Piano World Home Page Forums Our Most Popular Forums Piano Forum Piano Recording: Need Your Ears!

Please excuse the imposition, as this isn't a piano question so much as a psychoacoustic or audio technology question.

I recently purchased an album of solo jazz piano recordings, which is the premier release from a new label called Octave Records, owned by high-end audio equipment maker PS Audio. In two of the twelve tracks, I notice some brief distortion in a number of spots, but I've discovered that not all listeners can hear it. It sounds like clipping to me, and creates a slight "crackle" sound, not very different from the static-like surface noise you sometimes hear from LPs.

So I thought it would be interesting to obtain feedback from our pianist community, which undoubtedly includes some of the most critical piano listeners. If you are interested in participating in this study, just have a listen (the sample tracks are only about 30 seconds, so it won't take much of your time) and post back here to let me know whether or not the distortion issue is audible to you.

Sample tracks are at https://www.psaudio.com/products/out-of-thin-air/ and you will find them if you scroll about half way down the page. It is best to listen with headphones at a reasonably loud volume.

The track with the most issues is the first one, Classical September. When I listen to it, I hear distortion at times 0:19, 0:23, 0:26 and 0:27.

The only other track with distortion I hear, is Self-Quarantine, at time 0:13. The other tracks all sound clean.

Thanks very much in advance, for your assistance with my study in psychoacoustics!

FWIW, I don't hear a problem.

+1

I do hear the distortion, but it's hard to say where it is coming from. It could just be in the conversion, since the original file is DSD and I'm sure the streaming samples are not a super high bit rate.

Marianne,
I listened using my Acoustic Research AR4xa speakers and could not hear any distortion at all.
Ian

I did hear multiple somethings on the right channel between 0:23 and 0:27, somethings like a click. This can happen when over-driving a digital recording past 0db. It's hard for me to imagine how the engineers involved with a "premier release" from a high-end maker would have let something like that go. There are plenty of tools to help with that sort of thing.

Beyond that, I am not at all a fan of the sound on the recording. The playing is wonderful, but the piano, itself, is not. Beyond that, I find the balance between what must have been mics inside the piano with those used to get some "room" into the recording too heavily biased towards the very close up sound. Even as a performer, I'm further back (at the keyboard) from where those mics were listening.

FWIW - for comparison's sake,here are two recordings by Matt Herskowitz. The piano is a great one; the pianist equally so, and, to my ears, the recording is "like being at the keyboard".

https://www.youtube.com/watch?v=BUpfagdPZJk

...and this one, in a style more similar to what OP has posted:
https://www.youtube.com/watch?v=WL7unbLLdck

Thanks to all responders, for your time and input. It does appear that there are significant differences from person to person, in the way we hear. So far, only 35-40% of listeners have reported that they can hear the distortion.

A special thank you to Andrew, for your detailed comments and suggested reference recordings. After listening to them, I decided to purchase Matt's "Mirror Image" CD.

On the distorted tracks, I have done quite a bit of technical analysis, starting with the original DSD files that I purchased. I tested my DSD->PCM converter by manually editing a test pulse into the .DSF file, one at 75% of full scale and another at 50%. With the converter's output attenuation set appropriately, it passes the test pulses without distortion, yet the music peaks are still limited to a much lower level (3.5dB below the 50% test pulse).

I do not know if the single-rate DSD files I received are bit-for-bit the same as the master tracks, but if they are, there may be no recourse. On the other hand, if they were down-sampled from the masters, perhaps the masters are clean and the distribution DSD files could be redone.

There are two noteworthy features of the tracks that have distortion: Their peak level only reaches -3dB (Classical September track) or -2dB (Self-Quarantine track), while the other tracks exceed -1dB and none of them have distortion. Secondly, looking at a long section on a scope shows that all of the peaks reach almost exactly the same voltage, which is statistically impossible for piano recordings. Peak values should have a random distribution. It certainly appears that something in the signal chain was limiting the peaks.

I have submitted my analysis to the PS Audio CEO and he forwarded it to the mastering engineer, but I have received no replies. I have to wonder if both of them are in the 60-65% population that can't hear the distortion.

Psychoacoustics is a b**ch.

You shouldn't automatically assume that the recording is faulty and other people have bad ears, when there are some elements in the chain that cannot be reproduced by others. Your amplifier may not be linear, the membrane of your listening device my create misrepresentations and distortions. Your own ears may not do what they are supposed to do at certain frequencies in certain dynamic ranges and the assumption that your hearing is flawless and that 60-65% of the population have bad hearing is a far stretch at best.

A Fourier analysis of the original recording before any mix down (and you won't get that from the producers anyway) may lead you to possible clues of what element in the recording chain may not be flawless and if there isn't any, you'll need to start tracking down possible flaws in the reproduction chain on your side.

I agree with others who say that this is a badly prepared and recorded piano, which makes the whole discussion pretty pointless anyway.

I'm not automatically assuming anything. Enough testing has been performed, that I know the issue isn't the reproduction chain. I didn't say anyone had "bad ears" or that my hearing is flawless, only that many people do not notice the distortion. It is quite possible that they can hear other things, which I cannot. I consider it quite unfair on your part, that you are accusing me of making such assumptions.

