Hi Guys
"I was introduced to this forum by a coworker who restarted my interest for hi-end audio, but I mostly get some good laughs when I visit this forum.
Good laughs indeed"
Why not share your wisdom?
"I was introduced to this forum by a coworker who restarted my interest for hi-end audio, but I mostly get some good laughs when I visit this forum.
Good laughs indeed"
Why not share your wisdom?
Is there a small chance that you might address my three questions? I'm not looking to start a comment-war or anything, just my very old 3 q's. Thx! GoatGuy
One more answer to the question discussed contains the following link (see Section 1, page 1.108): ADI - Analog Dialogue | Op Amp Applications Handbook
AD specialists advise to provide from 5 to 10 full power bandwidth (FPBW) margin with respect to the maximum frequency of input signal. In our case it means 100..200 kHz of FPBW. In terms of slew rate it corresponds to ~ 18..36 V/micsek (for 28.3V output max and 20kHz).
Unfortunately, they don't explain that demand to the OP AMP. It seems to me, it is necessary in order to permit the FB of the OP AMP to fix from 5 to 10 harmonics of distorted signal of maximum frequency. But it's not clear, why fore this job of FB the slew rate should corresponds to the full power, taking into account march less amplitude of harmonics?
AD specialists advise to provide from 5 to 10 full power bandwidth (FPBW) margin with respect to the maximum frequency of input signal. In our case it means 100..200 kHz of FPBW. In terms of slew rate it corresponds to ~ 18..36 V/micsek (for 28.3V output max and 20kHz).
Unfortunately, they don't explain that demand to the OP AMP. It seems to me, it is necessary in order to permit the FB of the OP AMP to fix from 5 to 10 harmonics of distorted signal of maximum frequency. But it's not clear, why fore this job of FB the slew rate should corresponds to the full power, taking into account march less amplitude of harmonics?
in my experience with John you should just buy a lottery ticket, worry about being struck by lightning, falling space junk... instead - any time the answer would require using numbers, understanding or even acknowledging any but his own exact formulation of a problem
even a a basic like how many times so far has the need to Normalize slew limit numbers to the Vmax at which ever point in the system been mentioned - does John Curl take any notice?
I have tried to show IMD math, reasoning, sims when Curl repeats his faith based Otala "flat loop gain over audio" mantra - the results are marginal, even past hints of acknowledgement fleeting - the "conversation" restarts with John repeating his oldest position anew every 6 months as if we hadn't gone around and around many times before
Last edited:
In measurements, you can really see distortion shoot up when approaching the slewing limit. A safety factor of 5-10 seems about right.
Obviously a 20 kHz full power sine is somewhat of a worst-case assumption for an audio amp, as discussed earlier in this thread. People used to build audio gear using µA741 opamps back in the '70s, and it didn't sound as bad as you'd think despite their rather limited slew rate (a whopping 0.5 V/µs). I guess you better keep all kinds of RF away from the inputs though...
EDIT: I just grabbed I track that I thought may have highish slew rate requirements - track 8 from Dire Straits infamous hi-fi demo album "Brothers in Arms", A/C version of original issue, deemphasis applied, snare hit @ 1.46 s. Steepest transient, when the waveform is assumed scaled to 28.3 Vp, comes out to about 1 V/µs (so 10 V/µs ought to be fine for this). That's the equivalent of like 5-6 kHz fullscale if I have my math right, so definitely more than 2 kHz. Not 20 though.
Last edited:
Goatguy, I don't have any definitive answer to your questions, and this would be the wrong place to get started on answers, even if I had more facts to present.
Headroom? We design first and spec it afterward. i.e. we start somewhere to get the required voltage swing into 8 ohms, then adjust the output stage to get much more power out at lower resistance, you know 4,2 etc. This costs money, so sometimes similar amps are sorted out more by current drive than voltage swing. Finally, the actual limits of the best devices available become the limiting factor for voltage swing, and we have to make even more complex (more parts) circuits to get the most voltage swing, as well as higher current output.
What I am addressing on this thread is what is possible (reasonable worst case) for either rise time from the audio source or the slew rate necessary to make an 'effortless' transfer of the audio signal to the loudspeaker. Nothing more. Of course, there is more to successful amp design than just slew rate, and that is why many noted designers (some even here) have reduced their slew rates down from many hundreds to 50-100V/us over the last few decades.
