Journal Forum

Perceptual Effects of Dynamic Range Compression in Popular Music Recordings - January 2014
4 comments

Accurate Calculation of Radiation and Diffraction from Loudspeaker Enclosures at Low Frequency - June 2013
9 comments

New Measurement Techniques for Portable Listening Devices: Technical Report - October 2013
1 comment

Access Journal Forum

AES Journal Forum

Audibility of a CD-Standard A/DA/A Loop Inserted into High-Resolution Audio Playback

(Subscribe to this discussion)

[Engineering Report] Claims both published and anecdotal are regularly made for audibly superior sound quality for two-channel audio encoded with longer word lengths and/or at higher sampling rates than the 16-bit/44.1-kHz CD standard. The authors report on a series of double-blind tests comparing the analog output of high-resolution players playing high-resolution recordings with the same signal passed through a 16-bit/44.1-kHz “bottleneck.” The tests were conducted for over a year using different systems and a variety of subjects. The systems included expensive professional monitors and one high-end system with electrostatic loudspeakers and expensive components and cables. The subjects included professional recording engineers, students in a university recording program, and dedicated audiophiles. The test results show that the CD-quality A/D/A loop was undetectable at normal-to-loud listening levels, by any of the subjects, on any of the playback systems. The noise of the CD-quality loop was audible only at very elevated levels.

Authors:
Affiliation:
JAES Volume 55 Issue 9 pp. 775-779; September 2007
Publication Date:

Click to purchase paper or you can login as an AES member to see more options.

(Comment on this paper)

Comments on this paper

Default Avatar
Jonathan D. Boley
Comment posted August 27, 2008 @ 13:23:47 UTC (Comment permalink)

My compliments to the authors for performing this study. The result that immediately struck me as odd was the fact that female subjects scored 37.5% correct... well below chance. I wonder, was the variability greater for this subject pool?
Also, were the results analyzed only in terms of correct/incorrect? Or were the results also analyzed for false positives and false negatives?
For example, it would be interesting if the female subjects, or anyone else, tended to be biased toward answering either A or B, thus skewing the results. (This would be a great opportunity to apply Signal Detection Theory.)


Default Avatar
Douglas Rife
Comment posted August 31, 2008 @ 02:40:34 UTC (Comment permalink)

I read this paper with great interest when first published almost 1 year ago. Considering recent efforts promoting the supposed superiority of high resolution digital audio formats (24 bit, 96 kHz PCM as well as DSD) the authors' conclusion of no detectable audio quality difference between these formats compared to ordinary CD resolution (16 bit, 44.1 kHz PCM) would indeed be expected to create controversy. While I'm sympathetic to the authors' findings, it would have been a much stronger paper if the authors had revealed exactly what "well regarded CD recorder with real-time monitoring" was used to reduce the resolution of the audio signal to the CD standard. The authors claim their mystery CD recorder performed an A/D conversion using only 16-bit resolution and a 44.1 kHz sampling rate, followed by an immediate conversion back to the analog domain using a 16-bit, 44.1 kHz DAC (see Figure 1). But readers have no way to verify this claim without knowing the exact equipment the authors used in their tests and that they verified by correspondence with the manufacturer of this mystery CD recorder that the A/D/A conversion block shown in Figure 1 of their paper was actually being implemented. In other words, the authors need to leave no doubt that this block is really an A/D/A function at 16-bit resolution and not, for example, a purely analog monitor function or, one made at a higher bit resolution or sample rate.


Default Avatar
Author Response
David R. Moran
Comment posted September 2, 2008 @ 15:18:16 UTC (Comment permalink)

Mr Rife,
Please see whether the below information, which was posted shortly after the article was printed, answers your questions about our equipment's integrity. Of course we have always intended to make it possible for others to duplicate the experiment.

http://www.bostonaudiosociety.org/explanation.htm

I am sure Brad Meyer will have more commentary.

We also soon will try to provide some explanations for statistical distribution probabilities (binomial) for Mr Boley.

