I need to learn about up-conversion and digital filters

[wgscott]

I just purchased a Teac NT-503 DAC

 

It allows x1 (or off), x2, x4, x8, and DSD up-conversion. The owner's manual give the following conversion matrix:

 

 

The following digital filters are also available:

 

 

I've read that up-converting from PCM to DSD introduces noise, but that some people regard this as euphonic.

 

Apart from that, I think I understand up-conversion to some extent, but the filters have me scratching my head. I have no idea what any of them do.

 

To further complicate issues, I use Dirac room-correction filters as an AU plug-in. Currently, I have sets for 44.1, 88.2, 48 and 96 kHz. The newer version of the software will enable me to create 176.4 and 192 kHz. So let's assume I can make all those.

 

The DAC will accept up to 384kHz PCM input.

 

Am I best off using Audirvana to up-sample? Although Audirvana can upsample to 384, I will only have Dirac room correction filters up to 192kHz.

 

So, should I have Audirvana up-sample everything to 176.4 or 192 kHz, apply the Dirac Room correction filter, and then pass the corrected 176.4 or 192 kHz data stream to the DAC for further up-sampling, and if so, should I use further PCM up-sampling, or DSD up-sampling, and with what filters?

 

In addition, when I go to create the new room correction filters, I what up-sampling and filter settings would I use on the DAC? Does it matter if the room correction is only applying a correction curve say between 30Hz an 18kHz? Also, the correction curve imposes a rather sharp cut-off above (eg) 18kHz. How does this effect what choice of filter to use in the DAC?

 

I bought this thing because I thought it would be enlightening to play around with these parameters, but now I realize I know next to nothing about where even to start.

[beerandmusic]

As you know, i am a die hard DSD fan. HOWEVER, hardware upconverting to me doesn't usually sound better, or I haven't studied it hard enough to say one way or the other, but as much as I love dsd, i usually keep the upconvert off on the teac...but then again, i usually only listen to native dsd anymore. The software upconvert seems to sound better to me than the hardware upconvert.

 

Korg software will allow you to upconvert to 512, which to me sounds better than hardware upconvert....but best of all to acquire true native dsd.

 

This is just me...but i am sure others have different opinions.

[wgscott]

I only have 4 DSD albums, and to use Dirac, I need to play back PCM. But I think if I just turn off all upsampling and filtering on the DAC, I need to use Audirvana (or other software) to do the upsampling and filtering to avoid problems.

 

For completeness, here is what I am currently doing with Audirvana:

 

[jabbr]

Look at HQPlayer which can load a room correction kernel. Upsampling PCM to DSD256/DSD512 sounds fantastic. Your DAC does really well with this approach (it's going to SDM your PCM anyways if you don't do it).

[beerandmusic]

just out of curiosity, what settings do the professional recordings use for sacd?

do they somehow remaster the original analog recordings and so they don't really upsample?

[jabbr]

just out of curiosity, what settings do the professional recordings use for sacd?

do they somehow remaster the original analog recordings and so they don't really upsample?

 

Considering quantum effects they almost certainly down sample

 

What sample rate do you consider the "original analog recordings" to be at? Since you've said that you don't like upsampling, is this for all the DACs you have listened to which do SDM? Are you then concerned about which brick wall filters your DACs are using? Do you know? If not, what is the basis for your opinion?

[ecwl]

It's been many years and lots of reading and I think I'm still trying to learn how DACs really work.

 

The short answer to OP is whatever sounds good to you, just go with it.

 

The long answer, despite some interesting comments so far, is that I think for the Teac DAC, it can play DSD streams "natively" although I've already had some further discussions with people whether that's totally true. The DAC chip is a multi-bit SDM chip. So I don't know what frequency the AK4490 DAC chip runs at but basically almost all multi-bit SDM chips for simplificiation are running at say a 4-bit 11MHz (I'm guessing a little bit at the number and it obviously varies from chip to chip). So regardless of what PCM files you're sending the DAC, it'll eventually be converted to say 4-bit 11MHz before analog sound comes out. That's why these upsampling options are so confusing because the DAC will always upsample, regardless of what option you choose. So even if you choose not to upsample and play native at 16-bit/44kHz, what you're telling the DAC to do is to send the 16-bit/44kHz stream to the DAC chip and let the DAC chip upsample/noise shape, etc. What the options Teac is offering is that if you choose 8fs, you're using Teac's chosen algorithm to upsample to 24-bit/352kHz first before sending the data to the DAC chip and then the DAC chip will continue to upsample to say 4-bit/11MHz to play the music back. So you're essentially choosing whether the AK4490 DAC chip does the upsampling, the Teac uses its own algorithm for upsampling. By using software to upsample first, you're creating another method for upsampling. And in the case where you can choose to convert to DSD at whatever sample rate you want, you're once again telling Teac to use its own algorithm to convert 16/44 to DSD64/128/256/512, before letting the DAC chip play the DSD "natively".

