High-res audio comparison: Linn Records Free High Res Samples

[Newman]

The one take-home message from these videos seems to be that most recording studios haven't bothered shielding all their EM sources and have loads of stray radiation happening outside (or in some cases inside) the audible range...

Why taint all studios based on a sample of 1 or 2? It’s like you are defending Linn by saying everyone else is no better. I don’t see a good reason to do that?

 

[DSJR]

Actually, it was Linn/Naim, because Linn didn't yet make their own amps. ;-)

Did that in a big Linn/Naim system with a Sony 1610 pro processor system of all things and still couldn't tell if this thing set to A-D-A in the tape loop made any difference compared to the straight through feed - for what it's worth...

 

[Hector Achilleus]

"information in the gaps between the samples."

That's really incredibile, that a (reputable?) hw audio company spread that kind of BS in 2021.

 

[charleski]

Why taint all studios based on a sample of 1 or 2? It’s like you are defending Linn by saying everyone else is no better. I don’t see a good reason to do that?

Amir's posted six videos in this series so far, each looking at releases from a different label. Five of them showed significant high-frequency spuriae (the Naxos one focused on the DSD noise-shaping hump). Linn is no better than anyone else in this regard, and a bit worse when you look at the last track in this video.

The difference is that Linn's catalogue contains some top-notch performances from internationally-recognised professional artists (though you can certainly find some drek in there as well). So it depends on whether you think the music matters or not.

"My God! What has sound got to do with music! ... That music must be heard is not essential--what it sounds like may not be what it is"
-- Charles Ives (Essays 84)

 

[Serg]

So to conclude, is it the best way is to perform high resolution capture only during initial mastering from original analog sources and then down sample to CD for distribution purposes?
In my theory, this could ensure that maximum samples are captured in audible (i.e. the most important) band and any ultrasonic noise is successfully cut.

Agreed! Makes sense.

 

[Frgirard]

Music recording or produced how?
With mic and witch bandwidth.
Direct to the console for electronics and electric instruments.
With samples, vst and software like protool

With the audible 22 kHz max When Moogli or Pocahontas put their teeth in.

 

[Beershaun]

I appreciate all the information and discussion on samplong rate and bandwidth to help bring context to file evaluation. Thanks all for sharing your information and experiences.

I have another question in sample rate vs change over time. My understanding is the music information is the sum of amplitudes and frequencies constantly changing over time. If this is the case, is there a minimum sample rate required to ensure no change in frequency or amplitude is missed or lost due to quantization error?

 

[Jimshoe]

My takeaway from these videos is that 'high resolution' audio is complete BS and that redbook CD was right all along (i.e. all we need).

If we want anything its properly mastered, excitingly dynamic, 16/44 music. Stop compressing the hell out of everything and let do digital what it's capable of.

Sadly, it's never gonna happen in a world of marketing BS and ignorance.

Sigh

 

[SIY]

I appreciate all the information and discussion on samplong rate and bandwidth to help bring context to file evaluation. Thanks all for sharing your information and experiences.

I have another question in sample rate vs change over time. My understanding is the music information is the sum of amplitudes and frequencies constantly changing over time. If this is the case, is there a minimum sample rate required to ensure no change in frequency or amplitude is missed or lost due to quantization error?

>2 times the highest frequency in the recorded bandwidth. Nothing special about your case, it's still Shannon-Nyquist (assuming by "quantization" you mean "time quantization").

 

[AudioSceptic]

Ok, you do say that doubling of sampling rate is enough to capture "sufficient" data at a given analog signal frequency.
I know that 1 Hz has nothing to do with music but lets use it as an example to simplify mathematically and visually. So we have 1 full pure sine wave with 1 Hz, and we are sampling it 2 times per second. This means we're getting maximum 2 peak points of the signal which is very far from from representing an original waveform. Thus if we'd used say 20 Hz samplings rate, it would give us more clear shape of a sine wave. And the same should apply to the higher frequencies... Am I wrong here?
Or maybe all these does not matter because we finally reach the speakers stage for reproducing the sound waves, which by itself have mechanical limitations (precise cone movement over the voice coil) and thus all shortcuts done in digital processing are not noticeable??
Sorry for my English

You really need to watch Monty's videos, or watch again if you have already.

 

[AudioSceptic]

Nippon Columbia's machine was 1971...

And the BBC started using PCM to send audio to FM transmitters in 1972. <https://www.bbceng.info/Technical Reviews/pcm-nicam/digits-fm.html>

 

[sarumbear]

At 3 metres, at 40kHz the air alone has pulled about 4dB out, and at 100kHz this rises to closer to 10dB. Which is simply domestic listening distances. Attend live music and you are looking at way higher attenuations. At say 10 metres, which is pretty close for many venues, 40kHz is about 12dB down. 100kHz is of the order of 40dB down.

As an acoustics engineer I can verify that is very much the case but for the life of me I can’t recall a reference source to the actual measurements. May I ask your source, please.

