• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Help explain intersample overs, please?

j_j

Major Contributor
Audio Luminary
Technical Expert
Joined
Oct 10, 2017
Messages
2,267
Likes
4,758
Location
My kitchen or my listening room.
I consider myself fairly rational; I interpret this statement as it is highly unlikely that Benchmark has tested every single implementation of a DAC. They are merely trying to prove a point that this is a fairly common problem that is mostly not dealt with by DACs.

Some are much better than others. Some clip, some go completely spastic, one appears to actually wrap-around and creates terrible crunches.
 

cjm2077

Active Member
Joined
Jan 22, 2020
Messages
160
Likes
261
I think they're saying the dac and src chips don't inherently compensate for the issue, so unless an engineer explicitly designs their circuit to deal with it, it will cause issues.
 

Crazy_Nate

Member
Joined
Dec 21, 2019
Messages
52
Likes
68
Some are much better than others. Some clip, some go completely spastic, one appears to actually wrap-around and creates terrible crunches.

I was hoping that @amirm would eventually be able to come up with a test (or multiple, perhaps as part of a suite of tests) to show how DACs start to misbehave with oversampling. Perhaps what levels start to cause improper output and then, how bad does it affect the output (is it audible?). I could imagine how this would get to be a very difficult test case to derive as to not bias the results.

Perhaps this would also inform the audience and coax the manufacturers to do something about it?
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,381
Location
Seattle Area
I was hoping that @amirm would eventually be able to come up with a test (or multiple, perhaps as part of a suite of tests) to show how DACs start to misbehave with oversampling.
I have the test files but not time to implement them in my routine....
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
One thing I don't understand is why everyone mentions Benchmark's illustrations when talking about ISP. I mentioned it in 2006 here for example:

https://forums.dearhoney.idv.tw/viewtopic.php?t=52943&start=18

Yes, I mentioned the "0dBFS+ Levels in Digital Mastering" paper (at 2000) as well, Google it as it has some measurements of some old CD players. All these things happened before Benchmark's illustrations, Death Magnetic and so on.

I think many people watched this video already:

"... you own the volume knob, not the record producer". Yes, but analog volume controls often don't support presets, so it means you need to adjust the volume every time you listen to different music. On the other hand ReplayGain allows these things, plus you can edit the stored gain values if the default loudness algorithm don't fit your taste. Do it once and you can listen at your desired volume level without further intervention.

In hindsight, if people are accustomed to use digital volume management, loudness war would not be able to take root. If it still happens, blame those DAP/smartphones with weak output and require heavy compression and limiting to achieve the desired volume levels without clipping, intersample or not.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
Studios use heavy compressions because they sound good. Not because (or solely because) of loudness. In some genres, it sound better when the sounds blend together, it sound better when it sounds thick, it sounds better if some instruments are heavy compressed, it sounds better if we can hear better sustain of some instruments. It's not the age of choosing clearest radio channel anymore.
 

Hayabusa

Addicted to Fun and Learning
Joined
Oct 12, 2019
Messages
786
Likes
518
Location
Abu Dhabi
So, given that Benchmark has +3.5dB of headroom (gain factor of ~1.496), is this enough headroom?

I've processed the my music collection (all CD) with replaygain at 8x oversample and have a few offenders above 1.5 (wow!). Anyone else seen this occur in their collections? For reference all of my collection at 44.1kHz has peak levels for the albums at 0.999 or 1.0. 90%+ of my collection with 8x oversampling has peaks above 1.0. Three of the offenders are from Metallica: Garage Inc.

Just FYI for those reiterating the Benchmark statement... Intersample Overs in CD Recordings



I consider myself fairly rational; I interpret this statement as it is highly unlikely that Benchmark has tested every single implementation of a DAC. They are merely trying to prove a point that this is a fairly common problem that is mostly not dealt with by DACs.

They say every sample they TESTED...
 

Hayabusa

Addicted to Fun and Learning
Joined
Oct 12, 2019
Messages
786
Likes
518
Location
Abu Dhabi
I have the test files but not time to implement them in my routine....

