John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.
So you have an ideal brickwall AAF with no overshoot in the time domain whatsoever? :confused:

There is none in the A/D/A conversion, but bandlimiting can cause variable group delay issues, depending on implementation and your definition of "issues". The take-away word here should hopefully be implementation. I understand that some fancy new home D/A convertors give the user a variety of filters to choose from on the playback end. Might be fun for advanced homebrewers.

It's always worth keeping in mind that all mechanical widgets will add group delay "issues" to the sound, including speakers and rooms. There's no such critter as verbatim reproduction.

Thanks,
Chris
 
There is none. The bandlimiting itself can cause variable group delay issues, depending on implementation, but it's not inherent in the A/D/A conversion process.

Thanks,
Chris

None? No Gibbs type phenomena due to the sharp bandlimiting? No infinite negative or infinite positve ringing energy from the ideal sinc function based bandlimiting reconstruction filter? Really? You might want to have another look at those theories by Shannon and Nyquist. I think you'll find that the presence of such time-domain distortions are actually essential to perfect frequency-domain reconstruction. These are not group-delay or phase distortions. The time-domain distortions I'm refering to provide the missing frequency-domain signal content not directly captured by the original sampled points.

The sampling theorem itself features both positive and negative infinities as a signal requirement. That means perfect sampling and reconstruction assume a perfectly constant signal, constant to the degree that such signals should have existed since the big bang and will continue to exist until the end of the universe. Obviously, that is not possible, but what it does tell us is that while bandlimited sampling and reconstruction is theoretically perfect for signals which carry only frequency-domain sensitive information, it is not perfect for signal which also carry time-domain sensitive information. This, IMHO, has been THE flaw with audio CD from the start. The time-domain sensitive nature of music recieved little to no consideration when the format was defined.

So, CD gave us perfect sound forever, as far as spectrum analyzers are concerned, but not as far as the time-domain sensitive instrument such as the human hearing system is concerned.
 
Last edited:
sure - just list all of the properly blinded, psychoacoustically controlled listening tests that shows they're audible with music - even for 10% of the listening population

after all shouldn't "night and day" differences be able to AB/Xed?

Would that be the same population that is perfectly happy listening to MP3 that you're refering to? As far as the 10% who you assume to be more observant listeners, perhaps you would care to list those "properly blinded" listening test which show an inability for any listeners to distinguish sharply bandlimited CD digital recording and playback from say, 100% analog low-noise mastertape recording and playback.
 
The phonograph had 100 years of development before the CD was introduced. At that time most of the records sold were produced by well trained and experienced crews. Now the same level of skill is present in some CDs. But at the initial release there were lots of issues that were thought to be unimportant but turned out not to be. There was also a learning curve on how to best record for digital.

Now the difference between what can be delivered by the current CD standard and what is being encoded is converging. In live sound more resolution in the DSP is used. However most of the pros I know prefer the sound of cost no limit analog, but use digital mixing consoles for all of the other advantages.

Now my current system project uses a digital board as there are only two analog sources, the announcer and anthem singer. As the anthem singer uses a wireless microphone, the digital "losses" are not an issue. The big problem is the digital link from the console to the audio power amplifiers.

One can use Cobranet for the link, but it has design limits that I prefer to avoid. AES is up to the quality required but has transmission length issues. So I will have to build my own fiber transmission system. So to say digital is as fully developed as were records is probably not accurate.
 
Do you listen to impulses and step functions? There is a direct relationship between the time and frequency domain responses of a system.

If there is any frequency content near the upper bandlimit then yes, we are in part listening to the system impulse response. Otherwise, bandlimited sampling theory has no import.

Yes there is a direct mathematical relationship between the two domains. It is the Fourier transform. What one needs to understand is that the more rectangular (perfectly bandlimited) the signal channel is in the frequency-domain, the less perfect will be it's transform in the time-domain. This is the fundamental trade-off of bandlimited sampling and reconstruction.