 

Theoretically, with a powerful computer and appropriate software, you can upsample "better" than what the DAC chip and the Teac's internal hardware upsampling can do. But most software uses a pretty standard upsampling algorithm (e.g. JRiver) which is probably why most people would say, I've tried software upsampling but I prefer my DAC doing it because the DAC's upsampling algorithm is usually designed to sound good with the DAC chip. But then you also hear people say they prefer to use XXHighend or HQPlayer to upsample than using their DAC because those software generally use a lot of your CPU power to upsample, compared to what most DACs can put into their hardware chips for upsampling. That's probably why there is such a wide variety of opinions on what is "best" upsampling.

 

The biggest issue for OP is that he has to run Dirac first. Let's say we are dealing with a 24/96 Dirac signal at the end. And then he can decide how he wants to further process the Dirac signal. Would it sound better if the computer converts the Dirac 24/96 signal to DSD512? Would it sound better if the Teac DAC converts the signal do DSD with its internal hardware? Would it sound better without any upsampling and the AK4490 DAC chip just gets the Dirac 24/96 signal? Who knows?

 

On top of that, people have different preferences to sound so some people prefer one type of upsampling/digital filters and others prefer another. If I were to guess, the "best" sound will probably be for the Dirac signal to be upsampled by a "good" algorithm in the computer too to 24-bit/352kHz/384kHz but who knows whether the computer can handle it.

 

But this is all speculative. So it's probably best that OP experiments. Or just not worry about it.

[tne]

Look at HQPlayer which can load a room correction kernel.

 

Unfortunately the room corrections generated by Dirac Live are proprietary and cannot be used within HQPlayer currently. Therefore Dirac Live takes the output from HQPlayer and applies the room corrections before output to next device.

[Jud]

just out of curiosity, what settings do the professional recordings use for sacd?

do they somehow remaster the original analog recordings and so they don't really upsample?

 

Analog "sample rate" is infinite, so any digital format sample rate is downsampling (or "decimation").

[wgscott]

Even just getting some help with understanding the significance and the differences between these would be really helpful:

 

 

PCM digital filter

Use to set the digital filter when receiving PCM format signals.

 

OFF

The digital filter is disabled.

 

FIR SHARP

An FIR filter with a steep roll-off is used to sharply cut signals out- side the audio band.

 

FIR SLOW

An FIR filter with a slow roll-off is used to gently cut signals out- side the audio band.

 

SDLY SHARP

A short delay filter with a steep roll-off is used to sharply cut sig- nals outside the audio band.

 

SDLY SLOW

A short delay filter with a slow roll-off is used to gently cut signals outside the audio band.

 

o When receiving signals at 352.8 kHz or 384 kHz, the digital filter will be disabled during playback regardless of this setting. (Why?)

 

 

 

DSD digital filter

Use to set the digital filter when receiving DSD format signals.

 

CUTOFF 50kHz

50kHz cutoff frequency

 

CUTOFF 150kHz

150kHz cutoff frequency

[Jud]

I just purchased a Teac NT-503 DAC

 

It allows x1 (or off), x2, x4, x8, and DSD up-conversion. The owner's manual give the following conversion matrix:

 

[ATTACH=CONFIG]25141[/ATTACH]

 

The following digital filters are also available:

 

[ATTACH=CONFIG]25142[/ATTACH]

 

I've read that up-converting from PCM to DSD introduces noise, but that some people regard this as euphonic.

 

Horse hockey. Competent DSD conversion will move the noise high enough your dog should have real problems hearing it.

 

Apart from that, I think I understand up-conversion to some extent, but the filters have me scratching my head. I have no idea what any of them do.

 

Don't worry about it. Just switch between them at leisure (I wouldn't bother with trying to consciously A/B - that way lies stress rather than enjoyment IMO; listen to entire songs or multiple tracks). Start without Dirac, then include Dirac if you find you can't stand the sound without it. See whether you develop a preference.

 

To further complicate issues, I use Dirac room-correction filters as an AU plug-in. Currently, I have sets for 44.1, 88.2, 48 and 96 kHz. The newer version of the software will enable me to create 176.4 and 192 kHz. So let's assume I can make all those.