I do not believe there exists a multi-track pre-mix recording. As I understand the process, they only record the mixed stereo channels, i.e., it's direct to master, then the DSD files delivered to customers are exact copies of the masters. I was hoping that there were more generations involved, so that a better quality master might exist, but that is probably not the case.

I have been experimenting with declipping software, and the result is quite good, although not perfect. That will probably be the final solution, as I'm not expecting any help from the producer at this point.

That's what it says in their description and that's what we can see in the photo for the surround sound microphone setup.

As to the rest: You're right, I'm wrong.

It's interesting that the description mentions the original recording/mixing was done by Robert Friedrich at Moose Sound, while the editing/mastering was by Gus Skinas, possibly at a different location.

Explicit critical details are still missing. At one extreme, the mixing could all have been done in the analog domain upstream of the recorder, thus only requiring a stereo master recording with no further need for editing. On the other hand, the Sonoma system is certainly multi-track and capable of cross fades in post-processing, so each mic channel could have been recorded on a dedicated track as the first step. We don't know what Gus worked with as his source material, only that it was his responsibility to produce the PCM versions of the tracks.

I did find information on the Sonoma system which indicates it is strictly single-rate DSD at 2.8442Mbit/sec per channel, so no DSD down-conversion was involved.

It's frustrating that my emails have gone unanswered. It would certainly be nice if they could re-do the mixing and mastering to avoid the limiter action that occurred somewhere along the way. Considering that the album is advertised as an "ultimate reference" I believe they owe that to their customers.

That's pretty convincing, to me. I was thinking that loading one of the files into Audacity, and expanding the time scale, would reveal any flat-topping. Using an scope is "old technology", but it does the job.

Whether that glitch is _audible_ -- that's a different question.

Thanks --

Peaks are not clipped to flat tops in most cases. Rather, they appear to have been fast-attack limited.

Scopes are "old technology?" When was the last time you saw an oscilloscope? Today's DSO's are built with similar processing power to PC's, perform all kinds of mathematical analysis including FFT, and collect data orders of magnitude faster.


The point of the thread was for you to have a listen to the samples, then answer that question.

Editing and dithering, I know, can have unintended secondary characteristics show up in the final render. The issues don't happen on every playback system.

I don't think it's some listeners that hear it and others don't but some systems can play it fine and other systems exhibit a problem.

At each of the time marks you noted, the piano had a very robust sonority.

Still, sounded fine to me.

Marianne
You say the purpose was to listen to the thread and then answer the question.
I remember you bought a lovely GX Kawai piano. There were a few issues with it at one time which some suggested may have been your perception of the sound.

Is this interest in psychoaccoustics in some way related to what you perceive in the sound of your new piano ?
Considering that a number of us may remember your thread about your piano ,you may not get much objectivity. I am not saying people are untruthful at all !
They just may perceive things differently based on past experiences.
Memory is another aspect to this test and perhaps taste.

I'm late to the party but I can't hear it. I listened to it with headphones several times.

I created an A/B comparison track which plays repeated samples from two time ranges of the problematic Track 1. It compares a declipped version (A) to the original version (B).

The comparison can be downloaded from
https://cp.sync.com/dl/2da518a50/m3u6skb3-2y6b7hye-4kp3grqw-c6z2grg6

The first 30 seconds consists of five A/B comparisons of samples taken from time range 0:21-0:23. The distortion is audible in the right channel on the B samples.

The last 30 seconds consists of five A/B comparisons of samples taken from time range 0:25-0:27. The distortion is audible in the left channel, in two places.

If you were unable to hear the distortion while playing the original tracks, the comparison track will help isolate the time points for you, as well as providing a clean reference. I am interested to know if having the comparison makes the distortion easier to hear.

I would also like feedback on the quality of the declipping that was used to create the "A" reference samples.

Again, thank you in advance for your participation and feedback.

I don't recall reporting any issues with the GX-6, and I've been completely satisfied with it. At the next tuning, there are a few hammers which I will have voiced to even out the middle octave, but I don't consider that a major problem.

Now if it was an SK-6, I might be complaining about the voicing unevenness, but with the GX line I don't expect perfection and am willing to put some of the money saved into voicing.

I feel like Sargent Schultz from Hogan's Heroes, I hear nothing. I listened to the original recordings and your extracts with and without declipping. Sorry, I'm an old guy (66) who played in bands and has a touch of tinnitus. I like the rich chords that seem to be problematic, but they sound fine with my headphones. In an earlier post you mentioned the levels, -3 dB shouldn't cause any clipping. As one who does some recording and mixing your request for unlimited files will never happen. That's how studios avoid clipping in the first place. I can appreciate Golden Ears, I used to have a pair, but in a way life is simpler now, because I can concentrate on basics like quality of the music and arrangement.

I do hear something at the times you mention, but tbh to me they just sound like 'slightly' dissonant chords on the first piece - a bit unpleasant, but 'resolved' afterwards so not a problem. The second is similar, but actually more irritating to my ears at 0.11.