Do i really need high slew rate opamp at the output of my CD-player?

You don't. To pass the audio frequency range at 5Vp/p without distortion you need less than 1V/us slew rate (from my memory, but it is easy to calculate). The much appreciated OPA2111 has only 2V/us and any decent opamp will have 20-25V/us. More than that will only open the bandwidth to other potential problems.
 
Last edited:
A key question is, how effective is your I/V filter? Is the signal filtered pre or post I/V. Post I/V filtering would, of course, leave the I/V circuit fully exposed to the wideband output the DAC chip. This is recipe for A.C. errors in a feedback based active I/V stage, of which, slew rate limiting is only one.
 
Last edited:
I just went through a huge learning experience with testing different types of op amps and slew rates in a digital line. Once I figured out how slew rate affected everything, all the different opinion threads just went out the window (I tried xxx op amp and it's better than yyy op amp).

If you don't know, the output of the DAC chip itself is basically straight-line voltage injections of different levels (based on the 16-bit or 24-bit word from the digital source). This is seen by the I/V stage as a stepped square-wave form. There is nothing analog here. The I/V stage will attempt to reproduce this as best it can (with a slight low-pass filter if you use a capacitor negative feedback loop). A very fast op amp will try its best to keep the square wave form because the "slew" is how fast the op amp raises to the voltage to match the stepped voltage level. Since we only hear sound during changes in voltage (translated to movement of the speaker), the fast op amp will "slew up" extremely fast and the rest of the straight-line voltage injection from the opamp will be what it is - a flat line. Since only a small portion of this is actually a "raise in voltage", we will hear little of this. Also, since the slew is so fast, much of this will automatically get translated to higher frequencies that is what is intended. This is why a lot of solid state equipment sounds bright/harsh with no bass.

You really need a slew rate that is high enough to capture the high frequencies properly and not sound boring. If the slew rate is too slow, then you will actually have high-frequency roll-off. A slower slew rate will translate more bass in the signal, but may not have enough high frequency excitement. I have found that a slew rate of 20-28 V/uS is the best.

A convential op amp has a straight line slew (it will slew up at one rate to get to the voltage as fast as possible). A Class A circuit is completely different as it generally has an "S Curve" slew rate. In this circuit, more bass will get translated because the initial part of the slew is very slow. It can also be smoother and fuller in sound because of the S-Curve and that the negative slew is slower and not just a straight/fast nose-dive. See page 7 of the Sparkos op amp datasheet here:

http://sparkoslabs.com/wp-content/uploads/2014/06/SS3601_SS3602.pdf

There are different quality between op amps. I have found that the opa2134 (which is so popular), even though it has a slew rate of 20V/uS, is really not a great op amp. It's muddy sounding with a lot of distortion in the highs. It's also not fast enough to have good separation of instruments and high frequency excitement. A set of dual opa827 op amps (slew of 20V/uS) are much better. It is also cleaner sounding since the settling rate is faster. An extremely fast opa627 (slew of 55V/uS) will attempt to reproduce the wave form as exactly as it was seen on input. This is great for a preamp if you have an excellent source signature you just want to "pass through". (also great for recording/mixing equipment where you want to preserve the original analog waveforms). The Sparkos Class A op amp has a slew of 20V/uS (and 12V/uS on the negative slew side) sounds very clean and awesome when you pay attention to the power supply.

Once you know the engineering characteristics of op amps, you can understand where to best use them. There is no "best op amp" in any situation. Only different op amps with different expected characteristics.
 
Last edited:
Ideally, the circuit's slew rate should be at least 5 times the maximum output signal's slope
for minimum distortion.

How do you explain this? Slew rate is a circuit's ability to draw the function 2*Pi*f*V. For f=20kHz and V=5.6Vpp (2V RMS) the required slew rate is approximately 0.8V. Why do I need 5 times higher slew rate than what is required for non-distorted output of the input waveform?
 

jcx

Member
2003-02-17 7:38 pm
..
for simple bjt op amp there's a relation from Solomon's 1974 paper that links slew rate from dominant pole compensation to the amount of tanh distortion of a simple bjt diff pair

http://www.ti.com/lit/an/snoa737/snoa737.pdf

but you don't have to choose simple, dominant pole compensated or bjt for good performance today, several techniques modify the input stage distortion relation to slew rate
 
Last edited:
How do you explain this? Slew rate is a circuit's ability to draw the function 2*Pi*f*V. For f=20kHz and V=5.6Vpp (2V RMS) the required slew rate is approximately 0.8V. Why do I need 5 times higher slew rate than what is required for non-distorted output of the input waveform?