 

The DAC will accept up to 384kHz PCM input.

 

Am I best off using Audirvana to up-sample? Although Audirvana can upsample to 384, I will only have Dirac room correction filters up to 192kHz.

 

So, should I have Audirvana up-sample everything to 176.4 or 192 kHz, apply the Dirac Room correction filter, and then pass the corrected 176.4 or 192 kHz data stream to the DAC for further up-sampling, and if so, should I use further PCM up-sampling, or DSD up-sampling, and with what filters?

 

If working with Dirac, I'd suggest trying initially with the DAC's internal PCM filtering switched off and A+ set to upsample to 176.4 or 192. Some upsampling settings you can try/adapt: http://www.computeraudiophile.com/f11-software/izotope-sample-rate-convertor-15352/index12.html#post376630

 

In addition, when I go to create the new room correction filters, what up-sampling and filter settings would I use on the DAC? Does it matter if the room correction is only applying a correction curve say between 30Hz an 18kHz? Also, the correction curve imposes a rather sharp cut-off above (eg) 18kHz. How does this effect what choice of filter to use in the DAC?

 

I'd suggest trying both the DAC's internal filters and software upsampling without Dirac first before moving on to greater complexity. I frankly don't understand much if at all about interactions between filters, so when it comes to that someone like Miska, Yuri of Audiophile Inventory, or Alexey Lukin of iZotope will have to help.

 

I bought this thing because I thought it would be enlightening to play around with these parameters, but now I realize I know next to nothing about where even to start.

 

My suggestions:

 

- To simplify things, try without the Dirac filters at first. Again, don't A/B; listen at leisure. You own this thing, so you've got time.

 

- Try the internal filter settings, using RedBook material as input. (Higher resolution input will require less of the internal filtering, so you may as well use the input that maximizes any audible differences.)

 

- Next shut off the internal filtering and try upsampling with Audirvana, then HQPlayer. Use the max upsampling for both (PCM 352.8 or 384 for Audirvana; DSD256 I believe for HQPlayer, unless your computer hiccups, in which case use the -2s interpolation filter versions and back off to DSD128). Again, use RedBook material to highlight any audible differences.

 

- See if you've got a preference. Then maybe it'll be time to come back for further discussion about use of Dirac (or some other room correction software) with that.

[wgscott]

Thanks, Jud.

 

Also, I just found this: Archimago's Musings: MEASUREMENTS: Digital Filters and Impulse Response... (TEAC UD-501)

 

Sometimes a simple picture is worth a lot:

[Jud]

Thanks, Jud.

 

Also, I just found this: Archimago's Musings: MEASUREMENTS: Digital Filters and Impulse Response... (TEAC UD-501)

 

Sometimes a simple picture is worth a lot:

[ATTACH=CONFIG]25148[/ATTACH]

 

If you go on to look through the rest of the article, the graphs show the classic straightforward ringing versus aliasing tradeoff from the linear phase filters, along with the post-ringing from the "secret" minimum phase filter.

 

Have a look at SRC Comparisons and the iZotope 64-bit SRC charts there.

[wgscott]

Also, the accompanying THD measurement:

 

Both figures are from: Archimago's Musings: MEASUREMENTS: TEAC UD-501 PCM Performance (Part 2) [updated June 23, 2013]

 

These are the FIR filters from the UD-501. I am making the assumption that the NT/UD-503 will behave similarly.

[mansr]

I just purchased a Teac NT-503 DAC

 

It allows x1 (or off), x2, x4, x8, and DSD up-conversion. The owner's manual give the following conversion matrix:

 

[ATTACH=CONFIG]25141[/ATTACH]

 

The following digital filters are also available:

 

[ATTACH=CONFIG]25142[/ATTACH]

 

I've read that up-converting from PCM to DSD introduces noise, but that some people regard this as euphonic.

 

DSD has ultrasonic noise, but even at DSD64 it's well outside the audible band. If you look at a plot of the noise level by frequency, you'll see it start rising somewhere between 25 kHz and 30 kHz but remaining low up to 70 kHz or so (that's why a 50 kHz lowpass filter is recommended for DSD64). With higher DSD rates the noise is shifted to much higher frequencies.

 

Apart from that, I think I understand up-conversion to some extent, but the filters have me scratching my head. I have no idea what any of them do.

 

To further complicate issues, I use Dirac room-correction filters as an AU plug-in. Currently, I have sets for 44.1, 88.2, 48 and 96 kHz. The newer version of the software will enable me to create 176.4 and 192 kHz. So let's assume I can make all those.

