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: Point of transparency of Opus 1.2 (Read 43637 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: Point of transparency of Opus 1.2

Reply #25
Did some Foobar2k ABX testing using the "Bachpsichord" and "Applaud" samples:

Bachpsichord sample*
  Opus@64kbps          : Guessing Probability 0.0%
Applaud sample*
  Opus@64kbps, @96kbps : Guessing Probability 0.0%
Well, the more interesting question (at least to me) is: what an Opus bitrate setting does it take to have a) applaud, b) barpsichord being transparent to you.
lame3995o -Q1.7 --lowpass 17

Re: Point of transparency of Opus 1.2

Reply #26
Could someone please point me to where I can download these 'killer' samples? I'd like to try some ABX myself.

Also, OgGy, what made these so easy to identify? Is it pre-echo, hf roll off, something else?



Re: Point of transparency of Opus 1.2

Reply #29
... Anyway, as from opus@112kbps it becomes more difficult to ABX the samples, please find the results attached.
Thank you for testing.
Intereting as you get at 7/8 even @160 kbps with the applaud sample. Would you call the issue negligible at this bitrate? (I'm on holidays ATM and can't try for myself).
lame3995o -Q1.7 --lowpass 17

Re: Point of transparency of Opus 1.2

Reply #30
Actually, I believe I said that already in an other post but Chinese car stereos, mostly those cheap ones on ebay have a tendency to support open source codecs. Even though mine didn't advertise that it did, it supports APE, FLAC, OGG and OPUS. If you happen to own one of those I would suggest you to do a test. While on the outside some of them may look identical the internals might differ and they may support more or less codecs than the ones that they advertise.

I have owned two Chinese headunits (Android 4 + 5), and both were fucking horrendous. Constant issues with both software and hardware. Neither unit was what I would consider cheap, being around the £250 mark. The sounds quality was some of the worst I've eve heard (no, I didn't ABX).

The problem with using a non branded unit as suggested is that you get hardware which (in my opinion) has very little testing, and low test standards. The software is always junk. The reason good manufactures are limited to a small amount of formats is two reasons:

1) They are limited by available chips sets which are automotive grade. The Chinese ones tend just to use phone / tablet innards.
2) They need to limit the amount of support cost for the unit. I'd guess more formats means more customer support calls.

I've now moved to a Sony Apple Car Play unit now, which is of course, super limited in what it supports. But what it does, it does well.

Attached is a picture of the first Chinese unit I had. Looked ok, worked horribly.


These weren't the ones I was referring to. Anything Chinese Android based is terrible. What I was referring was basic 1DIN car stereos that go as much as 20 bucks on Ebay. I got one that actually plays all those formats I mentioned OPUS included, but an identical looking one that a friend of mine got didn't.

Re: Point of transparency of Opus 1.2

Reply #31
... Anyway, as from opus@112kbps it becomes more difficult to ABX the samples, please find the results attached.
Thank you for testing.
Intereting as you get at 7/8 even @160 kbps with the applaud sample. Would you call the issue negligible at this bitrate? (I'm on holidays ATM and can't try for myself).


Well @160kbps it is almost negligible but it suprises me how difficult it is to encode.  If I'm not mistaking the Opus Encoder uses only short blocks so that should not be the culprit ?

Re: Point of transparency of Opus 1.2

Reply #32
The reason good manufactures are limited to a small amount of formats is two reasons:

1) They are limited by available chips sets which are automotive grade. The Chinese ones tend just to use phone / tablet innards.
2) They need to limit the amount of support cost for the unit. I'd guess more formats means more customer support calls.
Boy! I'd love to know why chip sets manufactures insist on supporting almost-defunct (and, at least over here, unpopular) WMA and not AAC or Vorbis instead.
Listen to the music, not the media it's on.
União e reconstrução

Re: Point of transparency of Opus 1.2

Reply #33
The reason good manufactures are limited to a small amount of formats is two reasons:

1) They are limited by available chips sets which are automotive grade. The Chinese ones tend just to use phone / tablet innards.
2) They need to limit the amount of support cost for the unit. I'd guess more formats means more customer support calls.
Boy! I'd love to know why chip sets manufactures insist on supporting almost-defunct (and, at least over here, unpopular) WMA and not AAC or Vorbis instead.