There are tradeoffs here, just like in making a 200mph automobile. Not much need for one, especially in the USA. Same thing with very high slew rates, it might even make the amp even sound better, but not because of the need for a higher slew rate. Even the 50-100V/us slew rate recommendation might be more for reducing PIM distortion or FM modulation, than slew rate limiting itself. Still, the best solid state amps with global feedback added, benefit from a relatively high slew rate because of lowered TIM and PIM that is possible.
Headroom? We design first and spec it afterward. i.e. we start somewhere to get the required voltage swing into 8 ohms, then adjust the output stage to get much more power out at lower resistance, you know 4,2 etc. This costs money, so sometimes similar amps are sorted out more by current drive than voltage swing. Finally, the actual limits of the best devices available become the limiting factor for voltage swing, and we have to make even more complex (more parts) circuits to get the most voltage swing, as well as higher current output.
What I am addressing on this thread is what is possible (reasonable worst case) for either rise time from the audio source or the slew rate necessary to make an 'effortless' transfer of the audio signal to the loudspeaker. Nothing more. Of course, there is more to successful amp design than just slew rate, and that is why many noted designers (some even here) have reduced their slew rates down from many hundreds to 50-100V/us over the last few decades.
There are tradeoffs here, just like in making a 200mph automobile. Not much need for one, especially in the USA. Same thing with very high slew rates, it might even make the amp even sound better, but not because of the need for a higher slew rate. Even the 50-100V/us slew rate recommendation might be more for reducing PIM distortion or FM modulation, than slew rate limiting itself. Still, the best solid state amps with global feedback added, benefit from a relatively high slew rate because of lowered TIM and PIM that is possible.
Sgrossklass, thank you for some real input. Many here have not noted the difference in the open loop linearity between the input stage and the output stage when pressed to their extremes and nearing clipping. We know that a clipped input stage gives the slew rate limit, but below this limit, the input stage is creating both TIM and PIM in relatively large quantities.
Now to defend Matti Otala (RIP) a little: He was the first to promote wide open loop bandwidth, and he got high slew rate as well in his designs. The two went together in many designs. However, Matti seriously avoided the term 'Slew Rate' to the point of absurdity in the 1970's, and he got caught out by doing so, by Bob Cordell and others. However, by 1977, Matti realized that TIM could not be the only factor that changed the audio quality of an amp, and he started working in PIM. He was then short circuited by Cordell, and Lipshitz et al, so that he could not further publish in the AES, so he never developed PIM fully. Only when Barrie Gilbert brought out his article on op amp linearity, (15-20 years later) did we get back on track with PIM. It would appear that PIM or some relative to it, one of the fundamental limiting factors in making the best audio amp possible.
Now to defend Matti Otala (RIP) a little: He was the first to promote wide open loop bandwidth, and he got high slew rate as well in his designs. The two went together in many designs. However, Matti seriously avoided the term 'Slew Rate' to the point of absurdity in the 1970's, and he got caught out by doing so, by Bob Cordell and others. However, by 1977, Matti realized that TIM could not be the only factor that changed the audio quality of an amp, and he started working in PIM. He was then short circuited by Cordell, and Lipshitz et al, so that he could not further publish in the AES, so he never developed PIM fully. Only when Barrie Gilbert brought out his article on op amp linearity, (15-20 years later) did we get back on track with PIM. It would appear that PIM or some relative to it, one of the fundamental limiting factors in making the best audio amp possible.
Its a shallow question going nowhere - other than a figure- and what to do with it then ?
-throw it around the garden perhaps.
There needs to be discussion about how steepest audio transients are
attempted to be captured in recordings and then played back in the real world.
ftp://ftp.dbxpro.com/pub/pdfs/WhitePapers/Type IV.pdf
Which is a far more serious matter, with a generation lost in MP3 and
few if any hearing what is possible from audio equipment.
Cheers / Chris
-throw it around the garden perhaps.
There needs to be discussion about how steepest audio transients are
attempted to be captured in recordings and then played back in the real world.
ftp://ftp.dbxpro.com/pub/pdfs/WhitePapers/Type IV.pdf
Which is a far more serious matter, with a generation lost in MP3 and
few if any hearing what is possible from audio equipment.
