Uh, why? Just to put the silliness of the KV2 Audio claim in perspective, 1µs corresponds to 1 MHz. The sample interval of even 176.4 KHz is still 5 times longer.
Last edited:
Yes, KV2Audio is using 20MHz DSD for live audio transfers, and they are good at it, so that "requirement" for 1us might be lower for reproducing recorded audio.
But settling time of 20us corresponds to 50kHz and this might be some issue... or not?
Why am I asking? Because other Class D amplifier have according the experienced listeners problem with "highs", and I'd like to know if the nCore is better in this.
EDIT: but this might be also related to phase issues that are caused by a drop in the amplitude characteristic at high frequencies
But settling time of 20us corresponds to 50kHz and this might be some issue... or not?
Why am I asking? Because other Class D amplifier have according the experienced listeners problem with "highs", and I'd like to know if the nCore is better in this.
EDIT: but this might be also related to phase issues that are caused by a drop in the amplitude characteristic at high frequencies
Last edited:
But DSD sample rate really doesn't have much to do with settling time. What is the settling time (and slew rate) of 20 MHz DSD?
Only if you believe you can hear a 50 KHz signal.
Indeed. As others have pointed out, there are multiple more credible reasons for the perceived "Class D" sound (and the whole idea of a "Class D" sound is a similar stereotype as "Tube sound" or "Transistor sound").
even some high-end manufacturers who stay away from switching amps don't believe in the importance of flat amplitude response past 50kHz.
trying to look for "normally ignored" time-domain errors is a typical high-end game, but to my best knowledge no-one was able to produce credible data.
try to look at it this way. ringing in response to a square wave can be a product of a low-pass 2nd order filter. now, what does it do to a signal in the frequency domain? it does something to the amplitude and (implicitly) phase response.
like I said, non critically damped response (ringing) is indicated if memory serves me well by a small bump in the amplitude response plot at the corner frequency. the NCORE does display such a very small bump.
now, if you want to establish a correlation between settling time and a reported sonic trait, it's a matter of belief, no-one can stop you.
Last edited:
On my experience, every peak in the response curve of a power amp, near the cutoff frequency, is a proof of instability and produce audible artifacts (TIM). I use mostly current feedback amp, and try to get >2Mhz flat bandwidth.
My tip is, first, to work on the feedback loop in order to get no peak, then filter the input to get no overshoot on square waves.
The second tip can be applied to Class D amps as well: never apply a signal with faster slew rate than the amp's one.
I don't believe in TIM. I think it's one of the boogie-men of audio: you're scared of it, except it doesn't exist.
of course, there may be some academic merit in having 10GV/fs of slew-rate but I'm not convinced of the audible benefit of it. the maximum slew-rate that RedBook can demand from a 200W amp in 4 ohms is ~5V/us, and that is the theoretical case of a full-scale 22.05kHz sine, which is never found in music. as I wrote in the past ~78 billion times, Nelson Pass measured SR at the output of an amp and found 0.5V/us as local, if you would, maxima and 2V/us as an absolute one.
even John Curl, the very one who introduced TIM distortion in the '70s, wrote once that TIM is practically irrelevant with RedBook content. there's a post of him somewhere at Audio Asylum where he wrote that, black on white.
of course, there may be some academic merit in having 10GV/fs of slew-rate but I'm not convinced of the audible benefit of it. the maximum slew-rate that RedBook can demand from a 200W amp in 4 ohms is ~5V/us, and that is the theoretical case of a full-scale 22.05kHz sine, which is never found in music. as I wrote in the past ~78 billion times, Nelson Pass measured SR at the output of an amp and found 0.5V/us as local, if you would, maxima and 2V/us as an absolute one.
even John Curl, the very one who introduced TIM distortion in the '70s, wrote once that TIM is practically irrelevant with RedBook content. there's a post of him somewhere at Audio Asylum where he wrote that, black on white.
Band Width
It seems to me that Mr. Putzeys has mentioned that bandwidth limiting is a good idea for audio systems, perhaps to avoid the problems being mentioned here-otherwise we need to mkae the system flat out to ridiculous MHz levels....
Perhaps what people are mentioning here is another reason to try transformer coupling the Ncore input. Some will protest adding a lossy component like this, but I am interested in trying it.