To have both perfect frequency and time domain repsonses requires oversampling well beyond what's required to perfectly capture the frequency-domain content alone. For example, look at the sampling rates of digital oscillloscopes. You'll find that they sample at many times what's required to meet Nyquist for their maximum specified input signal bandwidth. This is because an oscilloscope is primarily a time-domain tool. Brickwall bandlimiting filters, such as utilized with CD, would produce unacceptable distortion of the oscilloscope's time-domain waveform.
 
This, IMHO, has been THE flaw with audio CD from the start. The time-domain sensitive nature of music recieved little to no consideration when the format was defined.

So, CD gave us perfect sound forever, as far as spectrum analyzers are concerned, but not as far as the time-domain sensitive instrument such as the human hearing system is concerned.
There is no flaw with CD, no more than any other type of sound reproduction mechanism. Yes, there all sorts of subtle distortions one can point the finger at, inherent in the process of acquiring and reconstituting the waveform structure, but the ear/brain is amazingly accommodating provided you both supply it sufficient information, and also not overload it with highly unrelated material. Faults which a lot of digital is prone to, unfortunately.

An analogy for getting digital sound reproduction right is trying to use an extremely powerful set of binoculars. If with the latter the focus is not perfect then the image will be a blurry mess, the depth of field is so restricted that maladjustment by just a small amount makes the viewing hopeless. Plus, the binculars have to held totally stationary with respect to the viewed object, otherwise you'll get a headache from trying to keep track of what's there. But, get it right, and then the amount of detail from so far away you see is amazing, and you appreciate what a marvellous instrument that viewing aid is. That's how digital is, or at least has been for me over all these years ...

Frank
 
but with no convincing evidence that it is "clearly audible"

the 2006 3rd edition of Fastl, Zwicker Psycho-Acoustics book includes CD res demos - with no apologies or warnings for the "poor sample rate", 22.1 kHz ringing or other "digital artifacts" - does recommend headphones or very low reverberation listening room


the burden is to offer Positive results in Blinded testing - "rejecting the null hypothesis"

once we have a few positive results, accepted protocols then we can talk population statistics - does Stereophile's readership, "High End Audio" customers care if a 3rd sigma tail of prepubescent girls can tell a difference between 20 kHz and 40 kHz anti-alias/image reject filters


even the developers and their proven "highly resolving" skilled listening testers who train on the artifacts can't AB/X modern tuned psychoacoustic compression at 320k on the vast majority of musical releases - today it takes both training and selection of "killer samples" hitting the weak spots of the algorithms - algorithms that toss out 75% of the Shannon-Hartley Channel Capacity information of CD res audio
 
Last edited:
Frank, I'm not sure what you are attempting to say. Is it that CD is perfect, or is it that CD no more flawed than other mediums?
All mediums are flawed, and so is CD. Just by virtue of "only" using 16 bits, and also possibly by the sampling rate if one's hearing happens to be able to go above 22kHz. But in my experience it is less flawed than other sound carrying mediums. But, I am not saying that many of implementations of CD replay are not severely flawed in key, seriously audible areas, this is an entirely different thing!

Frank
 
None? No Gibbs type phenomena due to the sharp bandlimiting? No infinite negative or infinite positve ringing energy from the ideal sinc function based bandlimiting reconstruction filter? Really? You might want to have another look at those theories by Shannon and Nyquist. I think you'll find that the presence of such time-domain distortions are actually essential to perfect frequency-domain reconstruction. These are not group-delay or phase distortions. The time-domain distortions I'm refering to provide the missing frequency-domain signal content not directly captured by the original sampled points.

The sampling theorem itself features both positive and negative infinities as a signal requirement. That means perfect sampling and reconstruction assume a perfectly constant signal, constant to the degree that such signals should have existed since the big bang and will continue to exist until the end of the universe. Obviously, that is not possible, but what it does tell us is that while bandlimited sampling and reconstruction is theoretically perfect for signals which carry only frequency-domain sensitive information, it is not perfect for signal which also carry time-domain sensitive information. This, IMHO, has been THE flaw with audio CD from the start. The time-domain sensitive nature of music recieved little to no consideration when the format was defined.

