Assessing The Sound Quality Of Lossy Compression Systems

[alanh]

EBU article

This article describes how and the results of subjective testing.

The only relevance to our system is that we use SBR in DAB+, the compression system is different to DAB so you cannot use the bit rate conclusion in our system. Since the error correction used in DAB+ is much more powerful, you cannot use the signal strength of failure either.

The speech from low data rate Information radio in Perth is much better than AM complete with some high frequencies on the letter 's'

AlanH

[Digital Penetration]

The speech from low data rate Information radio in Perth is much better than AM complete with some high frequencies on the letter 's'

AlanH

Please provide recordings to back up this claim. I would be interested to compare them.

[GoForMoe]

The only relevance to our system is that we use SBR in DAB+, the compression system is different to DAB so you cannot use the bit rate conclusion in our system. Since the error correction used in DAB+ is much more powerful, you cannot use the signal strength of failure either.

Considering the article is irrelevant to DAB+ in quite significant ways, why are you posting it? Would it not be logical that SBR has different impacts on MP2 audio quality than on AAC?

The speech from low data rate Information radio in Perth is much better than AM complete with some high frequencies on the letter 's'

Funny how this response comes in a different thread and about a month after my question to you over 6RPH's bit rate and sound quality. Can you prove they use SBR? Likewise are you claiming that SBS didn't on their same rate services?

[DrP]

I think everyone should consider carefully before responding to a thread where the topic author is so uninformed on the topic that he still calls it 'Spectrum Bandwidth Replacement', even when the very article linked correctly names it Spectral Band Replication.

I wonder how long before the Paramedic Steroids are rolled out too.

Time and time again the OP demonstrates that all he is doing is cutting and pasting stuff he stumbles across on the internet. There is no comprehension, and from the looks of it not even a brief read of the content. For supporting evidence look no further than the thread regarding DAB broadcasting in Spain.

[alanh]

Double post

Edited August 1, 2011 by alanh

[alanh]

All,

My objective is not to champion SBR but to show how extensive subjective testing is carried out to determine the acceptability of a particular transmission system.

Alanh

[Digital Penetration]

All,

My objective is not to champion SBR but to show how extensive subjective testing is carried out to determine the acceptability of a particular transmission system.

Alanh

Unfortunately the only subjective test that seems to count in the long run is "How many streams can we shove in versus how much money can we squeeze out of the punters until they start to turn off in droves."

[charlesc]

All,

My objective ....

Why not post that in the first post then?

[DrP]

Why not post that in the first post then?

Its an alanh thread. What else could you expect?

BTW, I'd like to point out that the linked article references BBC tests. Forum readers will recall that alanh couldn't locate the easily locatable BBC tests in earlier threads and therefore wrote off any contribution that the BBC tests made because (paraphrase) "BBC made no such tests". No doubt the real reason for alanh being "unable to locate" was due to the fact that the BBC tests contradicted his then position.

I have done an extensive search of the BBC's Research and Development Department's website. There is no studies comparing subjective effects of audio compression systems. So since you have named them as a source, come up with the link.

Edited August 1, 2011 by DrP

[MLXXX]

All,

My objective is not to champion SBR but to show how extensive subjective testing is carried out to determine the acceptability of a particular transmission system.

Alanh

All lossy perceptual audio codecs are subject to extensive subjective testing in their development in order to get the best 'bang per buck', i.e. the least noticeable or least objectionable degradation for a target bitrate.

With a DAB/DAB+ ensemble of streams, the more efficient and effective the audio codec, and the error correction supporting the compressed audio data, the greater the number of streams that can be accommodated at a "reasonable quality" for the total bitrate capacity of the ensemble. The lower the audio codec bitrate, the greater the challenge.

In Australia, broadcasters seem to have settled on 64kbps (nominal) DAB+ for 'standard quality', used for music; and 48kbps (nominal) used predominately for speech. A lower rate, 32kbps (nominal), is in use to some extent, but is poorly regarded. An audio quality noticeably above 'standard quality' can be heard from those stations that use a nominal bitrate of 80kbps or greater.