Audience is using transformer coupling on their forthcoming Nc 1200 amp, and we all know about the tall one' use of transformers on his class D based amplifiers...
It seems to me that Mr. Putzeys has mentioned that bandwidth limiting is a good idea for audio systems, perhaps to avoid the problems being mentioned here-otherwise we need to mkae the system flat out to ridiculous MHz levels....
Perhaps what people are mentioning here is another reason to try transformer coupling the Ncore input. Some will protest adding a lossy component like this, but I am interested in trying it.
Audience is using transformer coupling on their forthcoming Nc 1200 amp, and we all know about the tall one' use of transformers on his class D based amplifiers...
and to put things into perspective a bit (I didn't find the exact discussion I was referring to but it looks like this boogie-man called slew-rate is very popular there): Audio Asylum Thread Printer
a few quotes from John Curl himself:
For the record, it is difficult to make a low slew rate power amp that works well right up to the slew rate limit. This is because the input stage will start to add distortion, well before the slew rate limit is reached. If you add local feedback to the input stage to make it more linear before the slew rate limit is reached, then the slew rate will increase from its initial value. To reduce the slew rate back to it's initial value, it would be necessary to reduce the gain-bandwidth of the amp, increasing its high frequency distortion.
fact is that the NCORE has extremely low high frequency distortion, compared to linear amps which reproduce a near-perfect square wave. also, there's a paper by Bruno somewhere on the Hypex website where he explains the distortion mechanisms, which are a bit different in switching amps (I'll look for it). also, I remember him writing somewhere that the practical slew-rate of the NCORE is that of a MOSFET that switches from voltage A to B, which, if you think about it, is correct.
or this one: Audio Asylum Thread Printer
[...] I have to correct a wrong assumption on your part as to what is the optimum slew rate for an amplifier. It is important that the slew rate limit is NOT approached, as this represents VERY HIGH distortion. Also, only small amounts of hi frequency need be present to create a VERY HIGH RATE-OF-CHANGE of the audio signal. Think about a 100HZ sq wave with a 1us rise time. The info at 100KHZ will only be 1/1000 the amplitude at 100HZ, but if it is not there, then the risetime will be compromised. This is an extreme case, but it shows that sine waves can be a poor judge of slew rate potential. We have found for audio that a minimum slew rate of .5V/us for every volt of output either + or -, is necessary. This gives 50V/us for a 100W amp. Most serious designs, even Parasound, has more than twice this number. This info can be shown in Eero Leinonen, Matti Otala, and my paper: 'A Method for Measuring Transient Intermodulation Distortion' in the AES Journal in Apr.1977.
a typical red herring. John is forgetting that we're not listening to square waves, nor can they be produced by digital audio. the way he justifies that a high amount of HF needs to be present to cause SR limiting is ill-conceived. the fact that we can get close (at least graphically) to a square wave without a very large number of harmonic components (Gibbs) is irrelevant. the lower freq. harmonics contribute and, by all means, please do calculate what is the actual slew rate of a square wave that passes through an ideal A-to-D and D-to-A chain with sampling rate of 44.1 kHz. compute the first derivative of the first 'n' terms of a Fourier expansion of a square wave. or simply believe me: the same upper limit derived from first derivative of a full-scale sinusoid of Nyquist frequency applies.
To keep this distortion below 1% or so on this rapidly changing part of the audio signal, it is important that we not approach clipping the input or second stage of the amp. It is interesting that low slew rate amps usually have very nonlinear input stages, because emitter degeneration is typically omitted, and this makes the situation worse.
another red herring. the fact that some amps omit emitter degeneration of the input stage (in fact, I think few actually apply, including the cheapest designs from the '70s and many actually go to lengths wrt to linearization of the input stage) does not prove a thing. a broken Lamborghini does not run, does that prove the Fiat is a better car?
also, this is exactly the issue addressed in Bruno's paper that I mentioned above.
now follows the best one:
I have measured it myself with a fast storage scope and mistracking MC cartridges after RIAA EQ is applied. I hope this clarifies the situation.
mistracking cartridge? need I say more?
now on to Bruno's paper: http://www.linearaudio.nl/linearaudio.nl/images/pdf/Volume_1_BP.pdf
In 1970, Otala identified his infamous “Transient Intermodulation Distortion”, a fancily named manifestation of Slew Induced Distortion. SID is the Chihuahua to slew rate limiting’s Bullmastiff so let’s make sure we understand the latter first.