Because these were likely the prominent formats when the chipset was validated. Besides, no one in the real world knows what Vorbis is. These people are trying to sell mass market products, not appeal out out lying audio nerds like us.

Re: Point of transparency of Opus 1.2

Reply #34
Boy! I'd love to know why chip sets manufactures insist on supporting almost-defunct (and, at least over here, unpopular) WMA and not AAC or Vorbis instead.

Because these were likely the prominent formats when the chipset was validated. Besides, no one in the real world knows what Vorbis is. These people are trying to sell mass market products, not appeal out out lying audio nerds like us.
I guess it's something along those lines, and yes, you were right in pointing out Vorbis lack of consumer outreach.

But let us not forget that, outside Windows Media Player, WMA is not that popular either. So I wonder if there's also Microsoft's own doing in keeping such a supersed format alive for so long.
Listen to the music, not the media it's on.
União e reconstrução

Re: Point of transparency of Opus 1.2

Reply #35
How does transparency of opus at 96 kbps compare against vorbis at 112 kbps (nominal bitrate in both cases)? I know this varies from person to person, but I would like to know forum members' opinions on this.

Re: Point of transparency of Opus 1.2

Reply #36
How does transparency of opus at 96 kbps compare against vorbis at 112 kbps (nominal bitrate in both cases)? I know this varies from person to person, but I would like to know forum members' opinions on this.
You'll learn with time (the sooner the better, so, in case you haven't yet, you should read this community's TOS ASAP) that, the very scientific empirical method which HA abides by (unlike the average "feel the force, Luke" audiophile website) will show you that, asking about whether a listening test comparing this or that format has ever been carried out (which, in your case, a simple search will tell you in a jif) - or even better: conducting yourself an ABX listening test - are totally valid under said guidelines - more specifically TOS #8.

Other than that, and hopefully, you'll grow accustomed to the fact that, over here, opinions regarding anyone's perception of quality are utterly worthless (because, let's face it, they're just that), unless properly corroborated by the afore-mentioned tests' results. 

So much so that any answer to that, without said results, would mostly be in direct conflict with TOS 8 - not to mention a contracdition per se, since, as you mentioned yourself: it "varies from person to person".
Listen to the music, not the media it's on.
União e reconstrução

Re: Point of transparency of Opus 1.2

Reply #37
How does transparency of opus at 96 kbps compare against vorbis at 112 kbps (nominal bitrate in both cases)? I know this varies from person to person, but I would like to know forum members' opinions on this.
You'll learn with time (the sooner the better, so, in case you haven't yet, you should read this community's TOS ASAP) that, the very scientific empirical method which HA abides by (unlike the average "feel the force, Luke" audiophile website) will show you that, asking about whether a listening test comparing this or that format has ever been carried out (which, in your case, a simple search will tell you in a jif) - or even better: conducting yourself an ABX listening test - are totally valid under said guidelines - more specifically TOS #8.

Other than that, and hopefully, you'll grow accustomed to the fact that, over here, opinions regarding anyone's perception of quality are utterly worthless (because, let's face it, they're just that), unless properly corroborated by the afore-mentioned tests' results.

So much so that any answer to that, without said results, would mostly be in direct conflict with TOS 8 - not to mention a contracdition per se, since, as you mentioned yourself: it "varies from person to person".

Hmm. Maybe I asked the wrong question.

What I should have said: "I know this varies from person to person, but I would like to know the results of forum members' personal listening tests."

FYI I checked to see if relevant public listening tests had been carried out and they haven't. This is hardly surprising considering I am asking about a specific pair of codecs at non-identical bit rates.

I could carry out ABX tests of my own and maybe I will. However, I would also be interested to see some kind of a sample of results across different people.

Re: Point of transparency of Opus 1.2

Reply #38
I guess you are asking if the point of transparency of Vorbis is 112k for peope who consider Opus transparent at 96 kbit.
I can tell you for me, I'm 42 yrs old, and last time I've tested other codecs (mp3, aac, voris) the point of transparency was at 128k. That was few years ago, and I haven't tested them since, I am happily using them at slightly bigger bitrate (usually VBR which ranges from 100 to 150 k). Two months ago I've tested Opus, and it was seemingly transparent even at 64 k, and as safety net I've been encoding it at 80 k. But I am using it only for recorded movies and tv shows, which have mixed speech and music. For that it's more than perfect... for me :)
Error 404; signature server not available.

