[gregorio]

[1] Yours is a common opinion but I think manifestly false ...
[2] ADC-DAC systems measure very well on some common and traditional parameters such as THD, noise and audio band response, but they don't always sound too good. The brain-ear works differently to most test gear in my opinion.
[3] A true brickwall filter gives you sinc (x) output in the time domain with pre and post ringing of quite high amplitude at around the cutoff frequency. This is a hard fact.
[4] Certainly on THD and FR any CD are higher res digital system will outperform any headphone or speaker.

1. Then manifest it with some reliable evidence please.
2. ADC/DACs measure very well as far as all the properties of sound are concerned. The brain-ear is not concerned only with the properties of sound, it introduces many variables which have nothing to do with sound, which is why we can hear significant differences even when there are none in the actual sound whatsoever.
3. No, the amplitude of the ringing is quite low and that's even using a test signal designed to exacerbate it, a test signal which cannot even exist in the real world! So that would be an audiophile "hard fact" rather than a real hard fact would it?
4. Yes, as far as any actual property of sound is concerned a CD (ADC/DAC) will not just outperform but massively outperform any headphone or speaker. Hence why I disputed your assertion that the relatively tiny imperfections of the ADC/DAC process play a "big part" in SQ like headphones or speakers!

G

 

[SimonPac]

pinnahertz I have heard people say 44/48-16 is transparent and even hi-fi magazines said so for a while in the '80's. I spoke to a musician many years ago who used to work in the rock-pop industry in the UK. He said that his newly acquired DAT machine was clearly changing the sound of percussion. I have done my own experiments with a 192-24 file of 461 Ocean Boulevard by Eric Clapton re-sampled down. Not even a brilliant recording. I have listened to a 15ips 1/2 inch tape fed through a studio ADC at 192-24 and back through a Chord top end DAC into HD800s. No way does the digitally routed signal sound the same. I agree that an analogue tape machine is unlikely to have a tidy square wave or impulse response, but the distortions are linear in the sense of their transfer functions before clip could be modeled by linear mathematical parameters. Not so with digital.

A perfect brickwall filter would show the Gibbs phenomenon on a square wave; something with no parallel in the natural 'analogue' world.

Digital systems use FIR filters which are inherently non-linear at the end of the day. Even their implementation of brickwall is an approximation to the mathematical model because of finite filter tap number.

If they are transparent to you, fine. Technically perfect they are not. Transparent, for me, they are not.

I spoke to a designer at a well known top end manufacturer who conceded that they add short term phase dither to their gear because it 'sounds more natural'.

 

[castleofargh]

I don't think I'm confused, subjectively or technically. I used to design parts of PCM based systems for telecoms. Yours is a common opinion but I think manifestly false, and widely (though not uniformly) held as such within the industry. ADC-DAC systems measure very well on some common and traditional parameters such as THD, noise and audio band response, but they don't always sound too good. The brain-ear works differently to most test gear in my opinion.

A true brickwall filter gives you sinc (x) output in the time domain with pre and post ringing of quite high amplitude at around the cutoff frequency. This is a hard fact. Research FFT windows and conjugate variables, or even QM conjugate variables for an roughly analogous situation.

Certainly on THD and FR any CD are higher res digital system will outperform any headphone or speaker.

important part in bold. ^_^ what is it to us when it's a frequency that's above our hearing range?

the link you gave was typical(as in misleading marketing). the graph with the impulses made me laugh at first TBH. the analog portion of the graph should be labelled "impulse as capture in its natural analog habitat in a parallel universe". that type of demonstration should be illegal IMO. hey let's "demonstrate" in a surreal graph, how a perfect impulse isn't exactly reconstructed on a system aimed at reconstructing the audible range. what you failed? awwweee that's too bad. in our next episode, we'll show how 16/44 isn't good enough to reconstruct radio waves and infrared.

I have nothing against increased resolution, but that kind of BS justification is disgusting.

 

[SimonPac]

3. No, the amplitude of the ringing is quite low and that's even using a test signal designed to exacerbate it, a test signal which cannot even exist in the real world! So that would be an audiophile "hard fact" rather than a real hard fact would it?

important part in bold. ^_^ what is it to us when it's a frequency that's above our hearing range?

