anyone like to post examples of "modern" commercial source material which has slew requirements greater than that for 44.1kHz sampling Red Book CD?
In Jurassic days, there was stuff with supersonic content above 20kHz but all MUCH less slew demanding than 20kHz FS square wave which would be 20kHz sine wave under Red Book CD.
In Jurassic days, there was stuff with supersonic content above 20kHz but all MUCH less slew demanding than 20kHz FS square wave which would be 20kHz sine wave under Red Book CD.
@MarcelvdG Do you know offhand how to calculate the maximum risetime of white noise in the 20Hz - 20kHz band assuming white noise frequencies are all cos(w)=1 at t=0?
Also, besides the long time delay of a near-ideal brickwall filter, might not settling time be an issue too? Say, for example, if we have something like a high-Q notch filter, it takes time to stop ringing after an impulse, so doesn't take just as long to settle into filtering a sine wave that just suddenly starts up (a sine wave times a step function)?
One reason I ask is that some resampling filters in HQ Player are described as being more optimized for time-domain response as versus frequency-domain response. Depending on the type of music one would like to listen to in the audio band, it is suggest that one type of filter or the other might be a better choice.
That sort of thing being the case, what are the tradeoffs in designing an antialiasing filter for CD? And, how much of an impulsive signal might pass through one way or another?
With an ideal brickwall filter, you mean?
Also, besides the long time delay of a near-ideal brickwall filter, might not settling time be an issue too? Say, for example, if we have something like a high-Q notch filter, it takes time to stop ringing after an impulse, so doesn't take just as long to settle into filtering a sine wave that just suddenly starts up (a sine wave times a step function)?
One reason I ask is that some resampling filters in HQ Player are described as being more optimized for time-domain response as versus frequency-domain response. Depending on the type of music one would like to listen to in the audio band, it is suggest that one type of filter or the other might be a better choice.
That sort of thing being the case, what are the tradeoffs in designing an antialiasing filter for CD? And, how much of an impulsive signal might pass through one way or another?
Last edited:
Why do we need to know that? Isn't comparing input square wave with output square wave enough of a test for slew-rate limiting? If identical, then no slew-rate limit, right?
What numbers tells you?
No, but I think what you describe is a Dirac impulse filtered by an ideal filter. Without the lower bandwidth limit, the result would be a time waveform with a sin(x)/x shape that rings at 20 kHz, so apparently x = 2π • 20000 Hz • t. You would then have to take its derivative to time and find its maxima, for example by again taking a derivative to time and equating that to zero. With the lower bandwidth limit, the result probably won't change much, but the calculation gets more complicated.
As a veteren of the slew rate wars of the '80s (Spectral DMA100) let me help clarify what matters. First Slew rate is a large signal parameter and somewhat independant of small signal performance. However slamming an amp with a square wave to get a slew rate number doesn't get a useful number. Its like looking an an amp that is hard clipping to measure its power output. Useful slew rate number would come from for example the max frequency at full output with distorion below 3%. Its an indication that the amp is still in feedback control. On an early amp project (Spectral CPU100) I devised a slew rate meter. It was fun except the only thing that got much action was record pops and tics. That amp was good for 500V/uS. Also pretty sensitive to loads and marginally stable.
Another large signal parameter thats not discussed but could be significant for audio is settling time- the time for a system to recover to a small percentage from a transient. This is difficult to measure accurately but could explain some of the differences that distortion measurments don't show. The challenge is a clean transient so you are not measuring the source and high resolution of the transient if you are looking at .01% settling time.
I have found in my experience faster internal circuitry, higher slew rate and shorter settling time seem to translate into better sound. However I could be fooling myself.
Another large signal parameter thats not discussed but could be significant for audio is settling time- the time for a system to recover to a small percentage from a transient. This is difficult to measure accurately but could explain some of the differences that distortion measurments don't show. The challenge is a clean transient so you are not measuring the source and high resolution of the transient if you are looking at .01% settling time.
I have found in my experience faster internal circuitry, higher slew rate and shorter settling time seem to translate into better sound. However I could be fooling myself.
err..rh! Stuff which goes on to a CD is already brickwall filtered so has MUCH less slew demand than 20kHz FS square wave which would be 20kHz sine wave under Red Book CD.
Yes. I know there's theoretically SACD and other zillion MHz sampling stuff. But if you have an actual example, please post this in WAV or other zillion MHz lossless format so we can see this unicorn for ourselves.
PS No one seems to point out the one real life example which might exceed Red Book CD slew limits ... but this has even less relevance to domestic music reproduction 😊
20khz square wave has 1000x the sine wave slew rate. It becomes Triangle wave if it's slew limited.
With 'modern' brickwall band limited source material (ie ALL modern stuff), the only time you might meet a slew limit would be with a signal driving your amp WELL INTO CLIPPING. Then the issue isn't settling time but overload recovery. I agree with Bob Cordell that this is one of the biggest factors deciding the 'sound' of an amplifier.
Many (all?) Golden Pinnae amps show instability & obvious xover distortion when coming out of overload on real speaker loads ... sometimes for several seconds. No wonder they sound different
Well, there is sound reinforcement of live music. A close mic'ed small cymbal hit hard can produce some very high frequency ultrasonic content. Usually the mic preamp and or capsule will hit its limit before the cymbal does.
