Approximately 1khz on CD
One guy I met said he uses a Garrard 401 to help design digital equipment . His point being that the 401 has strong personality . Very often digital copies of digital sound fine . However the same equipment didn't sound like a 401 . The test can be rerun , where as real life it is gone before any chance to correct it . 401 right or wrong should sound identical on a CD copy . For identical I mean like FM can fool you . FM looks so poor on paper and CD so good . That's the big deal .
there's no basis for that, do you have a signal that exemplifies it?
it still looks like you're advocating 2 opposite views in the same reply.
no music contains fast slew rates - I agree and all the practical measurements agree.
but you go on and say that for some reason CD should be able to reproduce a "good" square wave. and I say it doesn't. and then I say that you just claimed the opposite.
such thing doesn't exist in music. those mild-ringing-like things are Nyquist and Shannon telling you "you know that's not possible".
let's say that there are square waves exiting the mic, the guitar pickup etc. somewhere inside the ADC there will be a LPF that will alter it, it won't be a square wave anymore. if it's done in the analog domain it will have rounded edges. if it's done digitally it will have whatever shape, depending on the filter type. ok, so maybe you say "wait a minute, there's the actual problem". well, if you want for an unexplained reason to maintain the shape of the wave you'll have to settle for a much higher sampling rate. which means much higher bandwidth. which means higher slew rate. but the actual question is why do you need that at the input of the amp, especially when it's you who said that the reason high SR is needed is not because it exists in music?
LOL
Approximately 1khz on CD
I wanted to post a similar pic but was too lazy to do so. where is the surprise? it's supposed to look like that. what's the point?
Then the measurements are wrong, i.e. not taken comprehensively enough.
For proof, do try reproducing Vangelis' piece "Metal Rain". There's a longer slow intro passage, on the soft side, and then some mean electronic devices come crashing in at full blast. The speed and volume will relieve you of your pants if you had turned on the intro loud enough. Perhaps even your head, if the missus is at home.
Just because the instruments used are electronic in no way denies them the identity of musical instruments - I've heard that argument a few times, like "they are not natural instruments" - natural? Really?
Approximately 1khz on CD
One guy I met said he uses a Garrard 401 to help design digital equipment . His point being that the 401 has strong personality . Very often digital copies of digital sound fine . However the same equipment didn't sound like a 401 . The test can be rerun , where as real life it is gone before any chance to correct it . 401 right or wrong should sound identical on a CD copy . For identical I mean like FM can fool you . FM looks so poor on paper and CD so good . That's the big deal .
Nige, just use 741 op amps at the output to get rid of that nasty ringing.
I have started reading the BBC history link.
That link states that the FM signal was transmitted as a 13bit 32ks/s.
Companding of the 14bit signal only got considered as an adjunct to bringing in Nicam.
Is this saying that FM got the companded signal in 1971, or only the TV sound channel?
That is not the way I remember it.
Is there any evidence that the link has got it's history wrong?
Last edited:
Nige, just use 741 op amps at the output to get rid of that nasty ringing.
I was building a SE valve amp the other day and got a mild version of that . At first I was upset , the cause was well understood and a trade off . Then I said forget it . OK for CD OK for my amp . Sounds great . Looking at all other tests nothing bad was found .
The valve amp in question started life as an Alex Kitic design of ECC 81 and EL 34/KT88 . This uses a sort of feedback from the anode of EL 34 to anode of ECC 81 . Not really anode to anode . The point is the alleged function of the ECC81 becomes a V to I converter feeding a I to V converter as we have we a so called trans-impedance stage ( VAS ) . I did many tests and can confirm it seems a realistic description . What one ends up with s a super triode by look of curves .
I abandoned the circuit but not the idea . I chose to find a valve that could deliver 4 times an ECC81 and give an actual 3 times the current in the application . The result seems to confirm one can not have too much drive current . Measurements say very little to confirm it .
So the maximum slew rate from the source would be 2 Pi f times amplitude (divided my a million if we want V/us), where I assume that the maximum f is one-half the sampling rate, or whatever can get through the filter.