Slew rate is measured by overloading the amp input stage. It overloads because it's not fast enough to follow the rise of a square wave. So slew rate is the amps recovery time it takes to get back into its linear region not a measure of its bandwidth (speed) That's why the 5 ( I've heard 10 ) times.
 
Slew rate is measured by overloading the amp input stage. It overloads because it's not fast enough to follow the rise of a square wave. So slew rate is the amps recovery time it takes to get back into its linear region not a measure of its bandwidth (speed) That's why the 5 ( I've heard 10 ) times.

I have a different understanding of what you said here. If an op amp slew rate is not fast enough for a given input frequency, a rolloff will occur, not overload. Also, I think that thew slew rate is measured in the fastest (straightest) part of a sine waveform.
 

rayma

Member
2011-04-29 8:37 pm
If an op amp slew rate is not fast enough for a given input frequency, a rolloff will occur, not overload.

Slewing is rather like clipping, but due to excessive signal slope, rather than excessive amplitude.
During slewing, the signal waveform is replaced by the amplifier's innate limiting output slope.
A sine wave input would indeed measure lower peak amplitude, but would also be altered to have
a sort of sloping "crossover distortion" near the zero crossings, which is what lowers the peak amplitude.

The maximum signal slope of a sine wave, which occurs at the zero crossings, is equal to the
peak amplitude x 2Pi x frequency, since d/dt (A sin(2Pi x f x t)) equals 2Pi x f x A cos(2Pi x f x t).
This has a maximum value of 2Pi x f x A, where t=0 and cos(0) = 1, at the sine's zero crossing.

For example, a CD player's maximum slope output with a 20kHz sine input is around
2V x 1.414 x 2Pi x 20kHz, or 0.36V/uS maximum slope.
 
Last edited:
Slewing is rather like clipping, but due to excessive signal slope, rather than excessive amplitude.
During slewing, the signal waveform is replaced by the amplifier's innate limiting output slope.
A sine wave input would indeed measure lower peak amplitude, but would also be altered to have
a sort of sloping "crossover distortion" near the zero crossings, which is what lowers the peak amplitude.

The maximum signal slope of a sine wave, which occurs at the zero crossings, is equal to the
peak amplitude x 2Pi x frequency, since d/dt (A sin(2Pi x f x t)) equals 2Pi x f x A cos(2Pi x f x t).
This has a maximum value of 2Pi x f x A, where t=0 and cos(0) = 1, at the sine's zero crossing.

For example, a CD player's maximum slope output with a 20kHz sine input is around
2V x 1.414 x 2Pi x 20kHz, or 0.36V/uS maximum slope.

Ye, but there is a feedback loop, anyhow any modern opamp of quality will be overkill in this application, no squares at the end of a dac current output or voltage output.

Lampizator is a great way to kill an excellent CD player :Olympic:
 

jcx

Member
2003-02-17 7:38 pm
..
as I pointed out - there is a relation of nonlinear distortion at, say, audio frequency and amplitude and the slew rate limit
this happens while the loop is still working, before hitting the slew rate limit - that's why the recommended numbers for slew rate limit are multiples of the max signal slew rate

but there are ways around the simple bjt dominant pole compensated case from 1974 - easy to find modern op amps where the relation is not applicable and, often, the slew rate limit is way higher too
 

dotneck335

Member
2012-12-11 10:10 pm
I just went through a huge learning experience with testing different types of op amps and slew rates in a digital line. Once I figured out how slew rate affected everything, all the different opinion threads just went out the window (I tried xxx op amp and it's better than yyy op amp). There are different quality between op amps. I have found that the opa2134 (which is so popular), even though it has a slew rate of 20V/uS, is really not a great op amp. It's muddy sounding with a lot of distortion in the highs. It's also not fast enough to have good separation of instruments and high frequency excitement. A set of dual opa827 op amps (slew of 20V/uS) are much better. It is also cleaner sounding since the settling rate is faster. An extremely fast opa627 (slew of 55V/uS) will attempt to reproduce the wave form as exactly as it was seen on input.
So you are saying the difference between these opamps (2134, 827, and 627) is due to the measured differences in settling time rather than slew rate?? The data sheets quote these figures: 2134---20v/uSec slew rate, 1000 nSec settling time; 827--- 28v/uSec and 850 nsec; 627---55v/uSec and 550 nSec.
 