It has been reported by a number of listeners on this forum that the subjective quality of 64kbps (nominal) DAB+ falls below the standard achieved with FM broadcasting in the prime reception area, with a good quality FM receiver and antenna. A number of people have uploaded captured audio samples to this forum to illustrate such quality differences. [DAB+ compression has the strengths of very low noise, and low harmonic distortion, when decompressed. However at lower bitrates it suffers from hollowness and "wooshiness". Also, the SBR tends to have a blandness and artificiality about it, compared with FM radio from an uncompressed source. These side effects of SBR are more noticeable at lower bitrates.]

Alanh, I note that you have not responded to this request:

The speech from low data rate Information radio in Perth is much better than AM complete with some high frequencies on the letter 's'

AlanH

Please provide recordings to back up this claim. I would be interested to compare them.

Edited August 3, 2011 by MLXXX

[MLXXX]

The speech from low data rate Information radio in Perth is much better than AM complete with some high frequencies on the letter 's'

Alanh has been conspicuous by his silence, when asked to back up his claim above.

Perhaps some 2SM streaming at 32kbps AAC+ will go some way towards acquainting listeners with the general standard to expect:

http://209.212.144.153:7100/listen.pls

To listen to 2SM using VLC Player please use this link.

Your VLC Player should load automatically to play our stream.

[from

http://www.2sm.com.au/listen.html

]

For my ears, the ''mushiness" of the AAC+ is revealed, even when listening to the caller in a talkback segment. How that quality of sound is "much better than AM" escapes me.

The lowest nominal bitrate currently in use in Brisbane for DAB+ is 48kbps. I can recall how the 32kbps SBS streams sounded at the start of permanent DAB+ broadcasting in Australia; in my view they were an insult to ethnic communities.

Alanh has provided no recordings, no information on distance from the AM transmitter, length of AM aerial, or whether at night or during the day. In short, he has made a vague and unsubstantiated claim.

[alanh]

MLXXX,

I live very close to a high powered AM transmitter, and all DAB+ sounds better than it because of the addition of high frequencies and stereo, which is not available to the general public. I would rather use DAB+ of the same signal than a diode receiver as you have suggested. When multiple high powered transmitters are close together it is also hard to prevent interference between them without reducing the bandwidth.

I will not put on line samples because the coding of digital signals causes interactions between the coders particularly when clock frequencies are not synchronised and the coding algorithms interact.

As I have pointed out before SBS did not feed their encoder correctly causing these artifacts.

Why do you think that DRM has been developed and is being broadcast using AAC V2 which is also used by DAB+ and low data rates if it sounds as bad as you say.

AlanH

[DrP]

MLXXX,

I live very close to a high powered AM transmitter, and all DAB+ sounds better than it because of the addition of high frequencies and stereo, which is not available to the general public. I would rather use DAB+ of the same signal than a diode receiver as you have suggested. When multiple high powered transmitters are close together it is also hard to prevent interference between them without reducing the bandwidth.

I will not put on line samples because the coding of digital signals causes interactions between the coders particularly when clock frequencies are not synchronised and the coding algorithms interact.

As I have pointed out before SBS did not feed their encoder correctly causing these artifacts.

Why do you think that DRM has been developed and is being broadcast using AAC V2 which is also used by DAB+ and low data rates if it sounds as bad as you say.

AlanH

Proof of statement required.

If no proof, beyond extraneous waffling, is provided it has to be assumed that as per usual, alanh is talking out of an orifice that his not his mouth.

Edited August 4, 2011 by DrP

[MLXXX]

Why do you think that DRM has been developed and is being broadcast using AAC V2 which is also used by DAB+ and low data rates if it sounds as bad as you say.

