Upsampling, up-converting and oversampling...confused

[semente]

My D/AC uses the PCM1795 chip which, if I understand the datasheet correctly, has an "8x Oversampling Digital Filter".

 

But according to TEAC's literature, "All digital input signals are converted to 192kHz via non-integral up-sampling, dramatically reducing jitter and allowing consistent D/A conversion based on the high-precision internal clock. ASRC conversion is used to produce a natural, 192kHz high-sampling rate digital audio signal."

 

Does this mean that the original sample rate is always converted to 192kHz and then oversampled by 8x before filtering?

 

Would it perform better if I upsampled my Redbook files to 192kHz on the player?

 

I'm a bit confused about all this...

 

R

[wakibaki]

I looked at the PCM1795 data sheet. There's no ASRC, so this must be being done prior to the DAC by another chip. I was unable to identify which TEAC DAC you have, but from your quote, it would appear that all input is converted to 192kHz (by ASRC) and then routinely upsampled *8 in the PCM1795 to facilitate filtering.

 

If you upsample your files prior to sending them to the DAC you may circumvent the ASRC, which you might consider desirable, but you might consider it undesirable, and anyway I don't know for sure, but...

 

...don't admit to being confused or somebody may sell you something.

[audiventory]

In general case "on fly" (real-time) algorithms strongly restricted by computation resources. Especially for hardware.

 

Quality depend on all components as integrated system "file" - "player" - "DAC". Also amp, speaker, room.

 

For each case need make measurements. At least make hearing comparing for different sampling rates and different places (offline file/player/DAC) of resampling.

[semente]

I looked at the PCM1795 data sheet. There's no ASRC, so this must be being done prior to the DAC by another chip. I was unable to identify which TEAC DAC you have, but from your quote, it would appear that all input is converted to 192kHz (by ASRC) and then routinely upsampled *8 in the PCM1795 to facilitate filtering.

 

If you upsample your files prior to sending them to the DAC you may circumvent the ASRC, which you might consider desirable, but you might consider it undesirable, and anyway I don't know for sure, but...

 

...don't admit to being confused or somebody may sell you something.

 

 

Thanks.

[semente]

In general case "on fly" (real-time) algorithms strongly restricted by computation resources. Especially for hardware.

 

Quality depend on all components as integrated system "file" - "player" - "DAC". Also amp, speaker, room.

 

For each case need make measurements. At least make hearing comparing for different sampling rates and different places (offline file/player/DAC) of resampling.

 

Amplifier, speakers and room are constant so I guess it's down to the D/AC's SRC performance vs. music player software on-the-fly SRC (iZotope, SoX, etc.).

I am using an old MBP (2.26 GHz Core 2 Duo, 64-bit, 8GB RAM) solely for this (music playback) which should be powerfull enough, I guess.

 

R

[Jud]

If the original ASRC is to 192kHz, then I doubt the chip is actually doing 8x oversampling. If it is like virtually all other chips, what it does is oversample to "8x" rates, meaning 352.8 (if it were being fed 44.1 sample rate or multiples) or 384 (48kHz sample rate or multiples). Since it's receiving 192kHz from the ASRC chip, it is very likely just doubling that to 384kHz.

 

That internal doubling step is built in and will always happen. What you are able to change is that you can upsample externally inline or offline to 192kHz and eliminate the filter in the ASRC chip. The result will then be the interaction between your initial upsampling and the filter in your DAC chip that's applied during the doubling step.

[Audio_ELF]

I looked at the PCM1795 data sheet. There's no ASRC, so this must be being done prior to the DAC by another chip.

There is according to -- http://www.ti.com/lit/ds/symlink/pcm1795.pdf Look at Page 6 (the block diagram).

[Jud]

There is according to -- http://www.ti.com/lit/ds/symlink/pcm1795.pdf Look at Page 6 (the block diagram).

 

Might be spam, Eloise (faulty, perhaps 1-pixel?, images).

[semente]

If the original ASRC is to 192kHz, then I doubt the chip is actually doing 8x oversampling. If it is like virtually all other chips, what it does is oversample to "8x" rates, meaning 352.8 (if it were being fed 44.1 sample rate or multiples) or 384 (48kHz sample rate or multiples). Since it's receiving 192kHz from the ASRC chip, it is very likely just doubling that to 384kHz.

 

That internal doubling step is built in and will always happen. What you are able to change is that you can upsample externally inline or offline to 192kHz and eliminate the filter in the ASRC chip. The result will then be the interaction between your initial upsampling and the filter in your DAC chip that's applied during the doubling step.

 

Right, I'll try upsampling to 192kHz inline then...

 

Let's see if I can find a thread about iZotope settings.

 

Cheers,

Ric

[Ajax]

Might be spam, Eloise (faulty, perhaps 1-pixel?, images).

 

The images looked OK to me.

[audiventory]

I am using an old MBP (2.26 GHz Core 2 Duo, 64-bit, 8GB RAM) solely for this (music playback) which should be powerfull enough, I guess.

R

 

It may be no mater of CPU performance. Algorithm may be restricted by manufacturer for capability with weak systems. Also some hi-res files has significant noise during playback at some software and hardware. Will noise or not also depend on used SRC and audio processing algorithms.