 

The DAC will accept up to 384kHz PCM input.

 

Am I best off using Audirvana to up-sample? Although Audirvana can upsample to 384, I will only have Dirac room correction filters up to 192kHz.

 

When playing PCM, the DAC chip upsamples everything to 352.8/384 kHz which is then fed to its sigma-delta modulator. The DAC firmware can do its own upsampling to bypass the chip's filters. The filter selection lets you choose the anti-aliasing profile for the internal upsampling. Fast roll-off gives less aliasing and more ringing, slow roll-off more aliasing and less ringing. Upsampling in software bypasses some or all of this.

 

So, should I have Audirvana up-sample everything to 176.4 or 192 kHz, apply the Dirac Room correction filter, and then pass the corrected 176.4 or 192 kHz data stream to the DAC for further up-sampling, and if so, should I use further PCM up-sampling, or DSD up-sampling, and with what filters?

 

If playing, say, a 96 kHz source, there is by definition no content above 48 kHz, so applying room correction before or after upsampling makes no difference. Moreover, your measurement mike probably doesn't capture much, if anything, above 40 kHz, and even if it does, your speakers (B&W CM7 IIRC) can't reproduce it anyway. This means running Dirac at a high rate is mostly a way to burn CPU cycles. It may of course be easier to set it up that way, but will have no impact on the sound.

 

In addition, when I go to create the new room correction filters, I what up-sampling and filter settings would I use on the DAC? Does it matter if the room correction is only applying a correction curve say between 30Hz an 18kHz? Also, the correction curve imposes a rather sharp cut-off above (eg) 18kHz. How does this effect what choice of filter to use in the DAC?

 

I bought this thing because I thought it would be enlightening to play around with these parameters, but now I realize I know next to nothing about where even to start.

 

Try various settings and see what you like best. In the end that's all that matters.

[beerandmusic]

Considering quantum effects they almost certainly down sample

 

What sample rate do you consider the "original analog recordings" to be at? Since you've said that you don't like upsampling, is this for all the DACs you have listened to which do SDM? Are you then concerned about which brick wall filters your DACs are using? Do you know? If not, what is the basis for your opinion?

 

i know an analog signal is a continuous sampling.

i just don't know if things are recorded digitally now or not? and if the original recordings are digital do they upsample to make sacd's? "i think i know what i don't know" maybe everything is still analog recorded and all media is downsampled differently depending on if they are making cd's or sacd's?

[beerandmusic]

Considering quantum effects they almost certainly down sample

 

What sample rate do you consider the "original analog recordings" to be at? Since you've said that you don't like upsampling, is this for all the DACs you have listened to which do SDM? Are you then concerned about which brick wall filters your DACs are using? Do you know? If not, what is the basis for your opinion?

 

i have no knowledge of dac design, nor do i really want to learn. I just want the professionals to guide me, and let my ears do the picking. (jk grin)

[beerandmusic]

Analog "sample rate" is infinite, so any digital format sample rate is downsampling (or "decimation").

 

does everyone still do analog recording?

i would guess for the old stuff, they have to physically break out analog masters, but would think that most stuff would be digitally recorded today?

[beerandmusic]

It's been many years and lots of reading and I think I'm still trying to learn how DACs really work.

 

The short answer to OP is whatever sounds good to you, just go with it.

 

The long answer, despite some interesting comments so far, is that I think for the Teac DAC, it can play DSD streams "natively" although I've already had some further discussions with people whether that's totally true. The DAC chip is a multi-bit SDM chip. So I don't know what frequency the AK4490 DAC chip runs at but basically almost all multi-bit SDM chips for simplificiation are running at say a 4-bit 11MHz (I'm guessing a little bit at the number and it obviously varies from chip to chip). So regardless of what PCM files you're sending the DAC, it'll eventually be converted to say 4-bit 11MHz before analog sound comes out. That's why these upsampling options are so confusing because the DAC will always upsample, regardless of what option you choose. So even if you choose not to upsample and play native at 16-bit/44kHz, what you're telling the DAC to do is to send the 16-bit/44kHz stream to the DAC chip and let the DAC chip upsample/noise shape, etc. What the options Teac is offering is that if you choose 8fs, you're using Teac's chosen algorithm to upsample to 24-bit/352kHz first before sending the data to the DAC chip and then the DAC chip will continue to upsample to say 4-bit/11MHz to play the music back. So you're essentially choosing whether the AK4490 DAC chip does the upsampling, the Teac uses its own algorithm for upsampling. By using software to upsample first, you're creating another method for upsampling. And in the case where you can choose to convert to DSD at whatever sample rate you want, you're once again telling Teac to use its own algorithm to convert 16/44 to DSD64/128/256/512, before letting the DAC chip play the DSD "natively".

