times the sinusoidal slew rate.
Last edited:
Your recommendation I assume? Based on what?
The most extreme I've seen is from John Curl, factor of 5, but that's already overkill.
Jan
The most extreme I've seen is from John Curl, factor of 5, but that's already overkill.
Jan
Not mine, Texas Instruments,
EDIT: https://www.ti.com/lit/ug/tidu765/tidu765.pdf?ts=1726853861402&ref_url=https%3A%2F%2Fwww.google.co.in%2F
Even I do not personally like to design anywhere near the limit, but the above recommendation is not mine.
EDIT: https://www.ti.com/lit/ug/tidu765/tidu765.pdf?ts=1726853861402&ref_url=https%3A%2F%2Fwww.google.co.in%2F
Even I do not personally like to design anywhere near the limit, but the above recommendation is not mine.
Last edited:
I read about early 1970's solid state based mixing desks that were full of LM741 opamps which had by todays standards, very compromised slew rates. Someone mentioned that some very famous records were mixed on these and the sound quality wasn't too good (Doug Self?). I guess in a mixing desk you will be dealing with +4 dBU level signals which would have strained the 741s a bit at HF.
The recommended SR for a power amp is 1V/us per peak output volt. so, if you build a 100W amp with +-40V rails, the recommended design target is 40V/us.
Tyhe good news of course is the NE5532/34 came along in about 1976 and we never looked back again! That was spec'd at 10V/us which is more than enough for line level signals even at +4dBU.
🙂
The recommended SR for a power amp is 1V/us per peak output volt. so, if you build a 100W amp with +-40V rails, the recommended design target is 40V/us.
Tyhe good news of course is the NE5532/34 came along in about 1976 and we never looked back again! That was spec'd at 10V/us which is more than enough for line level signals even at +4dBU.
🙂
You base your design decisions on the basis of someone's recommendation who's in business to sell you fast opamps? Hmmm.
The term 'as a conservative rule' has absolutely no meaniung. He could have said 20. Or 5.55, with the same justification. Worthless drivel.
Jan
My design preference does not come from this document, but I think it is just more comfortable when you've enough slew rate. Besides, the definition of 'fast' may have changed from what it was in the initial days of integrated circuits. I guess they said that because it's easier to have higher slew rates these days.
I really meant the comment to be more humorous than a recommendation of actual SR headroom. That said, it did come from a design I am currently stewing over. Mostly in that I have over 1K of an op-amp that I more or less got for free and it sure would be nice to use them for the three required for the RF amplifier I am working on (run of maybe 50-60 units). They will work just fine for two of them but for the last (output) stage they just run out of steam. As it happens, I have plenty of another to use for that last stage - it's just the point of it I guess. The ones I have a ton of (ADA4899-1) top out at a SR of 310 and, by formula anyway, I need a minimum 355 plus whatever headroom would be safe.
I know what you mean. It is a but of a variation of murphy's law. Another example might be I need a .1uF cap at 55V but all I have is 50V .1's. Or I need a 1/2W resistor but just have a stock of 1/4's. I did just run out of 1u's for a little project that I thought I had plenty of. I was short 4. The Digi order arrived today so I''l solder them in this weekend. Ordered 100 this time. That'll show Murphy!
The 10x factor is just engineering speak for "significantly greater than" (or "significantly smaller than" if you divide by 10). That doesn't mean that the circuit will fail if the slew rate of the selected opamp is 9.99x the required slew rate. Or 5x for that matter. Sometimes a little perspective can be useful.
Slew rate is one of those "proven by math" things. Either the system has a high enough slew rate to reproduce the highest frequency of interest or it doesn't. Not thereby said that having some margin is a bad thing, but the question is how much.
Human hearing is widely recognized to be limited to 20 Hz to 20 kHz, so if the circuit has a full power bandwidth of higher than 20 kHz it'll be able to reproduce all frequencies that (young!) humans can hear with full fidelity. You can do the derivation if you wish. The highest frequency that can be reproduced with a given slew rate limitation is F_max = SR/(2*pi*V_peak). Rearranging: SR = F_max*2*pi*V_peak, so for 2 V RMS line level (2.82 V peak) you need a minimum of 354 MV/s, also known as 0.354 V/us. For the 4 V RMS pro level you'd need twice that. And even if you want to reproduce 15 V peak as you could possibly sustain with ±17 V supply rails, you're looking at 1.88 V/us.
And let's just say that you want to make sure that even the 4th harmonic of 20 kHz is reproduced faithfully at 15 V peak, you're still only at 7.54 V/us. I would argue that's plenty of overkill for a system that's likely to operate at 2 V RMS with a listener who likely can't hear much past 16-17 kHz. We're not all seven years old anymore... 7.54 V/us is well within the capabilities of just about any modern audio opamp. LM4562 boasts 20 V/us and OPA1611 27 V/us, for example.
So it seems that for audio the opamp isn't the main slew rate limiter.
