Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Audibility of "typical" Digital Filters in a Hi-Fi Playback  (Read 367777 times) previous topic - next topic
0 Members and 14 Guests are viewing this topic.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #775
OK, so listeners reported roughness or edginess and cut-off reverb tails...

But there is no indication of causation, as it could be false positives generated by system artifacts, switching, etc.
It's up to Amirs camp pushing the Hi-Re$ $cam to confirm system transparency to test, so no assumptions about cause>filters need to be made as you did above.
No system/setup/test transparency data = no conclusions, only wishful thinking fallacy assumptions by those with pecuniary interests. Like Amir's camp.

cheers,

AJ
Loudspeaker manufacturer


Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #777
Sorry, no.  Dither is an optional processing step in quantization.  The term quantization therefore has no dependency whatsoever on dither being used or not.

Exactly, so why call it "16 bit quantization" when it could be a more unambiguous "truncated to 16 bits", for example.

Truncation to fewer bits is re-quantization.  And per above, it is customary to drop the "re" from that.

Kees de Visser talked about "truncation" but you cite some paragraph about dithered quantization and the term "requantization" which nobody even mentioned. So I have to assume you're confused again.
"I hear it when I see it."

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #778
First I interpreted (apparently wrongly) that 24-bit quantization artifacts were audible in a 24-bit system, since the 24-bit filtered versions (condition nos. 1&4) have more (24-bit) dither than the 24/192 original. I didn't see why that would be consistent with the hypothesis that 16 bits are not sufficient.

OK, so listeners reported roughness or edginess and cut-off reverb tails, even in the 24/48 TPDF dithered version, implying that the filter is responsible for this (since the 24-bit dither shouldn't be audible).
What I would like to see in Fig.3 is the difference signals of the 24/192 original and each of the 6 filtered versions, to verify that those differences are below the threshold curves.


1) See the noise levels in my previous post. Doing filtering and dithering again changes this density from -191 to -188 dB (-83 dB in the figure given their acoustic gain).
They are purely speculating at this point. There's nothing definitive in their results that would support this speculation. Also, saying that 24 bit is not sufficient is not consistent with saying that 16 bit theoretically isn't.
Also keep in mind that the noise floor of one of the most "revealing" tracks lies a couple dB over even 16-bit dither noise.

2) If they did the filtering right, the difference should be around -110 dB in the critical 4 kHz area (linear spectrum, not density, since the difference is music) given the passband ripple of 0.025 dB which prevents nulling down to the dither.
(Difference here means simply subtracting the original sample by sample from the filtered and quantized version.)
"I hear it when I see it."

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #779
Also keep in mind that the noise floor of one of the most "revealing" tracks lies a couple dB over even 16-bit dither noise.

In a 192 Khz sampling system.
Amir
Retired Technology Insider
Founder, AudioScienceReview.com

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #780
If peak playback level at 16 bit resolution is in fact 120dB SPL, you )(*&(*& well should be able to hear the lack of dither in a piano decay.
You should also be able to hear the difference between TPDF and RPDF dither, too, via noise modulation.

On the other hand who in the name of Monty Python would listen at that level? :horrors:

For that matter, in a decent listening room, someone ought to be able to hear the (*&*(&(*&( dither as well.

Which leads to the question: What in the heck are these people doing? Trying to blow up their equipment, or destroy their hearing, or what?

I don't entirely know what was set up, and I don't want to know at this point. 

All I can say is: Are you the type of ruffian that would say "ni" to an old woman?
-----
J. D. (jj) Johnston


Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #782
In a 192 Khz sampling system.

Yes, and 16-bit TPDF noise density would still be lower in a 44.1 kHz channel, crossing the fig. 3 192 kHz test signal noise floor at ~15 kHz, which is already far below the hearing threshold of humans.

Also, why would it matter? According to their interpretation of the results less bits = more hi-res like sound at 44.1 kHz.
"I hear it when I see it."