— David Moran (coauthor)


Default Avatar
Douglas Rife
Comment posted October 18, 2008 @ 23:06:08 UTC (Comment permalink)

Mr. Moran:
Thanks for the link above to the Boston Audio Society.
The HHB CDR-850 professional CD recorder was used as the A/D/A block. Fine.
Still, I found nothing in that link demonstating that this CD recorder's monitor function reduces the resolution to 16 bits and the sampling rate to 44.1 kHz. It's a reasonable assumption but, as I said in my original comment, it needs to be proven beyond all doubt if your going to publish a paper on such a controversial subject. The HHB web site does not even include the basic user's manual for this model. Don't forget Murphy's law. Don't forget the Hubble space telescope's main mirror, that everyone knew was the correct shape.


Default Avatar
Author Response
David R. Moran
Comment posted October 19, 2008 @ 18:15:41 UTC (Comment permalink)

>> that this CD recorder's monitor function reduces the resolution to 16 bits and the sampling rate to 44.1 kHz ... needs to be proven beyond all doubt if your [sic] going to publish a paper on such a controversial subject. ... Don't forget Murphy's law.

Assuming you are being serious here and that this is a serious worry being expressed, do you think it possible that the CDR we used has *better* performance than 16/44? In other words, and on the other hand, if it is of crummy performance, say, only 14-15 bits and 40k bandwidth or some such, what would that show? It would show that hi-rez can be degraded *way below* the CD standard and still be audibly indistinguishable from hi-rez. Uh, is that what you intended? I mean, think about it. Which direction does Murphy work in here?

I will ask Brad Meyer to weigh in about the performance of this recorder, what he has measured, if he has, and what the official spex are. I wonder if this will be satisfactory. Murphy did not make a law about paranoia!

The only review I could quickly find with a search reported:

\\\ I checked the performance of the built-in sample-rate converter with a varispeeded CD source and found it tracked extremely well, as it also did with both 32 and 48kHz stable signals. I also made several test recordings of familiar material copied from commercial CDs, both with and without the SRC in circuit (on adjacent tracks in some cases), but could not hear any significant difference. To my ears, and those of my friends and colleagues, all of the copies made on the CDR850 were indistinguishable from the source material when played back on the source machine (material was copied mainly from a Meridian 508 CD player using the optical or co-ax connections, co-ax giving the best results).
(Hugh Robjohns, http://www.soundonsound.com/sos/may99/articles/hhb.htm, 5/99)

I wonder if this is persuasive enough testimony. In any case, surely you do not think the unit might possibly exceed the CD standard, do you?


Default Avatar
Douglas Rife
Comment posted October 21, 2008 @ 00:31:14 UTC (Comment permalink)

Of course, I'm serious about this. Let's be clear that IF a recording is burned to a CD and subsequently that CD is played back from the start, there would be little doubt the resolution was being reduced to 16/44. However, that method would be quite cumbersome in practice as it would require pre-recording the hi res source to CD, "rewinding" both source and CD and then restarting the playback of the source in exact synchronization with the CD playback. Thus, as your paper strongly implies, you used the CD recorder's real time monitor function, which avoids the synchronization problem altogether since, being real time, the audio signal would not be delayed by more than a few tens of milliseconds and possibly much less. However, the real time monitoring function could very well be performed completely in the analog domain, without ever being converted to the digital domain! If that is the case then your A/B listening tests prove only that the analog signal path of this CD recorder is very transparent. Note that in the case of analog tape recorders having separate record and playback heads, it's easy to monitor the actual recording taken right off the tape in near real time. But CD drives do not normally allow for this type of monitoring so a CD recorder's monitor function could be implemented purely in the analog domain or, possibly in the digital domain but at a higher resolution than 16/44. Most A/D converters operate at a much higher resolution than 16/44. The reduction to 16/44 need not be done until the data is actually burned onto the CD. There is no requirement I've ever heard of that the real time monitor functions of CD recorders reduce the resolution to 16/44. That is what needs to be demonstrated but you can't use personal listening to do that since you are trying to test the audibility of lowering the resolution to 16/44 using controlled listening tests with many listeners and with several different hi res audio playback systems.


Default Avatar
Author Response
E. Brad Meyer
Comment posted October 23, 2008 @ 17:47:45 UTC (Comment permalink)

Mr. Rife; Of course, the CDR-850 does what we said. If you had it in your system you would quickly realize this.