Re: Point of transparency of Opus 1.2

Reply #39
Playing around with some 'QUICK ABX testing' on my PC, which has Klipsch Pro-Media speakers(which are better than average PC speakers for sure(i have had them since the early 2000's)), i started off at 32kbps and then went to 48kbps and then stopped at the 64kbps setting in that, without actually finishing a ABX test, i simply did a QUICK test without spending too much time and it seems once i hit 64kbps i lose confidence in being able to more easily notice the FLAC file vs the Opus 64kbps file by simply clicking back and fourth between Sample A and Sample B to where i can tell it's noticeably more difficult once i hit that point to where i lose confidence in claiming i can quickly spot the FLAC and Opus file without actually completing a ABX test.

so in other words... once i hit the 64kbps setting it's not easy for me to detect artifacts anymore, especially when spending little time without focusing TOO much on things. because as others have mentioned once you start to strain on finding artifacts that's probably a good general guideline of a minimum bit rate you would consider using for general listening to the music straight up to enjoy it. because it appears some people around here can go beyond 128kbps Opus occasionally but i figure at that point you got to be more in that hardcore nitpicking level to where it's almost a non-issue when just straight up enjoying your music if by some chance some minor artifact turns up very briefly on a random song you got as to me it's not worth loading up on excessive bit rate just to 'maybe' clean up that semi-rare artifact that turns up if say the vast majority of music is either transparent or close to transparent to most people at say 96-128kbps range.

so given that info... i suspect ill be very similar to those who mentioned the 80kbps being hard to spot artifacts crowd of people that IgorC mentioned in here. who knows if there would be much of a difference for me if i tested with headphones. even if there is, i can't see myself going all that much higher with the bit rates in Opus especially assuming what was mentioned in this topic that Opus @ 96kbps is 'approaching transparency' is a good ball park figure.

p.s. for whatever it's worth ill be 38 years old later this month.
For music I suggest (using Foobar2000)... MP3 (LAME) @ V5 (130kbps). NOTE: using on AGPTEK-U3 as of Mar 18th 2021. I use 'fatsort' (on Linux) so MP3's are listed in proper order on AGPTEK-U3.

Re: Point of transparency of Opus 1.2

Reply #40
hi
may i ask you about your settings?
do you use vbr or cbr
thanks


Re: Point of transparency of Opus 1.2

Reply #41
Nowdays there is no reason to go with CBR or ABR (unless for very particular reason like radio streaming with limited network speeds)

VBR is highly preferred for last 10+ years
VBR is both, more bitrate efficient and higher quality (true for any good lossy codec)

Re: Point of transparency of Opus 1.2

Reply #42
I was curious about this topic as well, and did some testing, I think something that is often overlooked is the quality of the dac, power amp, and speakers. I tested opus 128, 144, and 160, against flac. I quickly noticed that I lost interest in listening to 128, I could clearly hear artifacts, it was annoying to my ears, but the point where I no longer felt confidence that I could readily detect flac vs opus was somewhere between 128 and 160, so I decided to split the difference and go with 144. I think if I really pushed the testing, I might, but not for sure, be able to detect 144 vs 160, but definitely not on a phone, media player, or in a car, no way. So to me, given the relatively small sizes involved, I figured 144 and I'll call it good.

Note, I was using a very well recorded album, with complex sounds, one of the better recorded albums I've come across, wide dynamic range, some acoustic, played very loud, on good speakers, and that's where you could really hear the difference between 128 and 160, but when it got to 144, I was no longer positive I could tell them apart. I wouldn't listen to anything lower for any reason, sd cards are cheap, there's no reason to.

This is so dependent on the dac involved, and the power amp/preamp/speakers, when I used to run old school audio using basic computer audio outs, or various lower end sound cards, there's no way I could hear these differences, so I upgraded my gear until I hit the point where I could, and stopped there (Asus Xonar -> adcom pre -> adcom power -> quality kit speakers). So in theory, I guess with even higher end gear, mine is just decent midlevel stuff, maybe the point where you can hear the difference fairly predictably gets to a higher bitrate.