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #783
Guess I'll keep my CDs...and my hearing
Loudspeaker manufacturer

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #784
Damnit, Janet,  there is a reason my avatar has a manure fork in hand...
-----
J. D. (jj) Johnston

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #785
Damnit, Janet,  there is a reason my avatar has a manure fork in hand...

I thought that was for the well manicured vegetable garden....
Amir
Retired Technology Insider
Founder, AudioScienceReview.com

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #786
Damnit, Janet,  there is a reason my avatar has a manure fork in hand...


We could do with its services here.
Regards,
   Don Hills
"People hear what they see." - Doris Day


Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #788
TPDF dither was used

Yep, by the BS crew...and here is the result:





This gets so tedius!

From:
The audibility of typical digital audio filters in a high- delity playback system  AES paper 9174

Page 5

"
After filltering with either FIR fi lter, the signals
were either unchanged or were quantized to 16-
bit. The quantization either included RPDF (rect-
angular probability density function) dither or did
not.

Page 8

"A within-subjects ANOVA was performed on the
data for all eight listeners with two factors: filter
cutoff frequency (low or high) and quantization
type (none, 16-bit, or 16-bit plus RPDF dither).

Page 9:

"
Post-hoc Fisher tests based on a least-signi cant dif-
ference of means at a 5% level of 0.1394 for quanti-
zation showed that performance was worse for 16-bit
quantization plus RPDF dither (mean=1.773) than
for no quantisation (mean=1.902) or 16-bit quanti-
sation alone (mean=1.909).

Page 9-10

"Every condition where 16-bit quantization was ap-
plied gave performance that was signi cantly better
than chance. Performance was signi cantly worse
for 16-bit quantization plus RDPF dither than for
no quantisation or 16-bit quantization alone. This
suggests that the e ect of adding the RPDF dither
on top of the 16-bit quantization and FIR ltering
was to make it more dicult to identify that process-
ing had been applied to the signal, which is perhaps
counterintuitive. To try to explain this nding, we
turn to subjective descriptions of what listeners de-
scribed hearing in these tests.
"

Page 10

"It is possible that that these two qualitative ef-
fects on sound quality cancelled out each other for
the conditions where 16-bit quantization plus RPDF
dither was applied, which could explain why these
signals were harder to discriminate from their un l-
tered counterparts."

"All forms of processing tested here were audi-
ble, except for one condition where performance was
signi cantly di erent from chance at the 6.7% level,
including emulated downsampling lters at standard
sample rates and 16-bit quantization with or without
RPDF dither. Di erences were demonstrated here
in a double-blind test using non-expert listeners who
received minimal training.

"16-bit quantization with and without RPDF
dither can have a deleterious e ect on the listen-
ing experience in a wideband playback system.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #789
On the other hand who in the name of Monty Python would listen at that level? :horrors:
I've expressed my surprise about the high playback levels before.
One of the first recordings I made as a student recording engineer was Bruckner's 9th symphony. 16 bit was the best we had in the 80's. We had made a nice balance in the control room and were quite happy with the playback level. We didn't miss detail. Then the conductor came for a listen. "louder please". So I increased the monitor level by 6 dB, which felt uncomfortable for us, but well, anything for a happy conductor. "more", he said. We ended up with about 12 dB extra gain and still the conductor felt it was too soft, but it would do. We couldn't squeeze more out of our monitors. I guess around 105 dB was the limit, which is definitely not enough when directly compared with a live symphony orchestra.
So was 16 bit enough for the recording? Yes and no. I can't imagine anyone listening at realistic levels in a home situation. It all boils down to the definition of "typical".
Then there's another anecdote about airplane recording

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #790
TPDF dither was used

Yep, by the BS crew...and here is the result:





This gets so tedius!

From:
The audibility of typical digital audio filters in a high- delity playback system  AES paper 9174

Page 5

"
After filltering with either FIR fi lter, the signals
were either unchanged or were quantized to 16-
bit. The quantization either included RPDF (rect-
angular probability density function) dither or did
not.