Slew rate limiting means that an amplifier is trying to reproduce a very fast rate-of-change signal but can’t. The maximum rate of change it can produce is determined by the current available to charge or discharge the compensation capacitor. This current is provided by the input stage and the maximum is the tail current Ib. When a faster rate of change is demanded, the input stage is overloaded and becomes completely unresponsive to any further change. At this point, the feedback loop stops working and is no longer able to control the amplifier.
more follows and I'd advise you to give it a read.
there is some useful info here, if I recall correctly: Distortion In Power Amplifiers
all in all, I think that there's not much you can't see a THD vs. freq/plot or in a CCIF IMD test.
later edit: I think this is the discussion I've been searching: http://www.audioasylum.com/cgi/t.mpl?f=amp&m=21183
It is difficult to get classical TIM with an ideal CD, but SACD and DVD will easily allow it.
(again quoting John Curl)
I'd even contest the "SACD and DVD" part but there you go.
a few quotes from John Curl himself:
For the record, it is difficult to make a low slew rate power amp that works well right up to the slew rate limit. This is because the input stage will start to add distortion, well before the slew rate limit is reached. If you add local feedback to the input stage to make it more linear before the slew rate limit is reached, then the slew rate will increase from its initial value. To reduce the slew rate back to it's initial value, it would be necessary to reduce the gain-bandwidth of the amp, increasing its high frequency distortion.
fact is that the NCORE has extremely low high frequency distortion, compared to linear amps which reproduce a near-perfect square wave. also, there's a paper by Bruno somewhere on the Hypex website where he explains the distortion mechanisms, which are a bit different in switching amps (I'll look for it). also, I remember him writing somewhere that the practical slew-rate of the NCORE is that of a MOSFET that switches from voltage A to B, which, if you think about it, is correct.
or this one: Audio Asylum Thread Printer
[...] I have to correct a wrong assumption on your part as to what is the optimum slew rate for an amplifier. It is important that the slew rate limit is NOT approached, as this represents VERY HIGH distortion. Also, only small amounts of hi frequency need be present to create a VERY HIGH RATE-OF-CHANGE of the audio signal. Think about a 100HZ sq wave with a 1us rise time. The info at 100KHZ will only be 1/1000 the amplitude at 100HZ, but if it is not there, then the risetime will be compromised. This is an extreme case, but it shows that sine waves can be a poor judge of slew rate potential. We have found for audio that a minimum slew rate of .5V/us for every volt of output either + or -, is necessary. This gives 50V/us for a 100W amp. Most serious designs, even Parasound, has more than twice this number. This info can be shown in Eero Leinonen, Matti Otala, and my paper: 'A Method for Measuring Transient Intermodulation Distortion' in the AES Journal in Apr.1977.
a typical red herring. John is forgetting that we're not listening to square waves, nor can they be produced by digital audio. the way he justifies that a high amount of HF needs to be present to cause SR limiting is ill-conceived. the fact that we can get close (at least graphically) to a square wave without a very large number of harmonic components (Gibbs) is irrelevant. the lower freq. harmonics contribute and, by all means, please do calculate what is the actual slew rate of a square wave that passes through an ideal A-to-D and D-to-A chain with sampling rate of 44.1 kHz. compute the first derivative of the first 'n' terms of a Fourier expansion of a square wave. or simply believe me: the same upper limit derived from first derivative of a full-scale sinusoid of Nyquist frequency applies.
To keep this distortion below 1% or so on this rapidly changing part of the audio signal, it is important that we not approach clipping the input or second stage of the amp. It is interesting that low slew rate amps usually have very nonlinear input stages, because emitter degeneration is typically omitted, and this makes the situation worse.
another red herring. the fact that some amps omit emitter degeneration of the input stage (in fact, I think few actually apply, including the cheapest designs from the '70s and many actually go to lengths wrt to linearization of the input stage) does not prove a thing. a broken Lamborghini does not run, does that prove the Fiat is a better car?
also, this is exactly the issue addressed in Bruno's paper that I mentioned above.
now follows the best one:
I have measured it myself with a fast storage scope and mistracking MC cartridges after RIAA EQ is applied. I hope this clarifies the situation.
mistracking cartridge? need I say more?
now on to Bruno's paper: http://www.linearaudio.nl/linearaudio.nl/images/pdf/Volume_1_BP.pdf
In 1970, Otala identified his infamous “Transient Intermodulation Distortion”, a fancily named manifestation of Slew Induced Distortion. SID is the Chihuahua to slew rate limiting’s Bullmastiff so let’s make sure we understand the latter first.