Maybe if you're using a good Schiit headphone preamp with top end headphones too, but standard computer stuff, a regular phone, I doubt it, particularly not in a car with all its ambient noise. Though I wouldn't want to listen to 128 bit for any length of time.

Plus flac embedded cover images transfer smoothly to opus, with no fuss or bother, unlike oggs, which is a big plus.

Re: Point of transparency of Opus 1.2

Reply #43
I was curious about this topic as well, and did some testing, I think something that is often overlooked is the quality of the dac, power amp, and speakers.

It is generally recommended that you use headphones for testing codecs, not speakers since some kinds of artifacts are going to be less noticeable with speakers.  Most of the time when you see results here, it is going to be headphones.  The DAC usually won't matter, most people are going to have something without audible problems. 

I tested opus 128, 144, and 160, against flac. I quickly noticed that I lost interest in listening to 128, I could clearly hear artifacts, it was annoying to my ears, but the point where I no longer felt confidence that I could readily detect flac vs opus was somewhere between 128 and 160,

Did you actually ABX it?  Should be easy to know when you can tell a difference that way.

Maybe if you're using a good Schiit headphone preamp with top end headphones too, but standard computer stuff, a regular phone, I doubt it, particularly not in a car with all its ambient noise. Though I wouldn't want to listen to 128 bit for any length of time.

High end equipment isn't very important.  You just need headphones with reasonably flat frequency response, as you can interfere with masking if you have too strange a frequency response.  But generally compression artifacts aren't too dependant on the specific equipment as long as it is within reason.

Re: Point of transparency of Opus 1.2

Reply #44
Sorry, I didn't check back on this. A few things. I definitely noted a significant difference between a cheap low end generic D/A and a low end 'audiophile' dac, there was no question about it. I'd also noticed a difference between the built in PC audio chips and a dedicated though low end audio card. Also a good pre-amp made a significant difference. I'm not talking audiophile type snakeoil differences, these were clearly audible, at a relatively low cost, and I stopped getting new gear once I hit that level since the cost/benefit ratio doesn't do it for me after that point. But this is significantly better in my opinion than what many people use, and that's one reason the 'transparency' question is largely meaningless unless the playback tools are listed, the entire chain.

Note that the point of transparency to me is the point where it sounds as good as a flac given your ears, the playback chain, and your surrounding noise levels. While it may be possible to point to much lower bitrates when looking only at raw artifacts, that's to me a largely irrelevant point, something I'd missed initially. It's similar to talking about vinyl playback and noting the presence of pops or clicks, when actually the point is what the non clicking output is doing. It would be obvious that a true artifact, or what I'd say is a defect, is going to be audible at much lower bitrates than a really transparent reproduction. This might be a question of semantics, but when I'm talking about transparency, I'm talking about sounding good, good reproduction, very hard to distinguish from a flac. I'd guess the levels that clearly audible artifacts drift away is going to be significantly lower than the level where the opus sound output grows truly transparent. I know I'm not listening for artifacts, if I were, I could see why earphones or the playback system matter less, since those are relatively gross audio distortions, not lower quality overall audio, which is what I'm listening for. So that's a bit of a red herring, sorry I missed that point. If I hear actual artifacts, then that compression level is so far from usable that I wouldn't even be listening to it in most cases, though as some in this thread noted, you can produce those on certain recordings, but that's not what I'm listening for. Definitely not what I'd equate to the term transparent, which I consider a much higher standard.

I honestly think my speakers are better than my headphones, particularly at very loud volumes, even though I have decent Grado SR80s. But even there, the difference between the Grado's, which are currently the lowest fidelity earphones I'd use, and my older headphones, various mid level ones, is really big, really very significant in my view, and would easily mask the degradation of lower compression levels. And these grado's are lower end headphones when you're looking at the full range, but I'd say definitely in the 'good enough' category. If I had really good headphones, I might prefer them, don't know, but I don't. I don't really use headphones as a rule, just for testing now and then, or checking audio I'm working on for whatever reason. I know for some recent Audacity remastering stuff I was doing on some live recordings, I initially used headphones, which was a huge mistake, when I switched to speakers I started to hear the mistakes I was making much more readily. Could be really good woofers, hard to say.