Page 8

"A within-subjects ANOVA was performed on the
data for all eight listeners with two factors: filter
cutoff frequency (low or high) and quantization
type (none, 16-bit, or 16-bit plus RPDF dither).

Page 9:

"
Post-hoc Fisher tests based on a least-signi cant dif-
ference of means at a 5% level of 0.1394 for quanti-
zation showed that performance was worse for 16-bit
quantization plus RPDF dither (mean=1.773) than
for no quantisation (mean=1.902) or 16-bit quanti-
sation alone (mean=1.909).

Page 9-10

"Every condition where 16-bit quantization was ap-
plied gave performance that was signi cantly better
than chance. Performance was signi cantly worse
for 16-bit quantization plus RDPF dither than for
no quantisation or 16-bit quantization alone. This
suggests that the e ect of adding the RPDF dither
on top of the 16-bit quantization and FIR ltering
was to make it more dicult to identify that process-
ing had been applied to the signal, which is perhaps
counterintuitive. To try to explain this nding, we
turn to subjective descriptions of what listeners de-
scribed hearing in these tests.
"

Page 10

"It is possible that that these two qualitative ef-
fects on sound quality cancelled out each other for
the conditions where 16-bit quantization plus RPDF
dither was applied, which could explain why these
signals were harder to discriminate from their un l-
tered counterparts."

"All forms of processing tested here were audi-
ble, except for one condition where performance was
signi cantly di erent from chance at the 6.7% level,
including emulated downsampling lters at standard
sample rates and 16-bit quantization with or without
RPDF dither. Di erences were demonstrated here
in a double-blind test using non-expert listeners who
received minimal training.

"16-bit quantization with and without RPDF
dither can have a deleterious e ect on the listen-
ing experience in a wideband playback system.

I guess the question then becomes, exactly how "typical" is RPDF dither in a "typical" HiFi playback system in 2014?
Outside a certain Meridian test lab not using their own 518, etc. players, anyone know?

cheers,

AJ
Loudspeaker manufacturer


Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #792
TPDF dither was used

Yep, by the BS crew...and here is the result:





This gets so tedius!

From:
The audibility of typical digital audio filters in a high- delity playback system  AES paper 9174

Page 5

"
After filltering with either FIR fi lter, the signals
were either unchanged or were quantized to 16-
bit. The quantization either included RPDF (rect-
angular probability density function) dither or did
not.

Page 8

"A within-subjects ANOVA was performed on the
data for all eight listeners with two factors: filter
cutoff frequency (low or high) and quantization
type (none, 16-bit, or 16-bit plus RPDF dither).

Page 9:

"
Post-hoc Fisher tests based on a least-signi cant dif-
ference of means at a 5% level of 0.1394 for quanti-
zation showed that performance was worse for 16-bit
quantization plus RPDF dither (mean=1.773) than
for no quantisation (mean=1.902) or 16-bit quanti-
sation alone (mean=1.909).

Page 9-10

"Every condition where 16-bit quantization was ap-
plied gave performance that was signi cantly better
than chance. Performance was signi cantly worse
for 16-bit quantization plus RDPF dither than for
no quantisation or 16-bit quantization alone. This
suggests that the e ect of adding the RPDF dither
on top of the 16-bit quantization and FIR ltering
was to make it more dicult to identify that process-
ing had been applied to the signal, which is perhaps
counterintuitive. To try to explain this nding, we
turn to subjective descriptions of what listeners de-
scribed hearing in these tests.
"

Page 10

"It is possible that that these two qualitative ef-
fects on sound quality cancelled out each other for
the conditions where 16-bit quantization plus RPDF
dither was applied, which could explain why these
signals were harder to discriminate from their un l-
tered counterparts."