If you just send a signal to the analog inputs and advance the record volume control, there is no indication on the level meters and no audio comes out of the output.

With no disc inside, pressing the RECORD button puts the machine in MONITOR mode (as the user manual says, without spelling out specifically what this means). The meters move, and signal appears at the output. That the audio is coming from the D/A converter is shown by several things, most notably (1) the visible and audible hard clipping that occurs simultaneously with the activation of the overload lights, (2) the noise floor at 92 dBA below the clipping level, (3) a slight but (with very good equipment) measurable frequency response error, on the order of +0.2 dB between about 9 and 11 kHz (if memory serves), and (4) the very steep roll-off above 20 kHz due to the anti-aliasing filter. Good modern preamplifier stages are at least 20 dB quieter than this, and you would have to try hard to design one that performs as badly in just these ways.

Meanwhile, under the assumption that the previous is inadequate to prove the case to you, I am obtaining the block and circuit diagrams from the service manual. I talked to the service tech at Sennheiser in Connecticut, who says that yes, this is the way it works, but I have an email into the engineering staff in England for a definitive quote to post here, for the record. — E. Brad Meyer


Default Avatar
Douglas Rife
Comment posted October 27, 2008 @ 18:12:33 UTC (Comment permalink)

Mr. Meyer: Thank you very much for the additional information regarding the CDR-850. Another clue you did not mention would be to measure the input-output impulse response (IR). A large delay of several milliseconds or more would be expected for a digital signal path due to decimation and reconstruction FIR filters. Furthermore, such FIR filters, as you know, are typically linear phase, exhibiting symmetrical (in time) ringing behavior in their IR. The noise floor measurement needs to be made with full scale sinewave input to elicit the quantization noise of all the available bits of resolution. A statement from the manufacturer is also needed as confirmation. I'm surprised that my rather obvious question was not asked and answered long ago, indeed, well before the publication of your report. Doug Rife


Drew Daniels
Drew Daniels
Comment posted September 14, 2008 @ 02:16:48 UTC (Comment permalink)

[Recently an article on audio quality by Meyer and Moran (J. Audio Eng. Soc. vol. 55, pp. 775-779, [2007 Sep.]) has given rise to a number of letters sent to the Journal. There is a lot of intense debate and opinion on this topic that probably would be interesting to many AES members, and this might actually help to bring out the key aspects of the issue. Accordingly, we feel that an AES online forum that allows members to give their opinions and experiences will be valuable.]—JAES e-mail to members.

That debate about this issue is intense there is no doubt. Can such debate be borne entirely of electroacoustical differences which might actually be impossible to explain, demonstrate and reproduce in a statistically significant way, or is it perhaps possible that financial interests could be at risk if some ultimate determination of audibility suggested that a large number of digital audio products were overbuilt, overpriced, or even entirely unnecessary?

This author deplores fear-mongering and considers its use as a marketing tool on people with insecurities about their technical knowledge, a crime of fraud. The audio marketplace, its many commercially oriented magazines, and the manufacturers who advertise in these magazines, now rarely if ever, supply graphical engineering data or even detailed and accurate print information that might be used to make purchasing decisions. The vast majority of equipment manuals are a list of sales points and usually not helpful. More and more, manufacturers rely on convincing potential customers of the superiority of their products by subjective means and anecdotal comments from celebrities. This is marketeering—not engineering--and it harms consumers everywhere. Those who are forced to purchase manufactured products to remain competitive in their services to clients, products which can not be tested in personal labs before purchases are made, should speak up about what has become an excessively greedy and cynical marketplace where increasingly, we are victims of marketeers rather than beneficiaries of good engineering.

Drew Daniels


Subscribe to this discussion

RSS Feed To be notified of new comments on this paper you can subscribe to this RSS feed. Forum users should login to see additional options.

Join this discussion!

If you would like to contribute to the discussion about this paper and are an AES member then you can login here:
Username:
Password:

If you are not yet an AES member and have something important to say about this paper then we urge you to join the AES today and make your voice heard. You can join online today by clicking here.

 
Facebook   Twitter   LinkedIn   Google+   YouTube   RSS News Feeds  
AES - Audio Engineering Society