Slew rate limiting means that an amplifier is trying to reproduce a very fast rate-of-change signal but can’t. The maximum rate of change it can produce is determined by the current available to charge or discharge the compensation capacitor. This current is provided by the input stage and the maximum is the tail current Ib. When a faster rate of change is demanded, the input stage is overloaded and becomes completely unresponsive to any further change. At this point, the feedback loop stops working and is no longer able to control the amplifier.
more follows and I'd advise you to give it a read.
there is some useful info here, if I recall correctly: Distortion In Power Amplifiers
all in all, I think that there's not much you can't see a THD vs. freq/plot or in a CCIF IMD test.
later edit: I think this is the discussion I've been searching: http://www.audioasylum.com/cgi/t.mpl?f=amp&m=21183
It is difficult to get classical TIM with an ideal CD, but SACD and DVD will easily allow it.
(again quoting John Curl)
I'd even contest the "SACD and DVD" part but there you go.
Last edited:
in all honesty, I don't expect this TIM/slew rate controversy to be settled before the end of the world, irrespective of the number of hard facts. it's surrounded by much myth and a lot of people who misunderstand it defend its importance with their lives. you can't fight that with the regular weapons of logic.
+1
I owned a couple of Naim power amps decades ago and much liked them. At that time and likely even today, Naim prided itself on bandwidth limiting, eschewing modern high end "sacral" tenets.
Transformers make one of the best output buffers for tube preamps (VTL's most costly $17k preamp employs SS output buffer.) Lynn Olson's tube designs employ inter stage transformer coupling. Steve McCormack's vaunted DNA-500 employs transformer coupled input which he said trounced other types.
Maybe the only reason transformer coupling is not ubiquitous is that there is no such thing as a low cost quality transformer. They always cost real money. Too much quality raw materials and too much hands-on labor.
Last edited:
Please, audio is not a matter of believing. Just do the following: Set the output signal of an amp to the exact input level (attenuation = gain of the amp), then, subtract the input signal from the output one. Now, look (and listen) to the resultant signal with a real musical signal.
Do the same with sinusoidal signal and compare.
As far as i am concerned, and after had read such a incredible amount of pseudo science and unjustified magical smoke under the John Curl signature, I think that this audio designer can not be cited as a *serious* reference in anything related to electronic for audio.
As an example, John Curl is proud to pretend (in several posts):
"First, I showed him a short piece of BEAR's silver wire. I said this is pure silver, VERY expensive, per foot, and takes 30 days break-in on the reel, and then another 30 days once wired in the product. He was 'impressed'. "
And in an other post, that they were marking the cable on the reel (just a single conductor) for their mounting sens !!!!!
I don't expect this to be a valid test because of phase shift.
yes, he's funny at times (the Bybee stuff) but he did "discover" TIM.
I'm sure you can imagine ways to equalize phase shifts on both signals before comparing them
.And i believe you had read several papers on methods to measure transient intermediation distortion, including a recent one published by AES.
Hum, we where exploring that kind of distortion in the early 70th in my R&D office, proof it was *in the mood* at this time, then i had read stuff about TIM under Matti Otala (1970-1972), and Walter Jung signature.
The 1972 Matti Otala paper is, indeed, co-signed by JC: i bet this one was sitting at the right of Galileo Galilei when he discovered that the Earth was round, and used to eat apples and oranges with his friend Newton
At the end, nothing mysterious about the fact that a servo system has to be faster than the fastest signal he is supposed to follow and that closed loop feedback is a servo system.