Tom
Meh! You'll need 97 next time. 🙂
I've also found that I remember that one thing that I had to add to my next Mouser/Digikey order about 10 seconds after clicking submit on my current order. And, of course, it'll be a $17 item. Well below the threshold for free shipping.
Tom
That doesn't seem correct, could it be a typo? When you write out 354 MV/s in scientific notation, then write out 0.354V/us in scientific notation, the two numbers don't appear to be equal.
Actually, today, as ALL digital sources are brickwall band limited, the slew rate requirements are LESS than in da days of 30 ips master tape. Today, you'll only meet a slew limit at line level in an overload situation.
In overload, there is an argument that limited slew rate will lead to better sound as the yucky harmonics will be lower order .. a sorta 'soft clipping'
You're right. It should have been 354 kV/s = 0.354 V/us. Reason #201 to use scientific notation. 🙂
Tom
Apologies for contaminating this topic with yet another serious reply, but if you don't want any slewing-induced distortion at all, you need infinite slew rate.
When an op-amp is in slew rate limiting, its input stage is driven into hard clipping. The input-to-output characteristic of an op-amp input stage is never a straight line that suddenly goes horizontal, so it also distorts at lower rates of change.
What's worse, it can also cause intermodulation distortion. For example, when you play back a 100 kHz and a 101 kHz sine wave that together drive the op-amp close to but not into slew-rate limiting, you are bound to get a 1 kHz frequency component.
Fortunately, the rate of change of most audio signals is rather limited. If it weren't, FM radio pre-emphasis and de-emphasis wouldn't work.
When an op-amp is in slew rate limiting, its input stage is driven into hard clipping. The input-to-output characteristic of an op-amp input stage is never a straight line that suddenly goes horizontal, so it also distorts at lower rates of change.
What's worse, it can also cause intermodulation distortion. For example, when you play back a 100 kHz and a 101 kHz sine wave that together drive the op-amp close to but not into slew-rate limiting, you are bound to get a 1 kHz frequency component.
Fortunately, the rate of change of most audio signals is rather limited. If it weren't, FM radio pre-emphasis and de-emphasis wouldn't work.
Yes, and the exact words of the application engineer where he says 'eliminate any possibility of ....' etc. are worth noting, but Marcel above has taken a swipe on that as well.
I just very recently used LM358 for some audio but it also depends on the application. For instrumentation purposes where more precision is required, I'd like to have a good margin which is clearly not required for the communications / entertainment segments.
When compared to that of TI, the OP's finding of 120% is actually very modest.
Oh, and I've used an LM2904 as the input amplifier (and pre-emphasiser) for an FM based wireless headphone.
Last edited:
100 kHz and 201 kHz may be worse, as the input stages typically produce mostly odd-order intermodulation.
In the conventional textbook model (James Solomon) of an opamp, maximum slew rate is delivered when 100% of the input stage's diffpair Tail Current, is steered into (or out of) the Miller compensation capacitor. To steer ALL the current into one output leg or the other, requires a large differential input voltage. The big input signal "twists" the long tailed pair, either all the way left, or all the way right.
But, I don't want a big differential input voltage! Any voltage difference at the inputs, represents an ERROR: the output is not a perfect replica of the input. And to push the amplifier all the way to the onset of hard slewing, where 99.99% of the tail current is steered into only one of the output legs, requires a very large ERROR. No thank you! Thus my preferences are aligned with @MarcelvdG .
Instead, I prefer to use a rectally extracted guideline: no more than 60% of the tail current shall ever be steered into one output leg, and by symmetry, no less than 40% of the tail current shall ever be steered into one output leg. These are the numbers which happen to make me happy ; other persons may have different thresholds of happiness, of course.
My slew rate rule of thumb becomes: (i) when the output signal is a pure 40 kilohertz sinewave; (ii) at maximum output amplitude; (iii) no more than 60% of the longtailed pair "tail current" shall be steered into either leg.
This reassures me that normal musical signals as discussed in post #15, will not create huge differential inputs. Normal music will not create huge ERROR at the input.
_
But, I don't want a big differential input voltage! Any voltage difference at the inputs, represents an ERROR: the output is not a perfect replica of the input. And to push the amplifier all the way to the onset of hard slewing, where 99.99% of the tail current is steered into only one of the output legs, requires a very large ERROR. No thank you! Thus my preferences are aligned with @MarcelvdG .
Instead, I prefer to use a rectally extracted guideline: no more than 60% of the tail current shall ever be steered into one output leg, and by symmetry, no less than 40% of the tail current shall ever be steered into one output leg. These are the numbers which happen to make me happy ; other persons may have different thresholds of happiness, of course.
My slew rate rule of thumb becomes: (i) when the output signal is a pure 40 kilohertz sinewave; (ii) at maximum output amplitude; (iii) no more than 60% of the longtailed pair "tail current" shall be steered into either leg.
This reassures me that normal musical signals as discussed in post #15, will not create huge differential inputs. Normal music will not create huge ERROR at the input.
_