"All forms of processing tested here were audi-
ble, except for one condition where performance was
signi cantly di erent from chance at the 6.7% level,
including emulated downsampling lters at standard
sample rates and 16-bit quantization with or without
RPDF dither. Di erences were demonstrated here
in a double-blind test using non-expert listeners who
received minimal training.

"16-bit quantization with and without RPDF
dither can have a deleterious e ect on the listen-
ing experience in a wideband playback system.

I guess the question then becomes, exactly how "typical" is RPDF dither in a "typical" HiFi playback system in 2014?
Outside a certain Meridian test lab not using their own 518, etc. players, anyone know?


The dither in modern recordings can come from many places. Its function needs to be part of many common signal processing operations in AVRs such as gain and volume controls.

I did some quick tests to see if one can easily tell the difference between no dither, RPDF dither and TPDF dither being applied to a conversion of a pure -60 dB sine wave from 32 bit floating point to 16 bit fixed point. The effects of no dither is clearly shown in the output 65 k point FFT analysis, but the difference between RPDF and TPDF is not clear at all to me. 

If we are talking common signals with noise floors around FS -70 dB, none of the options make any apparent difference due to self-dithering.  For example mixing  a -60 dB sine wave with -70 dB brown noise (fair simulation of acoustical noise from natural sources) is sufficient to render the 32 bit -> 16 bit conversion free of any of the visible artifacts related to quantization error. Nothing coherent in the 65k point FFT noise floor other than the original 1 KHz signal.


Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #794
I did some quick tests to see if one can easily tell the difference between no dither, RPDF dither and TPDF dither being applied to a conversion of a pure -60 dB sine wave from 32 bit floating point to 16 bit fixed point. The effects of no dither is clearly shown in the output 65 k point FFT analysis, but the difference between RPDF and TPDF is not clear at all to me.

Much of this seems academic to me.
I've lost track of how many "Hi End" audio shows I've attended. I've personally exhibited my own products in at least 10, helped with more. I'm a member of the local audiophile club, with in excess of 100 members, that meets at minimum once a month (including today!) "officially", far more unofficially. Plus we occasionally link up with other clubs within FL.
So I have a rather large exposure to "Hi End" audiophiles who would be the exact target of "Hi-Re$".
I would guesstimate, based on this exposure, that less than 2% of "audiophiles" listen to classical music. The only type of music that might remotely approach the dynamic range/noise limits of 16/44, at absurd levels, in very quiet, highly atypical rooms for said audiophiles.
I've also lost count of how many, upon entering room at shows, have asked, is that "Hi-Re$"?
Only to be told, nope, just good ol' "low res" 16/44...using WMP (thanks JJ...and oh, yes, Amir ). I'll then be asked about 99% of the time, to play something akin to Pink Floyd/Neil Young, NOT classical.
Yes, I know, largely anecdotal....but I seriously wonder just how "typical" these contrived test conditions of the BS paper are, never mind how "typical" the "emulated" filters are, the playback levels for string quartet music, etc, etc....or the lack of any system transparency data.

cheers,

AJ
Loudspeaker manufacturer

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #795
Sorry, no.  Dither is an optional processing step in quantization.  The term quantization therefore has no dependency whatsoever on dither being used or not.

Exactly, so why call it "16 bit quantization" when it could be a more unambiguous "truncated to 16 bits", for example.

Truncation to fewer bits is re-quantization.  And per above, it is customary to drop the "re" from that.

Kees de Visser talked about "truncation" but you cite some paragraph about dithered quantization and the term "requantization" which nobody even mentioned. So I have to assume you're confused again.
The terms that Amir and the paper's authors used have been used in AES publications for decades. Just because they're not the terms we might use here does not make them wrong. As Amir said, the intended audience would understand them. This quantisation vs truncation argument is misguided IMO.


This isn't targeted at you xnor, but I think some of the other arguments we've had in this thread show a lack of understanding. When I finally got hold of the paper, I wondered if everyone who claimed to have read it really had.


