• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

To upsample or not...

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
15,236
Likes
25,072
Hard to believe.
At one time there was an old Arny Krueger file about where he recorded jangling keys at high sample rates and offered the 44.1 khz version from a down sample. I found I could pick out the low rez version using Foobar ABX software. I don't know what software he used. So I made my own jangling key recordings and downsampled and found I could not hear it. I was using the Sox downsampler which is a very good one. I took Arny's hires sample and downsampled it with Sox and I now could not hear the difference. I seem to recall Amir tried my downsampling, and found it much harder to discern than Arny's downsampled version. Though I believe he still could detect it.

BTW, why jangling keys? Because about 3 metal keys on a keyring jangled are actually pretty loud in the 25-35 khz range. Not so loud to us, but most of the energy is ultrasonic. What sounds like a tinkling to us is pretty loud to our dog or cat or mice.
 

kongwee

Addicted to Fun and Learning
Joined
Jan 22, 2022
Messages
593
Likes
146
Upsample to highest you can get for the DAC side will get the lowest floor noise. For wave file, 44.1khz upsampling don't do much. It will normally upsample automatically in DAC. Don't need to increase the file size.
 

BeerBear

Active Member
Joined
Mar 9, 2020
Messages
121
Likes
121
Of course this blind test of 44.1 khz vs 88.2 khz sample rates makes one wonder.
Free download of the paper.

They found some could discern 88.2 khz vs downsampled 44.1 khz using a very good downsampler.
They used Pyramix 6, which doesn't look very good here (v6.2.3): https://src.infinitewave.ca/. Even the newer versions don't look very good.
I'd also like to see some analysis of the files used for comparison and/or line out recordings of the files as they were played back for the test. That could give us some valuable information. But this is more of a suggestion to anyone who's going to attempt a similar test, rather than direct criticism.

Oh and when doing such ABX with Foobar you need to be careful with ReplayGain. Loudness matching is essential, of course, but ultrasonic content can mess with the loudness calculation, at least with EBU R128. Nowadays Foobar downsamples high samplerate content by default for ReplayGain...
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
15,236
Likes
25,072
They used Pyramix 6, which doesn't look very good here (v6.2.3): https://src.infinitewave.ca/. Even the newer versions don't look very good.
I'd also like to see some analysis of the files used for comparison and/or line out recordings of the files as they were played back for the test. That could give us some valuable information. But this is more of a suggestion to anyone who's going to attempt a similar test, rather than direct criticism.

Oh and when doing such ABX with Foobar you need to be careful with ReplayGain. Loudness matching is essential, of course, but ultrasonic content can mess with the loudness calculation, at least with EBU R128. Nowadays Foobar downsamples high samplerate content by default for ReplayGain...
Yes, the key point is when. nothing was done, concurrently recorded native 44.1 was not heard as discernable from 88.2 khz. IOW, 44.1 khz is enough. And hearing recordings in native sample rate is preferable to resampled versions.

PS my faulty memory was they had used iZotope which is very good, not Pyramix which is not so great. Sox or Audacity which uses Sox is much better. The resample I did was in Audacity 2.03 at the time.
 

Nobleman

Member
Joined
Jun 30, 2022
Messages
19
Likes
3
Why would anyone use a higher sample rate particularly just for playback?
For me it was not for improving quality. Not all songs are sampled at 24/192 in fact most of them are 16/44.1 or 48. In my case I moved from Tidal to Apple Music when Apple upgraded to lossless in June last year. But not all tracks are high res 24/192 but I did not want to lose out when I do play a high res album like the Coldplay albums. So I set the Mac mini outputo 24/192 to not miss the opportunity to listen to high red encoded tracks. However this means OSX up samples anything below 24/192...
 

Chrispy

Major Contributor
Forum Donor
Joined
Feb 7, 2020
Messages
3,941
Likes
3,013
Location
PNW
For me it was not for improving quality. Not all songs are sampled at 24/192 in fact most of them are 16/44.1 or 48. In my case I moved from Tidal to Apple Music when Apple upgraded to lossless in June last year. But not all tracks are high res 24/192 but I did not want to lose out when I do play a high res album like the Coldplay albums. So I set the Mac mini outputo 24/192 to not miss the opportunity to listen to high red encoded tracks. However this means OSX up samples anything below 24/192...
No native mode?
 

dougi

Addicted to Fun and Learning
Joined
May 31, 2020
Messages
640
Likes
526
Location
ACT, Australia
Why would anyone use a higher sample rate particularly just for playback?
I had a brief look at optimal sample rates for an aspect of my system in this post. So matching sample rate would seem to have measurable benefits in some cases, but there could also have been measurement error.
 

Chrispy

Major Contributor
Forum Donor
Joined
Feb 7, 2020
Messages
3,941
Likes
3,013
Location
PNW
I'm having the redirect problem again for looking at the post. I guess I'm just stuck on the concept of you can't put lipstick on a pig type thing....artificially increasing sample rate is useless afaik.
 

Chrispy

Major Contributor
Forum Donor
Joined
Feb 7, 2020
Messages
3,941
Likes
3,013
Location
PNW
At one time there was an old Arny Krueger file about where he recorded jangling keys at high sample rates and offered the 44.1 khz version from a down sample. I found I could pick out the low rez version using Foobar ABX software. I don't know what software he used. So I made my own jangling key recordings and downsampled and found I could not hear it. I was using the Sox downsampler which is a very good one. I took Arny's hires sample and downsampled it with Sox and I now could not hear the difference. I seem to recall Amir tried my downsampling, and found it much harder to discern than Arny's downsampled version. Though I believe he still could detect it.

BTW, why jangling keys? Because about 3 metal keys on a keyring jangled are actually pretty loud in the 25-35 khz range. Not so loud to us, but most of the energy is ultrasonic. What sounds like a tinkling to us is pretty loud to our dog or cat or mice.
Yeah somewhat remember the keys thing. But does it improve with upsampling was more my thought.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
15,236
Likes
25,072
I had a brief look at optimal sample rates for an aspect of my system in this post. So matching sample rate would seem to have measurable benefits in some cases, but there could also have been measurement error.
Looks like something in your chain doesn't like 44.1 khz and multiples of it. So in your case you'd likely benefit from even 44 to 48 khz conversion. I had an HDMI converter that split out sound either on RCA jacks or an optical signal. It worked much the same, didn't like 44.1 khz multiples. Was a cheap device, and I'm guessing they only had one actual clock crystal set for 48 khz.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
15,236
Likes
25,072
Yeah somewhat remember the keys thing. But does it improve with upsampling was more my thought.
Improvement would have to mean they sound different. 44 was not distinguishable from 88 in the blind test they did if no resampling occurred. My conclusion is if 88 sounds the same then there is no audible improvement. What you are describing would be taking 44 and upsampling then seeing if it sounded better. I've not done that as far as I recall.

The post by dougi in his thread shows a case where upsampling definitely improves the signal, but that isn't something you can apply to things generally. Some piece of his gear just doesn't handle 44.1 properly.
 

Axo1989

Addicted to Fun and Learning
Joined
Jan 9, 2022
Messages
614
Likes
482
For me it was not for improving quality. Not all songs are sampled at 24/192 in fact most of them are 16/44.1 or 48. In my case I moved from Tidal to Apple Music when Apple upgraded to lossless in June last year. But not all tracks are high res 24/192 but I did not want to lose out when I do play a high res album like the Coldplay albums. So I set the Mac mini outputo 24/192 to not miss the opportunity to listen to high red encoded tracks. However this means OSX up samples anything below 24/192...

Yes I do that for much the same reason. Doesn't make things sound worse and appeals to my numerical OCD (44.1 is an ungainly number). When I'm feeling zen at 3 am it sounds better too, but there's no science in that observation.
 

dc655321

Major Contributor
Joined
Mar 4, 2018
Messages
1,558
Likes
2,111
but I did not want to lose out when I do play a high res album like the Coldplay albums.

Upsampling will not improve Coldplay.


4A8362A62BF256FA9495AD9A1C8A060CC564FE08
 

samsa

Addicted to Fun and Learning
Joined
Mar 31, 2020
Messages
500
Likes
578
It used to be argued that a computer could do a better job upsampling than the DAC chips of the day could do internally. So bypassing the DACs internal upsampling was "worthwhile". It seems to me that whatever limitations there were in silicon of ancient DAC chips no longer exist in modern high quality DAC chips, as found, say, in the OP's Topping D90 or RME ADI-2. So I'm hard-pressed to see where the advantage would lie.

OTOH, CPU cycles are cheap and Roon and HQPlayer offer a range of filters different (not better, just different) from those available internally in the DACs in question. So there's nothing wrong with experimenting with your options.
 

dougi

Addicted to Fun and Learning
Joined
May 31, 2020
Messages
640
Likes
526
Location
ACT, Australia
Looks like something in your chain doesn't like 44.1 khz and multiples of it. So in your case you'd likely benefit from even 44 to 48 khz conversion. I had an HDMI converter that split out sound either on RCA jacks or an optical signal. It worked much the same, didn't like 44.1 khz multiples. Was a cheap device, and I'm guessing they only had one actual clock crystal set for 48 khz.
Yeah, and this is somewhere in my Lyngdorf dpa-1! It resamples everything to 96kHz for the DSP, but not from 44.1 very well it seems.
 

tuga

Major Contributor
Forum Donor
Joined
Feb 5, 2020
Messages
3,260
Likes
3,311
Location
Oxford, England
I have two DACs currently, an RME ADI-2 DAC FS on my desk and a Topping D90 in my main rack. I am currently running the RME with a 'Sharp' filter and the Topping with (as they describe it), the default Short_Delay_Sharp_Roll_Off_Filter. Any thoughts of these choices? These are obviously filter choices rather than upsampling of frequency rates but are also in my mind.

I also have the ADI-2 DAC which I feed with (mostly) Redbook upconverted to DSD256, the RME set to DSD Direct mode, and this sounds "smoother" than feeding the DAC Redbook.
HQPlayer has a load of filter and noise-shaping options that you can test. Some of them will produce subtle differences in "presentation". It will play for 30 minutes at a time in demo mode; you can try it for free.

The "best" measuring DAC to date at ASR also upconverts to high-rate DSD, as do dCS, Linn, Teac/Esoteric, etc.

@Miska has posted comparative measurements of many DACs fed by HQPlayer including the ADI-2 with different settings (NOS, pass-through, upsampled PCM, PCM upconverted to DSD). They're available on CA/AS.
He's discussed the merits of upsampling and DSD in here as well:

 

Snoopy

Addicted to Fun and Learning
Joined
Jul 19, 2021
Messages
665
Likes
347
I upsample everything to 32bit / 192khz in roon.
 

Rednaxela

Addicted to Fun and Learning
Joined
Mar 30, 2022
Messages
631
Likes
684
Location
NL
He's discussed the merits of upsampling and DSD in here as well:
So I clicked the link and read on for 15 posts or so.

Sadly I can’t say I found it all very enlightening. What I read was basically a peeing contest on who claimed what and who should provide proof for which claims.

A waste of time AFAIC.

Did you perhaps link to the wrong post?
 

tuga

Major Contributor
Forum Donor
Joined
Feb 5, 2020
Messages
3,260
Likes
3,311
Location
Oxford, England

Rednaxela

Addicted to Fun and Learning
Joined
Mar 30, 2022
Messages
631
Likes
684
Location
NL
Thank you and sorry for the negativity.

Doesn’t fit a beautiful Sunday morning.
 
Top Bottom