So yeah, a 22 kHz square wave would come out as a sine wave, and the truncated Fourier series of other waveforms would make little nasties, or deformations. Let's hope that the phase doesn't vary with frequency, in band, since that also makes those ringing-like artifacts.
But, internally, our power amplifiers would still need to be capable of a much higher slew rate, so that the feedback could work fast-enough. And, of course, amplifying the amplitude amplifies the slew rate. So the amplifier's maximum output slew rate would be the overall gain times the maximum input slew rate.
So let's see.... 6.28 x 22 kHz x 1V / 1000000 = 0.1375 V/us, maximum, from the CD source, assuming 1V amplitude. So with a gain of 30, the power amp's output should only slew at up to 4.125 V/us.
So yeah, a 22 kHz square wave would come out as a sine wave, and the truncated Fourier series of other waveforms would make little nasties, or deformations. Let's hope that the phase doesn't vary with frequency, in band, since that also makes those ringing-like artifacts.
But, internally, our power amplifiers would still need to be capable of a much higher slew rate, so that the feedback could work fast-enough. And, of course, amplifying the amplitude amplifies the slew rate. So the amplifier's maximum output slew rate would be the overall gain times the maximum input slew rate.
So let's see.... 6.28 x 22 kHz x 1V / 1000000 = 0.1375 V/us, maximum, from the CD source, assuming 1V amplitude. So with a gain of 30, the power amp's output should only slew at up to 4.125 V/us.
Last edited:
Being very simplistic . A full power square-wave at what frequency would be a good test ? A guess is great . From CD we might say 1 kHz . From my valve amp I have to . Above that ?
Someone said under certain conditions notching of waves can be seen when slew rate isn't enough . If real music I am convinced . If I need JBL tweeters to know it I am still convinced .
It it lovely making a sine-wave from a square-wave via a 10 pole filter . Interestingly the last four poles seem the same on the scope . Fourier analysis says differently .
Now did anyone find this ? If a wave seems to look pure it sometimes still has about 3 % THD . That amp will sound OK . Another amp of 3 % THD will not look good on scope and not sound good . I feel increasing if it looks OK it sounds OK . Perhaps it is obvious why ? Just surprised our eyes see it as clearly as ears hear it . Speakers must obey this law ? If not they would be unusable . I speak of the vast majority and not my panels .
Someone said under certain conditions notching of waves can be seen when slew rate isn't enough . If real music I am convinced . If I need JBL tweeters to know it I am still convinced .
It it lovely making a sine-wave from a square-wave via a 10 pole filter . Interestingly the last four poles seem the same on the scope . Fourier analysis says differently .
Now did anyone find this ? If a wave seems to look pure it sometimes still has about 3 % THD . That amp will sound OK . Another amp of 3 % THD will not look good on scope and not sound good . I feel increasing if it looks OK it sounds OK . Perhaps it is obvious why ? Just surprised our eyes see it as clearly as ears hear it . Speakers must obey this law ? If not they would be unusable . I speak of the vast majority and not my panels .
It was perhaps in a complicated way saying even 1 kHz is a bit high . I don't read too much into this except why didn't someone question how we interpret it .
I suspect why CD galloped into production was the Nicam had been trouble free and very neutral . A LP record sounded like a LP record via FM+Nicam . What I suspect wasn't right is that Philips did no golden ear listening , certainly not as the BBC did . One can sort of understand why . 13/14 bit is OK 16 bit better without any need to discuss it if Philips . Secrecy also . I can also understand how not monkeying about with compression must seem better . How many bits needs so as not to need Nicam ? Answer 16 bits , job done .
I wrote sometime ago to Tim de Paravacini to ask if he really said tape deck bias signal frequency relates to sampling rates . He wrote back and said mostly yes . Put simplythe bias frequency might easily convert into a good sampling frequency and the engineers should have known it . I am sure Tim would have made it more interesting and accurate if him saying it . I felt he said it was essentially working as digital does . Obviously parts of the process are not the same .
a good test for what? I can go on ad infinitum: was it not you who said that amps need slew-rate for internal workings and not because music has really fast signals? what are we testing? the amp or the digital part of the CD?