So, CD gave us perfect sound forever, as far as spectrum analyzers are concerned, but not as far as the time-domain sensitive instrument such as the human hearing system is concerned.

The infinities that arise are in the bandlimiting (and reconstruction, which is just another bandlimiting anyway) and not in the A/D/A conversion itself. But I'm starting to sound like a lawyer, even to myself.

I've never said CD's have perfect sound (the forever part is close enough to true for me) and it's funny I find myself defending 'em. Probably because everybody else of similar mind has something better to do tonight. Lucky me.

Thanks,
Chris
 
but with no convincing evidence that it is "clearly audible"

the 2006 3rd edition of Fastl, Zwicker Psycho-Acoustics book includes CD res demos - with no apologies or warnings for the "poor sample rate", 22.1 kHz ringing or other "digital artifacts" - does recommend headphones or very low reverberation listening room


the burden is to offer Positive results in Blinded testing - "rejecting the null hypothesis"

once we have a few positive results, accepted protocols then we can talk population statistics - does Stereophile's readership, "High End Audio" customers care if a 3rd sigma tail of prepubescent girls can tell a difference between 20 kHz and 40 kHz anti-alias/image reject filters


even the developers and their proven "highly resolving" skilled listening testers who train on the artifacts can't AB/X modern tuned psychoacoustic compression at 320k on the vast majority of musical releases - today it takes both training and selection of "killer samples" hitting the weak spots of the algorithms - algorithms that toss out 75% of the Shannon-Hartley Channel Capacity information of CD res audio

Can we agree that dissatisfaction with CD has been significant (though not universal) and persistent among the audiophile community pretty much since it's introduction? If we can't agree on that much then we are destined to get nowhere debating the technical/perceptual basis why.
 
It would be a shame if talented and influential people like John Curl were to feel that the issues of CD playback were so inherent that no effort would ever improve things.

I changed my mind when I was first able to make a CD-R from vinyl that I couldn't tell from the original. Hardware got better and I got older and at some point the lines crossed.

Thanks,
Chris
 
If we can't agree on that much then we are destined to get nowhere debating the technical/perceptual basis why.

Especially because the reason why is unlikely to be technical/perceptual but rather sociological.

Off to see Ira Sullivan today. I didn't even know he was still alive! Last time I saw him must have been in the early '80s. What a nice surprise, and he's playing at a terrific venue (Jazz Showcase). I will enjoy this more than Brad enjoyed his "new music."
 
Yes, perfect sound forever, in the frequency-domain. Chris, now tell us about the time-domain distortion inherent to ideal bandlimited sampling and reconstruction.

The issues are discussed here by Dave Griesinger of Lexicon, he makes most of his work available as free download. In essence there is no evidence that BW filtering music at 22kHz is audible, in fact he shows several SACD's from Sting, Steely Dan that have in fact no content above 22k.

http://www.davidgriesinger.com/intermod.ppt
 
Member
Joined 2005
Paid Member
John - have you ever designed the analog stages ( IV converter / filter buffer) of a CD player or DAC. I found significant improvement using Erno Borbely's discrete complimentary jfet IV / filter over an IC based balanced design using 3 IC's per phase where OP 627, BUF 134, and other quailty op amps were substituted. The change in sound quality between quality op amps was small at best compared to going to the direct coupled discrete jfet design.

In the last version of the DAC Borbely did away with the seperate IV stage and filter buffers and used a single stage combining the IV with lower roll off rate filter at the output with a TO220 2SK216/2SJ79 output stage biased at 70 ma for use with the higher sampling rate DAC chips. Shunt regulators and Caddock or Vishay bulk foil resistors ice the cake. Yes i know the recorded signal went through a 100 op amps in the studio, yet I hear a real step up on just about everything played back, not just some showcase disks, to the point where it challanges an Oracle MK 6 /SME V / Dynavector XX turntable setup with discrete jfet teflon RIAA caps phono - the best 200 gram vinyl re-releases are just a little better in resolution, but not a lot and the best of of the HD downloads sound better than any vinyl release.
 
Last edited:
Status
Not open for further replies.