May be you can do it on some next examples that come in? In that way we can see if at least the current generation of DACS handle these intersample overs correctly.
 

L5730

Addicted to Fun and Learning
Joined
Oct 6, 2018
Messages
667
Likes
434
Location
East of England
Looking at very early CDs, many had plenty of peak headroom. Highest sample peaks at -3dB or even lower were not that uncommon. Sure, some/many had pre-emphasis, but we'll leave that for another thread.

It would seem that the engineers at the beginning of the CD era were thinking of things and actively working to deliver a perfect product.
As time has gone on, however, it's almost as if a new generation of engineers have come along and see 0 dBFS / 100% as the maximum permissible amount (all DAWs to my knowledge permit that) and seem quite happy to slap a limiter on the master bus set to 99.9% FS.

Ignoring pressure from the label/artist to chase loudness for the sake of being competitively loud, how can an engineer deliver something that they don't really know how it's going to perform on other equipment? They may accept the sound in their studio pushing the sample peaks close to 0 dBFS and not hear a problem. But when this is played back on other equipment outside of their studio, the ISP issue becomes an unknown quantity.

Why would a DAC chip designer need to try and accommodate cases which shouldn't really exist, even if they are very common place?
The fault lies with the music industry pressures for loudness, the engineers not understand/not caring and DAW designers not handling this better which would teach the next generation better practices.

Remember old audio cassettes? The ones you could buy in the shops had quite low levels, and thus a pretty poor SNR. When mid 90's cassette desks could handle a pretty decent signal level without clipping, why should they play a shop bought album on tape and only see the bottom half of the VU lights moving? Conversely, what would happen if the album being released on cassette was mastered for playback on more capable HiFi players - all of the folks with cars, walkmans, boom boxes etc are gonna get a very woolly overloaded experience!

Digital has to be treated with the same respect as analogue. For goodness sake, leave some damn headroom!

Saying all the above, if one uses ReplayGain then it's unlikely they'll experience ISP issues all that much and when they do it'd likely be very fast transients which one could argue are too fast to really make an audible difference.
 
Last edited:

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
I consider myself fairly rational; I interpret this statement as it is highly unlikely that Benchmark has tested every single implementation of a DAC. They are merely trying to prove a point that this is a fairly common problem that is mostly not dealt with by DACs.
It is a fairly common "problem" that isn't actually that big a deal. If it really was as bad as Benchmark make it out to be, only the cheapest junk would be affected.

Given that it wouldn't be difficult at all for the DAC chip designers to include some headroom, one wonders whey they don't. I doubt they are all unaware of the phenomenon. Maybe they are ignoring it because that lets them put a slightly higher SNR figure in the spec sheet.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
Looking at very early CDs, many had plenty of peak headroom. Highest sample peaks at -3dB or even lower were not that uncommon. Sure, some/many had pre-emphasis, but we'll leave that for another thread.

It would seem that the engineers at the beginning of the CD era were thinking of things and actively working to deliver a perfect product.
Many early CDs are ADD or AAD anyway (rather noisy), plus DAWs are uncommon in for example, pre 1990. Digital metering was not as precise as DAWs these days, probably some LEDs on a DAT deck or PCM adapters, DASH and the like. Software dynamic processors were simply uncommon.

In fact I have some old CDs have actual 0dBFS clippings, but the tracks themselves have plenty of dynamics, like this:
https://forum.cockos.com/showthread.php?p=1852077#post1852077
I attached an audio sample in that post so please listen to it.

The waveform of the whole track looks like this:
doina.png


Yet it still clipped. So leaving some headroom due to the imprecise nature of metering in that era made perfect sense.
 

L5730

Addicted to Fun and Learning
Joined
Oct 6, 2018
Messages
667
Likes
434
Location
East of England
@bennetng Thank you for the sample, it was quite an obvious change in sound. Just listening on little desktop speakers and mainboard Realtek audio the clipped audio gave feeling as if the action was too low and the string was rattling against the fret wires.