[Jud]

The images looked OK to me.

 

Yeah, I didn't initially see Eloise was quoting wakibaki, because in fact there were two comments with 1-pixel "images" in the thread repeating word for word what wakibaki had said right above her response at the time I made my comment about spam. And those comments did turn out to be spam, it seems, because they're gone now.

[Jud]

Right, I'll try upsampling to 192kHz inline then...

 

Let's see if I can find a thread about iZotope settings.

 

Cheers,

Ric

 

More than you ever wanted to know....

 

http://www.computeraudiophile.com/f11-software/izotope-sample-rate-convertor-15352/

[semente]

More than you ever wanted to know....

 

http://www.computeraudiophile.com/f11-software/izotope-sample-rate-convertor-15352/

 

 

Hmmmm, 19 pages of that...thanks.

 

R

[Jud]

Hmmmm, 19 pages of that...thanks.

 

R

 

Yeah, I know - pain to wade through. I'll try, sometime today or tomorrow, to post a comment here with about four different sets of iZotope settings, and a brief explanation of what the various settings do (unless you are already familiar with the latter).

[semente]

Yeah, I know - pain to wade through. I'll try, sometime today or tomorrow, to post a comment here with about four different sets of iZotope settings, and a brief explanation of what the various settings do (unless you are already familiar with the latter).

 

That would be great, thanks.

 

Since the resulting file will be oversampled and filtered by the D/AC, why do I need to filter when converting the sample rate from 44.1 to 196kHz?

 

R

[audiventory]

Any resampling give aliases (mirroring of spectrum)

 

How Convert Sample Rate. Oversampling

 

How Convert Sample Rate. Divide. Downsampling

[semente]

Any resampling give aliases (mirroring of spectrum)

 

How Convert Sample Rate. Oversampling

 

How Convert Sample Rate. Divide. Downsampling

 

I see. Thanks.

 

R

[wakibaki]

There is according to -- http://www.ti.com/lit/ds/symlink/pcm1795.pdf Look at Page 6 (the block diagram).

 

I don't intent to be stubborn over this, but that is where I looked, and there's no mention of ASRC in the whole document. Of course I visually scanned it very quickly, but I did also search the entire document for ASRC, asynchronous, conversion.

 

I'm happy to accept that the up sampling might be <8* for 192 input, I thought this might be the case but I saw little point in elaborating.

 

I was hoping to get the TEAC product designation.

 

w

[Audio_ELF]

I don't intent to be stubborn over this, but that is where I looked, and there's no mention of ASRC in the whole document. Of course I visually scanned it very quickly, but I did also search the entire document for ASRC, asynchronous, conversion.

 

I'm happy to accept that the up sampling might be <8* for 192 input, I thought this might be the case but I saw little point in elaborating.

 

I was hoping to get the TEAC product designation.

 

w

Sorry your right... I registered you were saying there was no built in over sampler. Any ASRC would be being done by some form of external DSP.

[semente]

I don't intent to be stubborn over this, but that is where I looked, and there's no mention of ASRC in the whole document. Of course I visually scanned it very quickly, but I did also search the entire document for ASRC, asynchronous, conversion.

 

I'm happy to accept that the up sampling might be <8* for 192 input, I thought this might be the case but I saw little point in elaborating.

 

I was hoping to get the TEAC product designation.

 

w

 

This is it:

 

UD-H01 | TEAC

 

R

[wakibaki]

It looks like an outstanding piece of kit.

 

It says it supports asynchronous USB. Which means that it can use its own internal clock for playback.

 

Then it goes on to say that it converts everything to 192k using ASRC. That must be to accommodate 44k1 sample rates, since 48 is a divisor of 192, and if they're both running from the same clock, then strictly speaking ASRC is not required. Well, it's not required in any case, it's just a sample rate conversion even from 44k1, it's just more complicated. So I'm a bit confused, but you can't expect an expert write-up from a sales department for something like this.

[semente]

It looks like an outstanding piece of kit.

 

It says it supports asynchronous USB. Which means that it can use its own internal clock for playback.

 

Then it goes on to say that it converts everything to 192k using ASRC. That must be to accommodate 44k1 sample rates, since 48 is a divisor of 192, and if they're both running from the same clock, then strictly speaking ASRC is not required. Well, it's not required in any case, it's just a sample rate conversion even from 44k1, it's just more complicated. So I'm a bit confused, but you can't expect an expert write-up from a sales department for something like this.

 

The question is whether it would be best to upsample to 192kHz on-the-fly or to feed the D/AC with the original 44,1kHz sample rate.

 

R

[audiventory]

The question is whether it would be best to upsample to 192kHz on-the-fly or to feed the D/AC with the original 44,1kHz sample rate.

 

R

 

Exact answer only via measurements. Alternative (enough subjective) hearing test. Nobody, except manufacturers of DAC and audio player software, know how work it inside (a lot of details).

[semente]

Exact answer only via measurements. Alternative (enough subjective) hearing test. Nobody, except manufacturers of DAC and audio player software, know how work it inside (a lot of details).

 

I am always in favour of measurements but I don't think I have the equipment for that.

 

R