You referring to Benchmark AHB2 trying to drive Sound Lab electrostatic speakers? If so, maybe that's why I stopped using that amp on my newer Sound Lab 645 Majestic speakers. The Marantz MA9-S2 handle the loads a lot better.
It is filtered at some point, true, but then its typically played back through an oversampling multibit sigma delta dac. There goes all that brickwall filtering. There is so much RF spewing out of the dac chip analog output is difficult to find opamps that can actually handle it for I/V conversion. Yes, OPA1612 is a favorite, but a lot of RF garbage still exists after that first I/V stage. What's left can be hard on the balanced MFB filter that commonly follows. Of course, some folks don't like the sound of all those IC opamps, so they try to do the I/V conversion by other means...
Last edited:
I pointed this out for some Golden Pinnae amps whose designers are on this forum and was banned so you'll excuse me if I don't mention names
I was re-instated when my accusations were proved true 😊Actually the result of the brickwall filters, ie nothing of the original source above 20kHz STILL holds. RF crud on the output is a separate matter but the cure to that is a better DAC or CD player.
I'm not holding my breath for someone to post a zillion MHz uncompressed file off a commercial player ... with sufficient RF to demand more slew than a 'perfect' Red Book CD player ... but I'm prepared to be surprised.
In da previous Millenium, ALL properly conducted DBLTs on bandwidth limiting including mine, showed that of those who could reliably tell the difference, ALL of them preferred a brickwall 20kHz filter. Back then, many of the sources used still had significant analogue ultrasonic content. These were the early days of CD and digital recorders.
When I emerged from the bush some 15 yrs into this century, I was amused to find ALL properly conducted DBLTs on bandwidth limiting including some conducted by august members of this forum showed the same results even though by this time, all the source material was digital and hence already brickwall limited.
I can pontificate at length on why this was the case for the previous Millenium tests but don't have any sensible explanation for the tests this Millenium. One possible but highly improbable explanation is that even the very small amounts of ultrasonic crud was sufficient to induce slew distortion in modern amps.
The practical take on this is. A box with a 20kHz brickwall will actually increase the musicality of your siystem, If you make a zillion $$$ out of this, could I beg perhaps $0.01 for each device sold? 😊
I'll give you another observation in return: Sometimes a 256kbps or 320kbps MP3 can sound better than a CD, because the lossy encoder throws out some of non-musical grunge in an imperfect recording. Fidelity in that case can still be pretty close to CD if not quite perfectly there.
Actually, a well recorded CD can sound very good on a good system. IME, many or most people have no idea. However hi-res can sound a little better. Often they are mastered differently which may account for much of the difference in sound.
That does not seem surprising to me.
On a properly bandwidth-limited recording, any output above fs/2 on the playback end has a corresponding error below fs/2. The brickwall filter is putting the energy back where it belongs.
Ed
On a properly bandwidth-limited recording, any output above fs/2 on the playback end has a corresponding error below fs/2. The brickwall filter is putting the energy back where it belongs.
Ed
As you are in the habit of driving amplifiers into clipping, maybe the phase shift of the filter happened to reduce the peak level, so the clipping was less hard? I know some broadcasting stations used all-pass filters for peak reduction, although I don't know the details - in particular, whether they use fixed or adaptive all-pass filters.
If "properly conducted" means that side effects like phase shift in the passband, roll-off in the passband, clipping, slewing or anything else that causes intermodulation in the amplifier or loudspeakers have all been excluded as a cause for the heard differences (but your last sentence seems to contradict this), then your test subjects must have been able to hear frequency components above 20 kHz while listening to music (and as the human auditory system is not linear time invariant, this doesn't imply that they could hear sine waves above 20 kHz). You then need more than 20 kHz of bandwidth if you want the reproduction to be accurate rather than musical.
If I remember well, I have seen equipment with switchable low-pass filters in the past. I think you could even reduce the cut-off frequency to 7 kHz to make it even more musical.
The nice thing about CFA’s is they do not slew rate limit. If you feed a VFA with a sine wave to drive it to full power and then jack the frequency up, on most well designed modern amps, at about 150 to 200 kHz (design specific dependent) it will start to slew. On a CFA, you just get a reduced level sine wave out. Even if you increase the input level up further, it just remains a sine wave.
I still bandwidth limit my amps because with VERY fast rise/fall time signals I can force the input diamond buffer into class AB operation. But here I am talking <<1us full power stimulus. I usually set the BW limiting filter to >300 kHz, but on some as high as 500 or 600 kHz.
Slew rate limiting on modern VFA’s is not a thing. This issue has been solved as knowledge about the causes and how to compensate VFA’s correctly has spread through the design community (although you’d never guess it reading some of the reviews and anti-feedback dogma over on Stereophile).
I still bandwidth limit my amps because with VERY fast rise/fall time signals I can force the input diamond buffer into class AB operation. But here I am talking <<1us full power stimulus. I usually set the BW limiting filter to >300 kHz, but on some as high as 500 or 600 kHz.
Slew rate limiting on modern VFA’s is not a thing. This issue has been solved as knowledge about the causes and how to compensate VFA’s correctly has spread through the design community (although you’d never guess it reading some of the reviews and anti-feedback dogma over on Stereophile).