I'm not trying to sound clever. I'm suggesting that some people on both sides are so keen to argue that they don't take the time to figure out what the words they've read really mean.

Cheers,
David.

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #796
The effects of no dither is clearly shown in the output 65 k point FFT analysis, but the difference between RPDF and TPDF is not clear at all to me. 

If we are talking common signals with noise floors around FS -70 dB, none of the options make any apparent difference due to self-dithering.  For example mixing  a -60 dB sine wave with -70 dB brown noise (fair simulation of acoustical noise from natural sources) is sufficient to render the 32 bit -> 16 bit conversion free of any of the visible artifacts related to quantization error. Nothing coherent in the 65k point FFT noise floor other than the original 1 KHz signal.

Remove any extra noise and compare the RPDF dither noise floor with just a -60 dB sine vs. a -120 dB sine. You should see that the noise changes with signal level with RPDF.

With all the loads of room/acoustic noise in the real-world recording, none of this really seems to make a difference.
"I hear it when I see it."

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #797
... I did some quick tests to see if one can easily tell the difference between no dither, RPDF dither and TPDF dither being applied to a conversion of a pure -60 dB sine wave from 32 bit floating point to 16 bit fixed point. The effects of no dither is clearly shown in the output 65 k point FFT analysis, but the difference between RPDF and TPDF is not clear at all to me.


The problem as I understand it is RPDF dither [unlike TPDF] suffers from noise modulation, key word "modulation", so the problem wouldn't occur from a fixed level recording which is I believe what you attempted.

Remember the rival tape NR system to Dolby, "dbx"? It too was a complimentary, dynamic range "compansion" system [COMPression applied during recording/ re-expANSION in playback] but it worked over a broader range. Although the complaints about it were overblown (in my opinion) it was true that under some odd scenarios it suffered from "pumping and breathing artifacts". The classic one (that comes to mind) was attempting to record the solo heart beat from the opening cut of Pink Floyd's DSotM. Although different in causation, I think a dynamically changing noise floor is RPDF dither's similar issue.

edit to add: If we had access to the tracks used here we'd have  some examples to listen to:

http://www.coe.montana.edu/ee/rmaher/aes_c...information.pdf

My links often fail in this forum, I don't know why, the article is: Quantization, Dither, and Noise Shaping Demonstrations. by . Stanley P. Lipshitz,

I think I found them! [Exploring them now]: http://www.coe.montana.edu/ee/rmaher/aes_cd/1.2.htm

Bingo! As I suspected, RPDF dither, for instance used in tracks 18 -22, is similar to "pumping and breathing" artifacts, whereas track 23, made with TPDF dither, is immune to this noise modulation problem. [Track 1 is the clean, unmodified, reference version.]

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #798
The terms that Amir and the paper's authors used have been used in AES publications for decades. Just because they're not the terms we might use here does not make them wrong. As Amir said, the intended audience would understand them. This quantisation vs truncation argument is misguided IMO.


This isn't targeted at you xnor, but I think some of the other arguments we've had in this thread show a lack of understanding. When I finally got hold of the paper, I wondered if everyone who claimed to have read it really had.


I'm not trying to sound clever. I'm suggesting that some people on both sides are so keen to argue that they don't take the time to figure out what the words they've read really mean.

Cheers,
David.


Agreed .  'quantization' vs 'requantization' is not the droid we are looking for.

'Quantization' is just the digital representation of amplitude.  Going from 24bit to 16 bit, the 're' is hardly necessary, it's still a quantization of amplitudes.

But it would have been good for the Meridian folk to have been more explicit as to the order of downconversion operations (i.e,  application of dither). It would hardly have taken up significantly more space, just a minor re-phrasing.

Perhaps when/if this is published formally in JAES, there will be some modifications, or links to supplementary materals.

 

Audibility of "typical" Digital Filters in a Hi-Fi Playback

Reply #799
I think the suggestion that they'd get dither and quantisation the wrong way around, and not mention it, is going too far.