 

[sarumbear]

Head geometry might make a slight difference to the precise frequency response of the recovered audio, but that is about it.

You may be surprised to find that head geometry and alignment will make huge difference to the playback off a tape player.

Yeah, bias frequencies were all over the place. Lots of empirical tweaking.[/QUOTE]

AFAIK the bias tone is written by the erase head and not by the record head. The record head may also serve as playback head, but I could think that the erase head is optimized for the high bias tone frequency (smaller gap) which may not be the case for the playback head.

Also erase head and record head are a few cm apart. Does this distance play a role when using the bias tone for corrections? At least for fast flutter I can imagine that this is not identical for erase and record head.

Bias was simply used for pre-magnetising the tape and reduce noise, mainly hiss. It has nothing to do with playback. On most “modern” recorders like Studer A80 bias was mixed with the audio, hence the record head was used.

Unless we are talking about a pre-60s tape recorder using the bias on the playback for anything is wrong. The capstan motor was quartz Crystal controlled whereas bias was a free running sine generator.

What is missing in this conversation is the Dolby NR units, which was used on almost every analogue recording. Their (miss)calibration will alter the frequency response much, much more than anything mentioned so far.

 

[sarumbear]

These people are engineers. They have to understand how sampling works, how digital recording works, how D/A conversion works. It’s really shameful that they are propagating falsehoods to a public that largely doesn’t understand the physics of sound.

Their founder was a mechanical engineer who by education not expected to know either of your listed knowledge.

 

[sarumbear]

Nippon Columbia's machine was 1971...

I was recording on EMI machines!

 

[sarumbear]

Why taint all studios based on a sample of 1 or 2? It’s like you are defending Linn by saying everyone else is no better. I don’t see a good reason to do that?

Because that’s correct. As he said “most” never tested anything above 20kHz. For starters equipment was not available to do the tests!

 

[Beershaun]

>2 times the highest frequency in the recorded bandwidth. Nothing special about your case, it's still Shannon-Nyquist (assuming by "quantization" you mean "time quantization").

After rewatching Monty's video after our conversation in this thread, the explaination that helped me understand better, was: that there is only one possible mathematical solution to represent the 20khz waveform with certainty in a 44.1 khz bandlimited signal with only two samples. So I think I was missing the point about the band limiting creating a finite and mathematically definable number of possible solutions. Since we know every one of them, and band limiting means that each sample value combination can only coorispond to one output solution, we can look up the correct value for any sample combination to create the unique output that coorispond to the number and value of the samples representing it.

Did I get that right?

 

[SIY]

After rewatching Monty's video after our conversation in this thread, the explaination that helped me understand better, was: that there is only one possible mathematical solution to represent the 20khz waveform with certainty in a 44.1 khz bandlimited signal with only two samples. So I think I was missing the point about the band limiting creating a finite and mathematically definable number of possible solutions. Since we know every one of them, and band limiting means that each sample value combination can only coorispond to one output solution, we can look up the correct value for any sample combination to create the unique output that coorispond to the number and value of the samples representing it.

Did I get that right?

Pretty much, yes. Let me throw this one in: if the change in a signal that you proposed is rapid enough that the sampling and reconstruction "misses" it, then what was missed was frequency components that are removed in the bandlimiting process.

The key point is what you now understand: the sampling and reconstruction completely captures the waveform. It's not a matter of tricks or approximations or shortcuts, it is mathematically precise. Which, to me at least, is amazing and cool.

 

[Francis Vaughan]

You may be surprised to find that head geometry and alignment will make huge difference to the playback off a tape player.

Yup. I didn’t mean alignment, but the wording wasn’t the best. I was thinking more about the exact gap and pole geometry and how the final field as it affects the tape would subtly affect the response. Thus changes in head could make small changes.
As it is it seems that the Plangent Process uses a custom head anyway to recover the bias signal.

 

[Newman]

The key point is what you now understand: the sampling and reconstruction completely captures the waveform. It's not a matter of tricks or approximations or shortcuts, it is mathematically precise. Which, to me at least, is amazing and cool.

Yes, and you can understand why non-mathematical people see it as one of those 'free lunch' claims that can't possibly be true. And why same people find the idea, that digital 'steps' mean there is missing music between samples, to be perfectly logical.

What's really sad is that the mainstream audio/hifi media could have been, since 1984, leaders and standard-bearers for the truth -- the illogical, hard-to-credit truth that needs to be repeated -- about digital audio accuracy. Instead, they chose the exact opposite path. Chances are, if you have come across myths about digital 'steps', the sound of jitter etc, you probably read it first in the mainstream audio media. Quite likely the journalist writing about their interview with some guru who makes super-pricey gear, not just turntables etc but also very pricey DACs and purist direct-signal-path gear.

To their eternal shame IMO.

the sampling and reconstruction completely captures the waveform

I like to say it this way: the sampling and reconstruction completely captures the waveform that lies below half the sampling frequency and above the noise floor set by bit depth.