Cheers / Chris
"definitely more than 2 kHz. Not 20 though"
SGROSSKLASS, good day and thank you for reply.
You result, it seems, confirms acknowledged opinion, that in average the musical signal energy concentrates in the low frequency region.
About 10 years ago I experimented with FET output stage - measured it's input current on picks of output voltage. I had got about 30kom. With ~ 400p input capacity of output stage it gave ~ 14 kHz on picks of music. That is, fullscale level (we assume it 28.3v) is achieved at the high end of sound bandwidth. So, our results slightly differs🙂. May be because your results may be considered as more "average" and mine - as less "average"?
But I was going to talk about some different issue. Yes, I completely agree with AD engineers - to permit FB to eliminate harmonics of distorted signal, necessary to vast the bandwidth of Amp up to 5..10 times more, then max input signal frequency (our 20 kHz). But why it should be the full power bandwidth, when harmonics are very small?
SGROSSKLASS, good day and thank you for reply.
You result, it seems, confirms acknowledged opinion, that in average the musical signal energy concentrates in the low frequency region.
About 10 years ago I experimented with FET output stage - measured it's input current on picks of output voltage. I had got about 30kom. With ~ 400p input capacity of output stage it gave ~ 14 kHz on picks of music. That is, fullscale level (we assume it 28.3v) is achieved at the high end of sound bandwidth. So, our results slightly differs🙂. May be because your results may be considered as more "average" and mine - as less "average"?
But I was going to talk about some different issue. Yes, I completely agree with AD engineers - to permit FB to eliminate harmonics of distorted signal, necessary to vast the bandwidth of Amp up to 5..10 times more, then max input signal frequency (our 20 kHz). But why it should be the full power bandwidth, when harmonics are very small?
I have been arguing that point but without the theoretical knowledge to back up my "claims".
Peak transients are very similar all over the audio spectrum.
Average levels are very much concentrated in the middle range and very much lower in the treble region.
I suspect a lot of the "myth" of lower levels of treble transients is simply down to the peak indicating instruments not capturing and displaying the real peaks. Look at the traditional VU meter. It cannot display peaks. But good technicians know how far they can push without noticeable distortion in the recorded signal.
J.Curl seems to be confirming that treble transients are very fast and very short and thus very high slewing.
But if we clip them at the recording stage, we never see those fast transients in the reproduction stage.
Peak transients are very similar all over the audio spectrum.
Average levels are very much concentrated in the middle range and very much lower in the treble region.
I suspect a lot of the "myth" of lower levels of treble transients is simply down to the peak indicating instruments not capturing and displaying the real peaks. Look at the traditional VU meter. It cannot display peaks. But good technicians know how far they can push without noticeable distortion in the recorded signal.
J.Curl seems to be confirming that treble transients are very fast and very short and thus very high slewing.
But if we clip them at the recording stage, we never see those fast transients in the reproduction stage.
Last edited:
I think the people looking at transients and hence required slew rates were not that ignorant. Note that max slew rate might not correspond to signal peak, even in an amplitude sense. In a waveform sense, peak slew rate can never be at signal peak as signal peak (by definition) has zero slope.AndrewT said:
But if you stick an HF ripple on top of a mid or bass waveform, the peak slope is at high signal level.
And the slope up the side of that ripple is steep.
It is as if the treble peak was starting from the 0volts level and getting to that peak value in a tiny fraction of the time that the "whole" signal waveform took in getting from zero crossing to the peak.
Could someone simulate a dual frequency signal?
1.5Vac of 1kHz and 0.5Vac of 20kHz =2Vac of the simplest complex waveform.
What is the steepest slope anywhere in that waveform?
Peak value will be 2.8284Vpk
max slope is?
And the slope up the side of that ripple is steep.
It is as if the treble peak was starting from the 0volts level and getting to that peak value in a tiny fraction of the time that the "whole" signal waveform took in getting from zero crossing to the peak.
Could someone simulate a dual frequency signal?
1.5Vac of 1kHz and 0.5Vac of 20kHz =2Vac of the simplest complex waveform.
What is the steepest slope anywhere in that waveform?
Peak value will be 2.8284Vpk
max slope is?