I've noticed now and then at night clubs, where lazy DJs or sound engineers are running obviously lossy audio directly into the soundboard from their laptop using their laptop builtin audio out jacks for between set music, when that's running over a reasonably good sound system, like a good club will have, that sounds beyond awful, truly terrible, totally non subtle, and it's not from artifacts, it's from using lossy audio over a very  powerful system with real amps and real speakers that really make that painfully obvious. I almost wish I'd asked them what levels they were using, but I doubt they would have known since if they were aware of such things they wouldn't have been pumping lossy audio over a good club system in the first place. So I'd tend to say the better the speakers/amps, the more obvious lossy gets.

Yes, I was ab testing against flacs, since I was redoing my entire lossy collection from ogg 7 encoded to opus, I wanted to determine what point I could no longer reliably hear differences, which as I noted, was pretty clearly around 140ish vbr opus. If I was listening to very well recorded classical or accoustic music, it might be higher, but also as noted, since these opus files will never be played on a real, home/studio type stereo, I didn't consider that point to matter, I just didn't want to have to redo the stuff. Cars are really noisy, the dac on my phone sucks, lol, and I don't have good earbud earphones so 140 seemed a reasonable compromise. I believe on a technical level, if your system can reproduce the good and less good sound, the less good will actually make you tired, since it's not the complete natural sine wave, so your brain has to do more work to construct the music, which is why it's annoying to listen to low bitrate stuff. I never listen to it at home.

What struck me is that once you are putting this stuff through a good dac and a good stereo, things that would pass inspection on lower end playback equipment suddenly doesn't. That's the only way I can explain how something as low as 96 would be considered transparent, the generally degraded quality of less hifi gear in general, not really sure how anyone could determine anything of meaning if the playback chain isn't capable of reproducing the differences in the first place.

To me, if the point is playback over a device or a car, there's a point where good enough certainly is good enough, and better is unlikely to be audible, unless you have a very high end system in the car, and are not using the builtin dac output on a phone or media player. I used to have a car stereo where some of these things would be audible, assuming I had a high end playback/dac chain built into the stereo, but even Paul from PS Audio (I use him as an example because I tend to believe his company makes gear that is way beyond what most people can actually hear as a real difference, and it's expensive stuff, way out of any range I want to be in, so that's a guy really tuned into these allegedly ultrafine differences) said in a video that he thinks the Tesla sound system is really good, but with all the ambient noise when you drive, really good is plenty in a car I think.

The actual real question I'm more interested in is what level of system is needed to actually hear a difference between 24/96 flacs and 16/44.1 or 48, if one can be heard. I know it's higher end than my gear, but I am not willing to state categorically that given much better gear, it would not be audible, since I don't have gear like that (say, a very good DAC running into a preamp running out to concert type Meyer sound speakers, with someone blind A/Bing it. Now that would be interesting...). Obviously, if that can be truly AB tested as audibly different on high enough end gear, the notion  that opus could be 'transparent' on the same gear at such low bit rates would be fairly absurd I think. But that's just an aside, it's something I've wondered about. I guess technically a real waveform analysis of the electrical signals generated by both coming out of a good quality DAC would show if there is a difference. People have better and worse ears too, so there is that as well, that's probably a relatively massive range, from 'golden ears' of top end mastering engineers, to someone raised on mp3s and ears trained to think bad sound is good.

To me questions like transparency of opus vs flac would be more informative if the answers included the playback chain, the normal formats listened to, etc, that helps put the numbers into a more realistic perspective, and would very likely explain some of the lower ranges like 64, 80 etc. And the higher ones in this thread, like 160, 480, etc, given certain types of music etc.

Re: Point of transparency of Opus 1.2

Reply #45
What struck me is that once you are putting this stuff through a good dac and a good stereo, things that would pass inspection on lower end playback equipment suddenly doesn't.

People often assume this, but then struggle to actually observe better discrimination in practice.  I suggest trying ABX of tracks with both rather than assuming an answer.  This is often humbling.