maybe it looked like my CoolEdit snapshot meant to say "it seems perfect visually, so it has 0 distortion". I posted that pic because you mentioned something about SR and I wanted to emphasize that inter-sample delay has nothing to do with it. now, if some RedBook implementers deviate from ideal by a lot, it's not my problem. are we discussing if it can be done badly? I don't think so.
this is a FFT of a CCIF IMD test of the output of a Meridian 808 CD player.
we're not looking at waveform that seems perfect even if it has 3% distortion. we're looking at a FF(ourier)T that reveals that. largest intermodulation product is @ -100 dB and I'd even bet that's because of analog circuitry nonlinearity.
yes, we can go on and on and on and round and round and round if you tell me that a less than full-scale test signal would reveal larger distortion (really, would it?) but I'm still naive enough to think that one day we'll have less speculation and more facts here...
I still do not understand why you agreed with two opposed things and my initial question remains unanswered.
I am not familiar with Nicam and can't comment. but instinct tells me that there was nothing special about it and the reason it sounded good has to do with factors that are not intrinsic to the format. I'll look it up though.
Tom, you can also approach it from the other side. An amp that has to output say 40V peak at 20kHz needs a certain slewrate, and that's it.
According to your calculation, 6.28 * 20kHz * 40/1 million, comes out to 5V/uS IIGTR.
That automagically includes its amplification and input signal.
jan
Last edited:
Nigel, I don't understand why a square wave would be a good test? For what?
A 1kHz square wave will look reasonable, a 10kHz square wave will look horrible, on a 20kHz system.
But what's the point of that result?
jan
what I find most bafling on audio forums is that in time one gets to understand that the subjectivists are only another types of objectivists. they simply have their preferred set of measurements, but for unexplained reasons. the much-talked-about correlation with sound is much more speculative than would seem, maybe even less than with the standard set of measurements. in this case, a similar measurement is used in 2 cases, but for unexplained reasons one is preferred: the visual shape of sine waveform is not relevant (although it actually is in this case), while the shape of a square wave is (although it isn't). if that's not double standard...
Nige, just use 741 op amps at the output to get rid of that nasty ringing.
dvv;
Your basic understanding of bandwidth limited systems is limited.
Sure, feed a TTL generated square wave into an amplifier. Either you low pass filter it intentionally before signal hits active component or you work locally with active component to maintain linearity.
Cool Edit waveform display is typically quite good representation of what is happening when waveform is constructed with given bandwidth.
Gibbs effect isn't nasty; it is very real.
That 44.1kHz sample rate captures all meaningful bandwidth for human hearing reflects upon the true limits of hearing.
Push-pull, basic insecurity me thinks. Even as a die-hard subjectivist, you cannot really prove the result you think you hear. So one option is to shore it up with measurements.
But of course you can't rely on the measurements of the objectivists, that would be giving in. Hence the slew of nonsensical measurements that in addition are wrongly interpreted....
jan
But of course you can't rely on the measurements of the objectivists, that would be giving in. Hence the slew of nonsensical measurements that in addition are wrongly interpreted....
jan
ok, looked up NICAM. it's is some sort of adaptive PCM. when I was much younger and before I started having the most basic understanding of sampling theory it seemed to me rather obvious that intuitively adaptive quantization shoud be better. maybe it really is but if one thing is sure, we're stuck with linear PCM.
The only advantage of companded PCM over a longer word linear PCM is lower bit rate. The CD system has better low level resolution than NICAM.
The sampling tied to line rate version described by AndrewT put the data in the line sync pulses. We called it Sound in Sync and was a system used by the then Post Office, now BT, for video conferencing
Maybe NICAM sounded good because it was only 15 kHz. CD fed signals into the amplifier/speaker chain in the 15 to 20 kHz range that were almost absent in legacy sources and contains little music, but a lot of recording studio artifacts
The sampling tied to line rate version described by AndrewT put the data in the line sync pulses. We called it Sound in Sync and was a system used by the then Post Office, now BT, for video conferencing
Maybe NICAM sounded good because it was only 15 kHz. CD fed signals into the amplifier/speaker chain in the 15 to 20 kHz range that were almost absent in legacy sources and contains little music, but a lot of recording studio artifacts
- Status
- Not open for further replies.