 

Theoretically, with a powerful computer and appropriate software, you can upsample "better" than what the DAC chip and the Teac's internal hardware upsampling can do. But most software uses a pretty standard upsampling algorithm (e.g. JRiver) which is probably why most people would say, I've tried software upsampling but I prefer my DAC doing it because the DAC's upsampling algorithm is usually designed to sound good with the DAC chip. But then you also hear people say they prefer to use XXHighend or HQPlayer to upsample than using their DAC because those software generally use a lot of your CPU power to upsample, compared to what most DACs can put into their hardware chips for upsampling. That's probably why there is such a wide variety of opinions on what is "best" upsampling.

 

The biggest issue for OP is that he has to run Dirac first. Let's say we are dealing with a 24/96 Dirac signal at the end. And then he can decide how he wants to further process the Dirac signal. Would it sound better if the computer converts the Dirac 24/96 signal to DSD512? Would it sound better if the Teac DAC converts the signal do DSD with its internal hardware? Would it sound better without any upsampling and the AK4490 DAC chip just gets the Dirac 24/96 signal? Who knows?

 

On top of that, people have different preferences to sound so some people prefer one type of upsampling/digital filters and others prefer another. If I were to guess, the "best" sound will probably be for the Dirac signal to be upsampled by a "good" algorithm in the computer too to 24-bit/352kHz/384kHz but who knows whether the computer can handle it.

 

But this is all speculative. So it's probably best that OP experiments. Or just not worry about it.

 

this all seems informative (although still greek to me), but my main question has to do with your statement that you think it can't play dsd natively? i thought that is what i was doing when i play dff or dsf files and set upsampling off?

[Jud]

this all seems informative (although still greek to me), but my main question has to do with your statement that you think it can't play dsd natively? i thought that is what i was doing when i play dff or dsf files and set upsampling off?

 

Specialized terminology. "Native" DSD means "not DSD over PCM (DoP)." You need about twice the bandwidth for DoP as "native" to achieve the same DSD rate, so "native" saves bandwidth/overhead. Depends on your hardware and availability of drivers for various OSs. I don't know about the Teac as I've never bothered to find out.

[Jud]

does everyone still do analog recording?

i would guess for the old stuff, they have to physically break out analog masters, but would think that most stuff would be digitally recorded today?

 

Find me vocalists and instrumentalists who sing and play digital samples rather than analog, and I'll agree you don't need a decimation step.

[Jud]

i know an analog signal is a continuous sampling.

i just don't know if things are recorded digitally now or not? and if the original recordings are digital do they upsample to make sacd's? "i think i know what i don't know" maybe everything is still analog recorded and all media is downsampled differently depending on if they are making cd's or sacd's?

 

Recording is the reverse of playback. Just as in your DAC the digital bitstream is (1) upsampled to high rate PCM then (2) goes through a sigma-delta modulator to become a DSD-like signal, then (3) is put through a final filter to convert it to analog music, when recording an ADC (3) takes the analog music and puts it through a sigma-delta modulator to turn it into a DSD-like signal, which (2) is converted to some rate of PCM inside the ADC, usually 24/96, and is then (1) further decimated to lower rate PCM (RedBook).

 

SACDs can be the result of converting PCM recordings, or they can be taken from the ADC before decimation to PCM (caveat that at least parts of the recording will almost always have to be converted to PCM or analog for editing, then converted back to DSD).

[beerandmusic]

Specialized terminology. "Native" DSD means "not DSD over PCM (DoP)." You need about twice the bandwidth for DoP as "native" to achieve the same DSD rate, so "native" saves bandwidth/overhead. Depends on your hardware and availability of drivers for various OSs. I don't know about the Teac as I've never bothered to find out.

 

i think the digital readout said DSD 256 DF150K or something like that...i don't know if the DF 150K is pertinent?

[beerandmusic]

Find me vocalists and instrumentalists who sing and play digital samples rather than analog, and I'll agree you don't need a decimation step.

 

 

i thought the recording gear would be digital like everything else in the world?

[Jud]

i think the digital readout said DSD 256 DF150K or something like that...i don't know if the DF 150K is pertinent?

 

Nope.