A digital signal is treated by the filter digital electronics as a superposition of scaled impulse responses of sample rate width. Each impulse is transformed by the brickwall filter. An ideal (dirac) pulse comes out like this:

The point is that the superposition of impulses in the general case results in nonlinear temporal smear. With an analogue system, you get temporal smear but it's linear.

 

[SimonPac]

castleofargh I agree that you'd never get those impulse responses out of a real world analogue audio tape machine; not with a 3uS pulse and with the timebase the same throughout the graphs. They avoid declaring what sort of analogue it's supposed to be..... you certainly can easily design analogue electronics that good. They are PA people I believe.

 

[71 dB]

As I'm sure 71dB knows, two back-to-back zeners conduct in both directions when you exceed a certain voltage. A small zener such as a BZY88 acts like a normal diode, i.e. it conducts in the forward direction with an exponential curve having a 'knee point' of around 600mV. Below that it passes very little current. In the reverse direction the zener is much like a normal diode, tiny leakage current only, until it reaches its breakdown voltage, where it suddenly conducts strongly. The breakdown voltage is the prime factor in the zener specification and can be from about 3V up to 100V plus. When you have two back-to-back, as in the Beyer photo, you get very little conduction until you hit about 600mV+zener breakdown voltage. Same in both directions so works as a symmetric clipper for AC.

Yeah SimonPac, this is exactly what I mean.

Agreed. I agree with you on everything you have said about electrical equivalent damping and LCR elements. I also trust you have made good agreement in deducing equivalent lumped elements though I haven't tried to regenerate the curves from your parameters.

Well thanks! This stuff isn't "rocket science" to someone with my education, but that doesn't mean everything is 100 % correct, so please bring it up if something I say looks funky.

I agree that higher maximum rate of change of driver displacement at a given frequency means louder sound on a sine wave. However I'm not convinced that a maximum d/dt on displacement has no effect on music just because a 20KHz sine wave can be generated at adequate level. In a similar way I'm not fully convinced by 22KHz brickwall filters for audio either, having heard analogue master tapes fed through various digital codecs. I recently heard the LCD4's, and they have amazing delicacy and transparency on acoustic material, though I don't know for sure why. But light moving assembly was a prime design goal. Listening it is easy to hear the progression in resolution from 650 to LCD3 to LCD4. I still enjoy my 650's and 518's though.

It's a bit difficult to follow the logic in this one. You go from velocity to brickwall filters to the resolution of LCD4.

I agree with the technical analysis performed by 71dB concerning how headphones work, at least on the specifics discussed so far. I assume he is a good engineer or physicist.

I have an university degree on electric engineering major being acoustics and signal prosessing. That doesn't mean everything I write is correct. There is a lot to learn.

I am thinking about the implications of the LF over-damping issue. The LCD4s I mentioned I assume are mechanically highly over-damped and have a very light membrane and strong magnetic field. For me, if I'm right and judging by the end result, that is a good way to go.

The impedance curve for LCD4 is a straight line just under 160 Ω indicating VERY strong damping.

All these things are a compromise, as 71dB says. Some work better than others. Having said that, in the case of the HD650s and indeed '518s , I prefer them with some series R. I'm not totally sure why, which is why I was interested in the damping issues. I had not done the work there that 71dB has. incidentally he prefers low z out and overdamping on dynamic phones. On that, with my phones, recordings and system, I disagree, at least subjectively.

The target should be maximal precision of driver movements, and techically that is achieved with critical damping. However, I believe due to the properties of human hearing, overdamped system sounds as accurate or even more accurate than critical damped system. This target assumes that the recording is good, that it is worth precision. If series resistors make bad recordings sound better, that's fine. You are correcting the errors of production. However, tinkering with sound poses the danger of preference biases. That's why it might be a good idea to listen to good recordings without series resistors.

I haven't studied series R -thing with headphones much. I have a general idea of how it changes the sound.