I applied Izotope DeClip to your audio. Threshold -1 dBFS, Gain -6dB, High Quality. Dithered (TPDF) to 24bit from Reaper. The nasty rattle from the clipping is gone, but to be able to do perform this intelligent guesswork reconstruction the gain must be reduced to provide somewhere for the reconstructed waveform to go.

You are correct, of course. In the early days of CDs the DAW as we know it didn't exist. ProTools appeared right at the end of the 80's. Folks were still using analogue VU meters and PPM meters. Leaving headroom was just part of the nature. The difference being that analogue tape would soft-limit by nature, digital would clip transients (or extreme hard limit).
 

Attachments

  • DeClip_-6_T1_HQ_24b_dither.zip
    968.4 KB · Views: 90

cjm2077

Active Member
Joined
Jan 22, 2020
Messages
160
Likes
261
@bennetng Thank you for the sample, it was quite an obvious change in sound. Just listening on little desktop speakers and mainboard Realtek audio the clipped audio gave feeling as if the action was too low and the string was rattling against the fret wires.

I applied Izotope DeClip to your audio. Threshold -1 dBFS, Gain -6dB, High Quality. Dithered (TPDF) to 24bit from Reaper. The nasty rattle from the clipping is gone, but to be able to do perform this intelligent guesswork reconstruction the gain must be reduced to provide somewhere for the reconstructed waveform to go.

You are correct, of course. In the early days of CDs the DAW as we know it didn't exist. ProTools appeared right at the end of the 80's. Folks were still using analogue VU meters and PPM meters. Leaving headroom was just part of the nature. The difference being that analogue tape would soft-limit by nature, digital would clip transients (or extreme hard limit).

The loudness/compression wars also came along about the same time as mp3s becoming common. A good studio producer takes into account how the final music is going to be played back. Not everyone has a 1k+ stereo system they listen to everything on. You have to mix/master for cheap speakers, car radios, at one time you had to account for the limited bandwidth and snr of FM or AM analog broadcasts, and all of a sudden you had to account for the high likelihood that your music would be compressed into a 128 kbps mp3 and listened to on an audio player on the subway. So of course they turned to compression to even out the listening experience across all the formats. That's bad for us hi-fi people, but it was good for the much larger group of people listening to their ipods on the way to work or at the gym.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
Audio (dynamic range) compression and lossy/lossless compression are in no way related.
 

cjm2077

Active Member
Joined
Jan 22, 2020
Messages
160
Likes
261
Audio (dynamic range) compression and lossy/lossless compression are in no way related.

They are different processes, yes. They do completely different things. The point was not about them being the same, but that adding dynamic range compression could enhance the final sound of a file that had been compressed in size. Just like adding compression could enhance playback over a cheap boombox in the 80's. You were also getting many more people listening to mp3 players in noisy environments, like the subway, streets, or at the gym, because they were so small, light, and skip free. So higher dynamic range compression would be beneficial there too. It's bad for me trying to listen to the music at home, but I can understand why they did it at the time.
 

L5730

Addicted to Fun and Learning
Joined
Oct 6, 2018
Messages
667
Likes
434
Location
East of England
The loudness/compression wars also came along about the same time as mp3s becoming common. A good studio producer takes into account how the final music is going to be played back. Not everyone has a 1k+ stereo system they listen to everything on. You have to mix/master for cheap speakers, car radios, at one time you had to account for the limited bandwidth and snr of FM or AM analog broadcasts, and all of a sudden you had to account for the high likelihood that your music would be compressed into a 128 kbps mp3 and listened to on an audio player on the subway. So of course they turned to compression to even out the listening experience across all the formats. That's bad for us hi-fi people, but it was good for the much larger group of people listening to their ipods on the way to work or at the gym.
The popularity of the lossy CODEC (mp3) came about primarily because of illegal and legal file sharing and portable devices. Smaller size for the same music = less heavy on network connections (this was pre- broadband), faster downloads, more songs to store per unit of space (flash storage was in MiB not GiB).