I think we all agree on that, but also on that if that condition is satisfied, there is no distortion.
no, but count me interested and I'd bet other readers are. this subject does interest me a lot but that 70s Otala study I don't count as relevant, if only for the simple reason that it tried to prove a point the wrong way.I'm sure you can imagine ways to equalize phase shifts on both signals before comparing them
And i believe you had read several papers on methods to measure transient intermediation distortion, including a recent one published by AES.
another way of putting it: let's take two imaginary amps of very low distortion at all frequencies. let's equalize the phase, level match their outputs and compare them. what will we see that eludes, say, that infamous CCIF IMD (19k+20k sines at equal levels) test?
and yet another way of putting it. many so-called fast amps have pretty high distortion at high-frequencies. rumor (good luck with finding any measurements because there aren't any) has it that Spectral amps are ultra-fast (high SR, that is) but have moderate linearity. I hope we all agree this means that as long as we eliminate phase-shift from the equation, that full-scale 20kHz waveform will not look exactly like a sinusoid. how does it differ from a sine? what is the error waveform and in which way is it "better" than that of a "slower" amp?
well, yes, a lot of stuff keeps floating around until one day a guy decides to put his name on it (the light bulb is an example, if I recall correctly).Hum, we where exploring that kind of distortion in the early 70th in my R&D office, proof it was *in the mood* at this time, then i had read stuff about TIM under Matti Otala (1970-1972), and Walter Jung signature.
The 1972 Matti Otala paper is, indeed, co-signed by JC: i bet this one was sitting at the right of Galileo Galilei when he discovered that the Earth was round, and used to eat apples and oranges with his friend Newton
yes, but when you actually observe the behavior of that system with the fastest signal that will ever be present at its input (again, the full-scale sine of Nyquist frequency) and there aren't any weird things going on, what's left?
of course, one can devise a theoretical nonlinearity that passes a sine undistorted but reacts differently to more complex signals. but are such complex nonlinearities present in an amp?
maybe it's awkward and hard to believe, but I know people who are able to design an amp from scratch with the aid of a piece of paper and a pencil only, but who also hold plainly wrong beliefs about what their amp is actually amplifying and the slew-rates actually present at the input. I actually ahd a conversation with a guy who designed his own amp but also believed (believes?) that when you add sines of < 20kHz frequencies, insane slew rates are born.
this is why I'm rather skeptical wrt TIM.
I was affraid
As it is analog, it is "more or less"
To get back to the class D subject, if the switching is PWM the feedback is analog. Means that the (fast) slew-rate will not not correlated with the feedback bandwidth.
The actual speed of power mosfets is not fully satisfying to get enough margin (eq resolution) at high frequencies, like in the beginning of Digital recordings.
For basses and medium reproduction, it is perfect, for treeble, and high end reproduction, i still find an advantage in fast analog amps. Not so boring: with multi-way amplification we can use efficient and economical class D amps where we need much of the power, and less efficient analog amps where we need less power.
Anyway, comparing Ncore to analog amps, even in treble reproduction, if the difference is audible, nothing so catastrophic when listening to real music in real world. And, nothing so different in character or presentation.
As it is analog, it is "more or less"
Even in analog you have the situation of "if A is larger than B, then A is not limiting B", so you can have a "no problems (at all)" situation.
I don't understand this. In a power amp, each stage (and specially the inverted input one in a voltage feedback topology) introduce its distortion and non linearities (and delays). And we can suffer more from the (theoretical) 0.01 % distortion of a power amp, than from the 4% of distortion of our loudspeaker.
At the end, hifi reproduction is a "make believe" story.
About analog power amps, my religion is to use current feedback topologies, (because they offer *constant and fast* slew-rates), try to make the open loop distortion and bandwidth as good as possible, get a response curve and square waves with no peak or ringing, then low pass the input signal as much as possible.
My input/output comparison process is very disappointing, as it reveal a huge amount of distortion, even in high end systems, during big transients like kick drums. And the awful power supply influence on them.
Playing with the protection system schematic you can see in my signature, even when the system is tuned to be perfectly stable at 20Khz sin wave full power, the first kick-drum at medium level will fire on the protection. You'll need to adjust-it with more margin.
Yes, we need overkill margins on everything (current, bandwidths etc) to get a decent reproduction system. Then how many records with good enough recordings and mixs ?
Last edited:
- Status
- Not open for further replies.