At the risk of drifting off thread, digital audio is a complex business largely because brickwall filters do not perform well in the time domain, as has been realized. An acquaintance of mine in professional recording told me 25 years ago that the first DAT machines were not transparent at all in his opinion. Opinions vary, but I agree with him. Even 192/24 is not transparent on a setup I heard. The corollary of brickwall in the frequency domain is temporal smear, although this should be less of a problem with higher sample rates.. Meridian are attempting to rectify this with MQA. There's also an interesting graph of impulse response on this page: https://www.plusmusic-us.com/index.php/technology/sla/dynamic-vs-sampling

In my opinion that link is anti-digital propaganda. They give the impression (twisting facts a bit) that digital audio has bad problems and then they tell how they have "solved" these problems in their products. If analog was that superior, why would anyone use digital audio for anything? Digital impulse responses look bad to our eyes, but do they sound bad to our ears? People who don't know how hearing works tend to think what looks bad also sounds bad. A classic marketing trick to demonstrate how your product is "better."

The brickwall filter problem is miniscule to say the least. It has been addressed in many ways:

1. Phase shift to move pre-ringing to longer after-ringing.
2. Rounder filter to reduce ringing.
3. Oversampling + more relaxed filter.

I think the total system impulse response is of very high relevance to sound quality. The ADC/DAC system including the filtering and any rate change interpolation is a big part of that, as obviously is the headphone or speaker.

Yes, the total impulse response it what counts. If the chain is linear, it doesn't even matter in what order the changes happen. If the impulse responses of your headphone amp and headphone swapped places, the sonic result would be the same (assuming they are linear enough).

 

[SimonPac]

71dB I respect your opinion and technical education. I have a university degree in Applied Physics/Electronics/Maths. LCR circuits are not 'rocket science'; S parameters will do. IMO digital has issues and so does analogue. I personally prefer analogue if it is high quality, say fast reel-to-reel. I don't say this because I like the known distortions, I say it because I think as far as the human consciousness is concerned, it captures more information. Exactly why, I don't know.

 

[gregorio]

An ideal (dirac) pulse comes out like this ...

How can you have an "ideal" pulse (or anything) if can't actually exist in the real world? What would an ideal unicorn or an ideal flying pig look like?

What you've got is pre/post ringing of a relatively small amplitude, at an inaudible frequency, on a signal which cannot exist ... and that's your "hard fact" is it? That's the best you can do when asked for "reliable evidence"? Surely you've got to have something better than that to demonstrate why the common opinion is manifestly false? The only thing which is "manifest" to me is that digital audio is not ideal at recording and reproducing signals which can't exist, although what that's got to do with commercial audio content (and recording/reproducing it) I can't quite fathom.

I don't think I'm confused, subjectively or technically. I used to design parts of PCM based systems for telecoms.

Hmm. If you worked on something which gave you an insight into how modern digital audio actually works, how come you fell for that nonsense marketing BS article enough to actually post it on the sound SCIENCE forum?

G

 

[gregorio]

[1] IMO digital has issues ...
[2] I personally prefer analogue if it is high quality, say fast reel-to-reel. I don't say this because I like the known distortions, I say it because I think as far as the human consciousness is concerned, it captures more information.

1. Please detail any issues digital audio has within the audible band.

2. Neither reel-to-reel analogue recording nor digital audio are in the least bit concerned about your human consciousness. All they're concerned about is audio signal fidelity, accurately storing/reproducing actual sound waves. Digital audio does this with greater fidelity than reel-to-reel, it captures/preserves more information. If you say you prefer reel-to-reel then it can ONLY be because of reel-to-reel's distortions and/or some bias in your human consciousness unrelated to sound fidelity.

G

 

[71 dB]

71dB I respect your opinion and technical education. I have a university degree in Applied Physics/Electronics/Maths. LCR circuits are not 'rocket science'; S parameters will do. IMO digital has issues and so does analogue. I personally prefer analogue if it is high quality, say fast reel-to-reel. I don't say this because I like the known distortions, I say it because I think as far as the human consciousness is concerned, it captures more information. Exactly why, I don't know.

Well, I have never heard fast reel-to-reel. All I can say is that CDs when produced well have NOTHING I would complain about. My CD-player allows selection of reconstruction filter and with some material the filter seems to change the sound, but it is very very subtle. The width of the soundstage changes, but cross-feed makes 1000 times bigger changes to the soundstage. For me it's the same information presented a little different way:

Information
Information​

Different fonts, but the information content stays the same

Also, at this age my hearing is long gone above 16 kHz, so why would I worry about some ringing around 20 kHz? When I was a child and heard 20 kHz, I wasn't into audio so I missed the opportunity…

Are you sure the greatness of fast reel-to-reel isn't just placebo effect? Seeing those reels spinning fast must have some effects…

Also, fast reel-to-reel feels ridiculously expensive and cumbersome solution for my music collection… …in studio they can use digital or analog, I don't care as long as it sounds great.