Dynamic compression for the sake of making the song louder was around in the disco era. There was a loudness war of sorts, but the limitations of what vinyl could do prevented levels reaching what we have today on digital mediums. It's funny really, because radio processing made things a general homogeneous loudness and DJ mixers would have allowed some loudness equalisation when cueing up the next single. So I don't really understand why 'cutting a hot record' was even necessary.
Hot clipped garbage was being put on CD in 1999 ion the form of the Red Hot Chili Peppers Californication. Charles Dye admits to shoving a Waves Limiter on the master bus of the ProTools session and just adding something like 6dB of makeup gain to Ricky Martin's 1999 hit Livin' la Vida Loca.
Levels were steadily increasing all the way through the 90's and could be plotted on a graph. It's no surprise that limiter plugins because more available and transparent towards the end of the 90's, and virtually all albums were being recorded and mixed digitally.

..but we are digressing.

With the metering available, quality of converters and such, there are zero reasons why 3 or 6 dB of headroom cannot be baked into the digital file when it is distributed to the consumer, either as a CD or HiRes download. Yes, that reduces the dynamic range by some, but it's already fantastic what is available with just 16 bits.
Turning the digital audio down yourself, either destructively (in a DAW and rendering a new file) or on-the-fly with playback gain will, in most cases create enough headroom to avoid ISP clipping.

However, when we look at something I recently posted somewhere else on this forum, we see that a track looks to have been digitally clipped (flat top waveforms) and then reduced in level, after the fact, to fit on the album. There is clear clear flat-top clipping but now at a reduced gain. I suppose the sound will be more consistent across DACs as they won't necessarily clip internally, but how is that wave being created with the samples in the shape they are? Urgh, yuk.
It's just like the sample provided by @bennetng above. Turn it down 6 dB and it still sounds strange. It's still clipped and sounds like it, it's now just not clipping the DAC.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
They are different processes, yes. They do completely different things. The point was not about them being the same, but that adding dynamic range compression could enhance the final sound of a file that had been compressed in size. Just like adding compression could enhance playback over a cheap boombox in the 80's. You were also getting many more people listening to mp3 players in noisy environments, like the subway, streets, or at the gym, because they were so small, light, and skip free. So higher dynamic range compression would be beneficial there too. It's bad for me trying to listen to the music at home, but I can understand why they did it at the time.

Yes, indeed very valid for most popular music or bands that want 'extremely loud' themselves. Fortunately there are also plenty of good recordings around or with tastefully done dynamic compression.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,422
Likes
4,029
Location
Pacific Northwest
... Anyone else seen this occur in their collections? For reference all of my collection at 44.1kHz has peak levels for the albums at 0.999 or 1.0. 90%+ of my collection with 8x oversampling has peaks above 1.0. Three of the offenders are from Metallica: Garage Inc. ...
I haven't tested every album but I've sampled a few in my collection. The offenders I've seen -- not just intersample overs, but actual clipping -- all happen to be rock / pop recordings. No surprise there, thanks to the loudness wars. I have found exactly 1 classical music recording (in a sample size of thousands) having this problem.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
However, when we look at something I recently posted somewhere else on this forum, we see that a track looks to have been digitally clipped (flat top waveforms) and then reduced in level, after the fact, to fit on the album. There is clear clear flat-top clipping but now at a reduced gain. I suppose the sound will be more consistent across DACs as they won't necessarily clip internally, but how is that wave being created with the samples in the shape they are? Urgh, yuk.
Exactly. That's why using ISP solely to judge mastering quality can be misleading, as well as overly obsessed with some other metrics like DR meter, LUFS and so on. You can have a song with zero ISP yet still sounds clearly clipped.

What's the point of leaving headroom if the waveform itself is already squashed, like these songs in my collection?
imas wasted.png


Compared to these examples, songs normalized to 0dBFS with some ISPs, but with reasonably enough dynamic range and no baked-in clippings, like some of the mid-90s masterings are completely benign.

As for looking for flat tops to verify clipping, no flat tops does not mean something is not clipped. The flat tops can be easily eliminated visually with something like a high pass filter at 10Hz, or simply ran a clipped song through a cassette deck like this:
https://www.audiosciencereview.com/...kamichi-dragon-cassette-deck.5595/post-124666
 
Top Bottom