So you are saying the difference between these opamps (2134, 827, and 627) is due to the measured differences in settling time rather than slew rate?? The data sheets quote these figures: 2134---20v/uSec slew rate, 1000 nSec settling time; 827--- 28v/uSec and 850 nsec; 627---55v/uSec and 550 nSec.

Well, you're close, but it not really fair to say the settling of the 827 is 850ns. You really need to compare based on within "xx percent". Different spec sheets have different ways of measuring. If you look that the specs for "settling within 0.01%" of the signal, you will get the following:

OPA2134 ==> 1000 nsec
OPA827 ==> 550 nsec
OPA627 ==> 550 nsec

When an op amp slews up to match the input signal, it typically overshoots the mark (it's does not track it perfect). Then it bounces up and down some until it finally hones in on matching the signal. The "settling time" is how fast it will dial in to match the exact input signal. When the OPA2134 is bouncing all around, it's causing a bunch of noise/junk. The OPA827 only slews half as fast as the OPA627, but it settles just as quickly -- which is a really great feat for an FET op amp. A good reference:

http://www.analog.com/media/en/tech...863287538299597392756AN359.pdf?doc=CN0269.pdf

Now a capacitor in the negative feedback loop can help remove some of this settling noise, but I can still hear problems with the opa2134 (even with a very good power supply). I hear the same character with the opa2107 (which has a worse settling time). They both have what is known as that classic "burr brown" warm signature. Some may like this, but I have found it muddy. It's definitely not the "crystal clear" sound. I don't have an oscilloscope and I don't know exactly what's happening with this, but I can definitely hear it. The 2134 and 2107 do sound slower/warmer and muddier. At the same time, the high frequencies can come across as a little harsh/bright. I believe this is from the settling errors as it sounds very much like clipping (or flatlining) when you starve an op amp circuit.
 
I just checked and OPA627s at Mouser are $25 EACH!!!! Anyone know of a cheaper source?

Not unless you want to buy a fake 627 from ebay, lol. That's the going rate, there are no "special deals" to be found anywhere. Actually, your looking at the AU version, which is nice. However, I have found the BP version (at $35) to be vastly superior. It is definitely a lot cleaner sounding. Make sure you have a 0.01uf MKP connected directly to the +/- power pins. Just bypassing each power rail to ground with 0.01uf is not enough, as the 627 can still oscillate some and sound a little brittle and "solid state" in the high frequencies without this. It is definitely smoother on the highs with that 0.01uf between power pins. I'm using OPA627BP in all 5 positions within a circuit on a Rane stereo line mixer. It's an excellent (yet expensive) op amp for this purpose. I've heard it said that this particular op amp is expensive because it's manufactured with old machine technology that is very expensive for TI to maintain. It also could be they are charging a high amount because of the "prestige" this op amp has generated. I would not recommend the BM metal can version. It's very clean in the highs, but it lacks body and the gain is just not there. Not sure if I got a bad batch, but all 4 op amps from Digikey had this character.

You could try looking at using the Analog Devices ADA4627-1BRZ (make sure you get the BRZ). This is, essentially, the AD version of the OPA627 and it has very respectable specifications. People that have tried it have been very impressed. Many have said it beats the 627. The BRZ are only $10 each, but they are only available in SOIC. You'll have to buy the 627 if you need a DIP version. I have not had the opportunity to try the ADA4627 yet, but it's on my list of things to do.

You definitely want to run these op amps with a very good higher voltage power supply (+/-12V split power supply at a minimum). My equipment is running +/-15V. You may get poor performance on stuff like headphone amps that run only at +9V.
 
Last edited: