Upsampling 16/44.1 collection a good idea?

[mike7877]

I've been thinking of upsampling my 44.1kHz collection to 88.2, 176.4kHz, 352.8, or even 705.6kHz for playback...

We're all familiar with the noise being filtered above 22.05kHz when it's being rendered by a DAC. Depending on the filter you choose, you can eliminate (effectively) all of it, or, you can allow the rendering artifacts above 22.05kHz up to about 28kHz, variously attenuated as illustrated above.

I assume that when you convert a 44.1kHz file to 88.2kHz, observing the output with a spectrum analyzer there would be effectively no energy between 22.05 and 44.1kHz. And likewise, when the new rate is 705.6kHz, there's no energy between 22.05kHz and 352.8kHz...

Although my amplifier is flat past 100kHz and I don't know where its rolloff begins, I do know my DAC's -3dB point should be somewhere between 70 and 75kHz (from its 40kHz being spec being -1dB SE, 0.8dB BAL), so whatever might come out of it above 352kHz (or even 176), is probably pretty attenuated.

Is this a good idea to resample my files? Something I'd really like to be able to do is apply the slowest 88.2kHz filter when a resampled 705.6kHz file is being played back... But, alas, that's impossible with my DAC (and most DACs... but that's no consolation!)
\Thoughts?

Also, has anyone else contemplated using slower filters designed for half the sample rate of a high sample rate file (ie. >= 88.2kHz) for playback?

 

[Chrispy]

I can't imagine wasting my time like this. You can hear some sort of difference?

 

[Brian Hall]

I've been thinking of upsampling my 44.1kHz collection to 88.2, 176.4kHz, 352.8, or even 705.6kHz for playback...

We're all familiar with the noise being filtered above 22.05kHz when it's being rendered by a DAC. Depending on the filter you choose, you can eliminate (effectively) all of it, or, you can allow the rendering artifacts above 22.05kHz up to about 28kHz, variously attenuated as illustrated above.

I assume that when you convert a 44.1kHz file to 88.2kHz, observing the output with a spectrum analyzer there would be effectively no energy between 22.05 and 44.1kHz. And likewise, when the new rate is 705.6kHz, there's no energy between 22.05kHz and 352.8kHz...

Although my amplifier is flat past 100kHz and I don't know where its rolloff begins, I do know my DAC's -3dB point should be somewhere between 70 and 75kHz (from its 40kHz being spec being -1dB SE, 0.8dB BAL), so whatever might come out of it above 352kHz (or even 176), is probably pretty attenuated.

Is this a good idea to resample my files? Something I'd really like to be able to do is apply the slowest 88.2kHz filter when a resampled 705.6kHz file is being played back... But, alas, that's impossible with my DAC (and most DACs... but that's no consolation!)
\Thoughts?

Also, has anyone else contemplated using slower filters designed for half the sample rate of a high sample rate file (ie. >= 88.2kHz) for playback?

All you will accomplish is making the files bigger. Unless you have a really bad DAC, it will already filter out the noise. You don't need to do anything to help it.

 

[boxerfan88]

Upsampling files isn’t really worth doing, it’s better to keep the original content intact.

On the other hand, if you prefer to listen at higher sampling rates, why not use a real-time upsampling filter/DSP during playback?

 

[sam_adams]

Ripped from a CD at 16/44.1:

Ripped from a CD at 16/88.2:

There's nothing in the source material above 22KHz but empty bits. Upsampling to anything higher just means more empty bits and larger file sizes for no gain in fidelity.

 

[Keith_W]

You would be wasting a lot of disc space for very marginal benefit.

If you really want to do it, why not consider a real time upsampler? Most music players can already upsample PCM (e.g. JRiver). But if you want more control, you could use HQPlayer. For this to work, you would have to route audio through HQPlayer. I am not familiar with the new version so I am not sure if Miska has included an ASIO input. I know it is possible to build another PC running a dedicated HQPlayer OS (i.e. HQPlayer Embedded) and send audio to it for upsampling. I just don't know if it is possible on the same PC.

I have tested the output of my DAC when fed different sampling rates using a trial version of HQPlayer. The result shows a slight reduction in jitter and marginally improved SINAD (from -111 to -114dB if I recall) depending on whether DSD or PCM was chosen.

But I am going to go with the other replies: it's not worth doing. If your intention is to hear an audible improvement, I guarantee you won't hear one. But if your intention is to feel good about miniscule but inaudible improvements, then go for it.

 

[MaxwellsEq]

I wondered the same thing (in my case, I wondered if I was missing anything) and concluded it's not worth it:

What am I missing - up-sampling and DSD conversion?

I see people up-sampling 44k/16 PCM CD quality source files to, e.g. 96k or 192k. Or even converting 44k/16 PCM files to DSD. But I don't get it, what's the benefit? PCM up sampling I can see one possible benefit from up-sampling to 192k, that any in-band artefacts from the playback DAC...

audiosciencereview.com

 

[AnalogSteph]

+1 for realtime upsampling if you need it. The SoX resampler DSP for Foobar2000 for example is highly optimized and generates barely any load even on a 15-year-old CPU. I am using that to tame a crummy old IDT HDA sound chip in a Dell Latitude E6330 (I suspect either excessive periodic filter ripple or poor resampling)... should really try measuring the darn thing now that I bought the requisite cables.

 

[digitalfrost]

I resample everything I have to 48khz.

Higher sample rates can be harmful https://people.xiph.org/~xiphmont/demo/neil-young.html

I've personally had issues with high frequency noise causing audible problems, so I actually want to get rid of it. I choose 48khz because it offers some more headroom compared to 44.1 and movies/games/web are often in 48khz. I use SoX resampler in foobar2000 to upsample during playback.

I would not store resampled files on the disc, upsamling is fast and takes surprisingly low CPU so you can just do it via DSP.

 

[Lambda]

You don't need to do this before you can do it in real time with no noticeable delay

 

[Rednaxela]

Would you do it with your digital photos?

 

[Tell]

Offline upsampling should in theory be better than realtime, but seeing that even that won't give any audible upgrades why even bother with any kind of upsampling?
Personally I just set my Windows machine to 96khz since that's what my MiniDSP is working at, but I doubt it matters anyway.

 

[Vincent Kars]

Most DAC's are over- or up-sampling.
My veteran Benchmark DAC1 run internally at 126 kHz.
Cambridge advertises one of their DAC's to upsample to 382.
Don't see how things improve by being up-sampled on the PC first and than being re-sampled again by the DAC.

 

[Lambda]

Don't see how things improve by being up-sampled on the PC first and than being re-sampled again by the DAC.

Because in the PC you can chooses what ever algorithm you want and you have infinity more computation memory and time available.

Sinc filter - Wikipedia

en.wikipedia.org

As the sinc-in-time filter has infinite impulse response in both positive and negative time directions, it is non-causal and has an infinite delay (i.e., its compact support in the frequency domain forces its time response not to have compact support meaning that it is ever-lasting) and infinite order (i.e., the response cannot be expressed as a linear differential equation with a finite sum). However, it is used in conceptual demonstrations or proofs, such as the sampling theorem and the Whittaker–Shannon interpolation formula.

Sinc-in-time filters must be approximated for real-world (non-abstract) applications, typically by windowing and truncating an ideal sinc-in-time filter kernel, but doing so reduces its ideal properties. This applies to other brick-wall filters built using sinc-in-time filters.

So theoretically you can of cause come closer to ideal with an PC and the proper algorytem...
Dose it matter to the ear? Probably not.

But still most DAC Reviews here include a test of the DACs filter Pass band and attenuation

 

[danadam]

I assume that when you convert a 44.1kHz file to 88.2kHz, observing the output with a spectrum analyzer there would be effectively no energy between 22.05 and 44.1kHz.

That depends on the filter used by the program that did the conversion. For example, here's what default settings in SoX and FFmpeg do:

sox IN OUT rate 88200
ffmpeg -i IN -ar 88200 OUT

zoom in:

Also, has anyone else contemplated using slower filters designed for half the sample rate of a high sample rate file (ie. >= 88.2kHz) for playback?

I'm not sure what you mean exactly, but if it is applying low-pass filter on hi-res files before sending them to DAC then I think @dualazmak is doing something like that. IIRC he uses low-pass filter at 25 kHz.

 

[mike7877]

I can't imagine wasting my time like this. You can hear some sort of difference?

Ever hear of batch conversion?

 

[Hayabusa]

Ever hear of batch conversion?

That was not his main concern, if there is no audible difference, its a waste of time no matter how efficiently you do the conversion.

 

[mike7877]

Ripped from a CD at 16/44.1:

View attachment 362370

Ripped from a CD at 16/88.2:

View attachment 362372

There's nothing in the source material above 22KHz but empty bits. Upsampling to anything higher just means more empty bits and larger file sizes for no gain in fidelity.

Good to see no upsampling artifacts there (like someone once told me there would be). You know all the filters everyone's always wondering which to use? Well, this seems to circumvent that! Just resample to 750 and you get analog signal up to 22kHz and silence up to 325kHz. No reverse images (potentially) coming back and (potentially) causing (potentially) bad things potentially, nothing!

 

[mike7877]

That was not his main concern, if the is no audible difference, its a waste of time no matter how efficiently you do the conversion.

Wasting 30 seconds is still a bit different than wasting dozens of hours...

If I can accomplish the same thing by just resampling on the fly with foobar2000, that'd be great! I wonder if there's a plugin which allows you to switch to upsampling to a multiple of 44.1kHz when the source is 44.1 (or 88.2 etc), and a multiple of 48kHz if it's 48k (or above, up to but not including 768kHz).

I get that artifacts should be well below the threshold of hearing when converting 44.1 to, say, 192 when using a proper method, I just would like peace of mind, knowing things are going to work the best they can at every step.

For some reason a lot of people here are big on justification "why you wanna do that?"

I want to keep every step in the chain working as perfectly as possible, so that at the end of the chain, all these "barely perceptible" and "questionable if perceptible" effects are minimized to the maximum degree, so that when summed their total is as close to imperceptible as possible. That is always my justification, and it should be every hi-fi enthusiast's. I don't like getting into arguments with people saying "you can't hear that!" A lot of times, I can, in fact, hear differences. Quite easily, actually. But nobody believes what I say and demand double-blind testing, which, if I did, the results would then be questioned...

I think that whenever anyone is doing anything to improve audio fidelity, the reason should be assumed to be: "I want to keep every step in the chain working as perfectly as possible, so that at the end of the chain, all these "barely perceptible" and "questionable if perceptible" effects are minimized to the maximum degree, so that when summed their total is as close to imperceptible as possible" with "no step too small!" stamped underneath

 

[Hayabusa]

Good to see no upsampling artifacts there (like someone once told me there would be). You know all the filters everyone's always wondering which to use? Well, this seems to circumvent that! Just resample to 750 and you get analog signal up to 22kHz and silence up to 325kHz. No reverse images (potentially) coming back in, nothing!

The red curve has artifacts. Its filter only attenuates by +/20dB at 24KHz . This means a 20KHz signal also mirrors at 24Khz at the level.