Last edited:
Yes, and the amplitude of the HF will be small so the slew rate will be small.AndrewT said:
Why simulate when you can calculate? The steepest slope will be if the phase relationship is such that the maximum slopes of both components coincide. This will be at the zero crossing.
2 pi ( 1.5 x 10^3 + 0.5 x 2 x 10^4 ) V/s. I make it 72V/ms.
This is the situation folks: You CAN have virtually any rise time from a square wave. In theory it is instantaneous from one voltage to another. ONLY an added filter to the square wave can give you a defined or controlled rise time. Generally the audio sources have worst case rise times, now up to 6us, although 10 us rise time might be more universal. It can be shown that 10us rise time can be transmitted by analog tape, MC phono cartridges, or hi end digital. But not CD or MP3. So those of you who are exclusively into CD or MP3 should not worry much about rise time. However, in hi end, these are not the only sources. That is why we have to design amps to pass a 10 us or faster transient effortlessly when we design beyond mid-fi.
A spectral display alone can give misleading view of what rise time can be. You can't just take 1 or 2 or 3 individual tones and generate a realistic worst case rise time without appearing to be unbalanced at the high end of the audio spectrum. However a SQUARE WAVE, can generate a rise time virtually removed from what the spectrum appears to be over the audio bandwidth. For example, a 100Hz square wave could have a 1ns rise time (10us is more probable) yet on a spectrum analysis be 40dB down at 10K from 100Hz and still falling at higher frequencies. This can be misleading, because it might be almost impossible to pass the 100Hz square wave at modestly high output levels (100Hz audible) without slew rate limiting of the amplifier that is not designed for 10us rise time and faster audio signals.
Now what makes high transient signals? Not a summation of a few sine waves, but percussion or just hitting a couple of wood blocks together might do it. Serious audio designers design for every possible combination, not just a chosen few, like a sine wave, or even 2 tones.
This is why we developed the TIM(30) test signal from a rise time limited 3.15KHz square wave with an added 15KHz tone at 1/4 the level of the square wave. The bandwidth limiting comes from a 30kHz R-C filter deliberately added to the test waveform. This is the test standard today for low TIM and included in much modern test equipment. Pass this and you are OK, fail this and stick to CD's or MP3. Your choice.
A spectral display alone can give misleading view of what rise time can be. You can't just take 1 or 2 or 3 individual tones and generate a realistic worst case rise time without appearing to be unbalanced at the high end of the audio spectrum. However a SQUARE WAVE, can generate a rise time virtually removed from what the spectrum appears to be over the audio bandwidth. For example, a 100Hz square wave could have a 1ns rise time (10us is more probable) yet on a spectrum analysis be 40dB down at 10K from 100Hz and still falling at higher frequencies. This can be misleading, because it might be almost impossible to pass the 100Hz square wave at modestly high output levels (100Hz audible) without slew rate limiting of the amplifier that is not designed for 10us rise time and faster audio signals.
Now what makes high transient signals? Not a summation of a few sine waves, but percussion or just hitting a couple of wood blocks together might do it. Serious audio designers design for every possible combination, not just a chosen few, like a sine wave, or even 2 tones.
This is why we developed the TIM(30) test signal from a rise time limited 3.15KHz square wave with an added 15KHz tone at 1/4 the level of the square wave. The bandwidth limiting comes from a 30kHz R-C filter deliberately added to the test waveform. This is the test standard today for low TIM and included in much modern test equipment. Pass this and you are OK, fail this and stick to CD's or MP3. Your choice.
Last edited:
Sleving rate in term of voltage rise time has nothing to do with real acoustical step response of the real speaker. If you were able to absorb 5V/us converted into sound with 86 dB/W efficiency with your ears from normal listening distance you'd probably immediately lose your hearing.
I've counted no less than 14 errors, misunderstandings, conflicting things in the above text.
Now, to stay in this forum rules, should one poor EE take these 14 errors and spend at least half a day trying to debunk them one by one (likely to no end result, since it will be either ignored, "I know better, I'm a famous designer" or simply dismissed based on listening experiences (not DBT, of course)?
Or calling the text a pile of <ahem> will do?
Or simply ignore the post and it's author?
- Status
- Not open for further replies.