I think you have misread my remarks. I do not doubt that HE-AAC v2 is an efficient and effective codec. It's well known that at low bitrates it outperforms mp3 and mp2.

A bitrate that is unacceptably low for use with mp2 may be acceptable for use with HE-AAC v2.

However, the fact an efficient audio perceptual codec is intelligible at a low bitrate does not mean it is well suited to the broadcast of music as entertainment. That is why for example 32kbps HE-AAC v2 if used at all for local broadcasting, tends to be restricted to use for speech rather than music.

With DAB+, the ABC use 48kbps nominal for news radio, 64kbps nominal for general broadcasting, and 80kbps for music. For me, with classical music, 80kbps noticeably surpasses 64kbps but still falls short of FM in a strong signal area with a good quality receiver, and good antenna. As you will recall, I have provided uploads illustrating the differences between 80kbps DAB+ and FM, for ABC Classic FM as received by me in Brisbane.

If you make simultaneous recordings of DAB+ and AM radio, you may find it convenient to use the freeware Audacity, to edit the recordings so as to align them in time. As regards matching amplitudes, I find that is best done by ear. Especially at 32kbps the perceptual codec will so massively modify the waveform that it is impossible to match the levels in the conventional manner of aligning peak or average level. In any event there may well be different processing of the audio prior to the signal being passed as audio to the DAB+ and AM modulators.

I will not put on line samples because the coding of digital signals causes interactions between the coders particularly when clock frequencies are not synchronised and the coding algorithms interact.

If you set your pc (line input) to sample at 44.1KHz, or 48KHz, that will be perfectly adequate to capture the line output of your DAB radio, faithfully, for the current purpose. (I am assuming you are using a standalone DAB+ radio rather than a DAB+ radio incorporated into a USB stick.)

Unless and until you, or someone else in your region, supplies actual recordings it will be impossible for members of this forum in other parts of Australia to ascertain how good or bad the RPH broadcasts in your region sound.

Edited August 4, 2011 by MLXXX

[MLXXX]

Proof of statement required.

Indeed. And it's so often hard to know what the "statement" even means. I make a guess below.

As I have pointed out before SBS did not feed their encoder correctly causing these artifacts.

Perhaps alanh is referring to the encoder being set for ordinary stereo rather than parametric stereo.

There is a dilemma at 32kbps for HE-AAC v2. Enable PS to improve the basic mono sound but lose accuracy of the stereo image; or disable PS for a proper stereo image but a poorer basic sound. The latter is the choice SBS made, at least that's the conclusion I came to for the broadcasts I heard in Brisbane. Artifacts were clearly audible in stereo and at times the music revealed a complex and realistic stereo image beyond what I would expect from PS.

I suggest it was a valid choice open to the SBS engineers to take. Ultimately it is a question of personal preference. Some of what SBS broadcast was stereo music. For my ears the music itself was a writhing, phasey thing; artifacts of a low bitrate. But the stereo effect was good!

What RPH are doing in Perth with their nominal 32kbps stream, in relation to using or not using PS, who knows. However I gather it would not be unusual to disable PS at 32kbps. A Wikipedia webpage, albeit unfortunately lacking in source references, claims:

Because only one audio channel is transmitted, along with the parametric side info, a 24 kbit/s coded audio signal with Parametric Stereo will be substantially improved in quality, compared to a discretely stereo coded audio signal coded with conventional means. Thus, the additional bitrate spent on the single mono channel (combined with some PS side info) will improve the perceived quality substantially of the audio compared to a standard stereo stream at similar bitrate.

However,

this technique is only useful at the lowest bitrates (say 16 - 32 kbit/s, sweet-spot at 24 kbit/s)

to give a good stereo impression, so while it can improve perceived quality at very low bitrates, it generally does not achieve transparency, since simulating the stereo dynamics of the audio with the technique is limited and generally deteriorates perceived quality regardless of the bitrate.

[emphasis added]

Edited August 4, 2011 by MLXXX