 

[71 dB]

Hmm. If you worked on something which gave you an insight into how modern digital audio actually works, how come you fell for that nonsense marketing BS article enough to actually post it on the sound SCIENCE forum?

G

That's what surprised me too. I started to read the text without checking (my mistake) whose pages I am on, thinking it's some audio guru telling something I can learn from. It didn't take many seconds to have my What-moment and realising I am reading some marketing BS by an audio company

 

[bigshot]

Analogue tape does improve the sound of recordings with limited upper frequency response. A digital recording that sounds slightly muffled because a high end roll off has been applied might sound fine on tape, because the tape hiss fools the ear into thinking it's hearing frequencies that aren't present in the recording. When I was restoring and mastering 78s for CD release, I'd sometimes take a recording of high frequency hiss and lay it in at a low level to take the curse off of the fact that there are no frequencies much above 8kHz on 78s.

 

[pinnahertz]

pinnahertz I have heard people say 44/48-16 is transparent and even hi-fi magazines said so for a while in the '80's. I spoke to a musician many years ago who used to work in the rock-pop industry in the UK. He said that his newly acquired DAT machine was clearly changing the sound of percussion. I have done my own experiments with a 192-24 file of 461 Ocean Boulevard by Eric Clapton re-sampled down. Not even a brilliant recording. I have listened to a 15ips 1/2 inch tape fed through a studio ADC at 192-24 and back through a Chord top end DAC into HD800s. No way does the digitally routed signal sound the same.

Ah, then you've never actually tested your theory, then. Do I assume correctly that each of your "comparisons" were fully sighted (not blind, double-blind)? You've fallen victim to your own expectations. It's a case of knowing "too much" about what's "wrong" with digital forming a massive anti-digital bias. Let me know when you've removed the bias and done an actual test.

I agree that an analogue tape machine is unlikely to have a tidy square wave or impulse response, but the distortions are linear in the sense of their transfer functions before clip could be modeled by linear mathematical parameters. Not so with digital.

No, no, no. You've clearly never had much hands-on with a tape recorder. The distortions are not linear, in the sense that phase varies with frequency, similarly to digital but not to the same rate of change, distortion is a function of frequency AND level, AND tape formulation AND tape speed, AND choices made during alignment (the HF response vs distortion trade-off choice of bias level), AND includes very significant IMD, AND incidental frequency modulation products, AND non-stable interchannel delay, not to mention the noise floor. Digital has none of that. None.

A perfect brickwall filter would show the Gibbs phenomenon on a square wave; something with no parallel in the natural 'analogue' world.

Not true. Anything that band-limits frequency response will show the phenomenon to different extents. Analog tape is rather significantly frequency limited by a combination of linear speed (magnetic wavelength) head gap (again, wavelength), bias and EQ. The lumped result is a multi-pole response, with the HF roll-off as a result of the inability of HF EQ to fully compensate for bias HF self-erasure being a rather steep curve. Add gap loss, you've got a multi-pole roll-off monster. People think analog tape is so "pure" and free of time-domain issues, because it's darn hard to measure with all that variable time delay in there. But it's hardly free of time-domain issues, they're there in spades, mangled up with all the other issues.

If they are transparent to you, fine. Technically perfect they are not. Transparent, for me, they are not.

I recognize that some people prefer the distortions of tape. However, ABX tests of undigitized live analog signals vs digitally coded and decoded version of the same signal reveal what is truly transparent. You can repeatably fool someone into thinking the digital process is live, but tape will never...I mean Never...accomplish that. It tips its hand at every corner.

I spoke to a designer at a well known top end manufacturer who conceded that they add short term phase dither to their gear because it 'sounds more natural'.

How can something sound "more natural" by introducing a distortion not found in nature?

Since you seem to like anecdotal "evidence"....many years ago again in the infancy of digital audio I entertained an audiophile club at our studio, tour, etc. One member was staunchly opposed to digital recording. Part of the evening was listening to some music. I put on a CD and told him...he hated it. I put on an analog tape master, told him...he loved it. And so it went for a few other selections, until I put on a digital recording we'd made of an orchestra. I lied and told him it was analog tape. He was emphatic! He said, "See! That's what I'm talking about! That analog tape sounds so incredibly wonderful...."(I might have the quote mangled, but you get the idea). I let him enjoy his "analog" experience for a bit, then called his attention to the control room window, stopped the digital tape and held up the video tape it was recorded on. His constituents actually applauded. My point to him was, the sound he heard had nothing to do with the medium, it was the mic positions, hall, and mix....and his expectation bias.

 

[SimonPac]

Hmm. If you worked on something which gave you an insight into how modern digital audio actually works, how come you fell for that nonsense marketing BS article enough to actually post it on the sound SCIENCE forum?

That's what surprised me too. I started to read the text without checking (my mistake) whose pages I am on, thinking it's some audio guru telling something I can learn from. It didn't take many seconds to have my What-moment and realising I am reading some marketing BS by an audio company

I'll stick to the technical points. If people here don't want to talk subjective, I'll respect that. You have my subjective opinions.

The plusmusic website I didn't read, I just found those impulse responses following a google search. The analogue ones are entirely achievable with electronics. Not tape. They are not talking tape. The sinc x plot I inserted is from wikipedia. These things don't go out of date any more than fourier transforms do. they're maths. The sync x plots at the plusmusic site still represent most PCM audio and therefore lossless compressions thereof.

The sinc x function represents the single impulse response of the filter. Digital audio uses the superpostion of time shifted impulses of differing amplitude. That is what PCM is. That hasn't changed. PCM is still PCM. There is a spilling over of each sample in the time domain in both directions with a linear phase, linear frequency brickwall LPF filter. It is only limited in terms of duration of time smear by the FIR implementation. What do you disagree with there? The sinc function may in isolation ring at an inaudible 21KHz but it is superposed with adjacent filtered samples. The ringing is not that low in amplitude, the scale is linear. Please correct my statements as necessary....

Yes you can oversample to reduce the smear. You can change the filter implementation. But don't think you are getting exactly the original analogue signal back. There is the issue of how ADC was done. To tackle this, apodising filters which turn over significantly before 22KHz have been used to remove most pre- and post- ringing, even that embedded in the recorded data due to the ADC nyquist filter. And filter algorithms have been derived which totally eliminate pre-ringing but they are not linear phase and have a fair amount of out of band noise.

Not true. Anything that band-limits frequency response will show the phenomenon to different extents. Analog tape is rather significantly frequency limited by a combination of linear speed (magnetic wavelength) head gap (again, wavelength), bias and EQ. The lumped result is a multi-pole response, with the HF roll-off as a result of the inability of HF EQ to fully compensate for bias HF self-erasure being a rather steep curve. Add gap loss, you've got a multi-pole roll-off monster. People think analog tape is so "pure" and free of time-domain issues, because it's darn hard to measure with all that variable time delay in there. But it's hardly free of time-domain issues, they're there in spades, mangled up with all the other issues.

A system with poles and zeros is a linear system in the mathematical sense. Google 'Linear system'. A digital system implemented as FIR is never really linear, just a working approximation. I agree that a perfect brickwall would exhibit Gibbs and pre-ringing if implemented in analogue, but it would also have infinite time delay......

 

[SimonPac]

Pinnahertz I changed my mind, I'll talk subjective for a minute. I would not claim to always be able to tell digital from analogue on a short burst especially on a system I don't know or with music I don't know. And if it was recorded and mixed in digital it will probably sound cleaner, agreed. Digital does some things better. But over a period of time, with material I know well, I would say analogue masters are better played with good analogue gear, and stuff that has been digitized works better at higher sample rate and/or bit depth.

How can something sound "more natural" by introducing a distortion not found in nature?

I can ask the gentleman in question and if he wants to be quoted if you wish. I have been in email contact. They are smart guys. I think it is about camouflaging systemic behaviour signatures. I prefer my PCM63 based DAC with some dither on the 20-bit input.

I assume you guys think all DACs sound the same?