I need to learn about up-conversion and digital filters

[Jud]

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

 

Nearly all of it is. The point is it has to convert the analog signal that's coming in through the mics to digital samples, and at that point it's changing from a continuous signal to a sampled one, so it's decimating the infinite analog "sample rate" to some finitely sampled digital format.

[beerandmusic]

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).

 

i will definitely read all the messages in this thread, and try to think of an intelligent response, but i likely won't be able to (grin)....but i will at least fool myself into thinking i am learning something.

[YashN]

Specialized terminology. "Native" DSD means "not DSD over PCM (DoP)."

 

I'd rather separate the notions of 'native DSD streaming' from that of 'native DSD conversion' at the DAC.

 

e.g. you can have DoP from your main computer (non native DSD streaming), but still native DSD conversion at the DAC, case in point being a Mac connected to the iFi iDSD Nano where the latter's Burr-Brown chip converts native DSD.

 

The point about overhead with DoP is, of course, correct.

[wgscott]

(Regarding Dirac Room correction)

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.

 

Assuming I have the facts right, the following happens in this order when I use Audirvana with Dirac filters applied as an AU plugin:

 

(1) Audirvana reads in a file and (if the option is selected) up-samples it to a pre-determined sampling frequency.

(2) The Dirac room-correction filter is applied.

(3) Playback

 

Dirac generates a set of filters subsequent to a single set of measurements. Whether the room correction window is the full audible spectrum or restricted say from 35Hz to 150Hz, filters are generated for all of the standard sampling frequencies (44.1, 48, 88.2, 96, ... and up to 192kHz or the limit imposed by the DAC, whichever is lower). So if I elect to up-sample to 384 kHz, I cannot apply any of the Dirac-generated filters.

[Jud]

Assuming I have the facts right, the following happens in this order when I use Audirvana with Dirac filters applied as an AU plugin:

 

(1) Audirvana reads in a file and (if the option is selected) up-samples it to a pre-determined sampling frequency.

(2) The Dirac room-correction filter is applied.

(3) Playback

 

Dirac generates a set of filters subsequent to a single set of measurements. Whether the room correction window is the full audible spectrum or restricted say from 35Hz to 150Hz, filters are generated for all of the standard sampling frequencies (44.1, 48, 88.2, 96, ... and up to 192kHz or the limit imposed by the DAC, whichever is lower). So if I elect to up-sample to 384 kHz, I cannot apply any of the Dirac-generated filters.

 

Can Dirac-generated filters be applied offline? (I have a feeling this is a dumb question.)

[wgscott]

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

 

 

This is what I had in mind:

 

The noise in the 8kHz to 80kHz range (the series of spikes) is significantly increased. Granted it should still be below the noise threshold for my listening environment (currently polluted by leaf-blowers, chainsaws and other two-stroke engines). But it is in the audible region.

[wgscott]

Can Dirac-generated filters be applied offline?

 

No.

 

Even the AU plugin is experimental. The typical way it works is it is set up as a virtual output device, and thus at the penultimate step of playback.

[Jud]

This is what I had in mind:

 

The noise in the 8kHz to 80kHz range (the series of spikes) is significantly increased. Granted it should still be below the noise threshold for my listening environment (currently polluted by leaf-blowers, chainsaws and other two-stroke engines). But it is in the audible region.

 

120dB down, which is the difference between a relatively quiet listening room (~40dB) and standing next to a top fuel dragster without hearing protection (~155-160dB). (By the way, the vibrations from the latter at that volume will blur vision and cause difficulty swallowing.)

 

Edit: I would expect the noise would be significantly down with conversion by HQPlayer or AuI Converter versus the measurements you show.

 

Further edit: Please also note as the DSD rate goes up, the noise in or near the audible range is reduced. This is a reason I recommended trying DSD256 if your computer will do it, DSD128 if not.

[Norton]

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

 

FWIW I have an Esoteric SACD player/DAC which uses the same chips and I suspect near identical processing.

After experimenting, for PCM I have settled on 8x upconversion and SDLY sharp filter. Turning upconversion and filters off sounded crude (but I haven't tried yet with HQP) while PCM> DSD was just too soft sounding for me.

 

On DSD the options are different - just filter on or off -I stick with the former.

[jabbr]

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

 

Yes most modern recordings are digital but you specifically asked about digitizing analog tapes

 

There are outstanding PCM approaches which use R2R/ladder DAC and are truly PCM "native".

 

Most DAC chips actually SDM and can be best though of as doing a type of DSD conversion. Alternatively you can do a very high quality DSD conversion in software. The reason to upsample is to "push" the digital noise higher into the MHz spectrum and thus place less demand on the analog filter that is used in the last stage of the DAC and to prevent ultrasonic trash into your audio system. For example you can use a 100khz filter which has much less effect on audible signals than say a 20-40khz filter.

 

So from a technical point of view there are very rational reasons to upsample and unless you are delving into the nitty gritty details of DACs there is no reason to prefer no upsampling. That of course does not mean that all upsampling is the same and there are good and poor ways to upsample. Pick a good one

[ecwl]

I've seen this argument from pro-DSD people a lot that most DAC chips are essentially a type of DSD. Except for most multibit SDM chip, if you're summing up 16 different streams of DSD signal, you end up with a 4-bit/2.822MHz signal that's is very unlike DSD. And most of these chips don't run at 2.822MHz but at much higher frequencies.

[beerandmusic]

Yes most modern recordings are digital but you specifically asked about digitizing analog tapes

 

i said do they remaster analog tapes (vs having stored digital copies of everything that they either up or downsample)...my guess is that they just store digital recordings. And curious if they were highrez or not....since only cd's "were" popular in the day, maybe they digitized just slightly higher than necessary for cd?

[JohnSwenson]

I've seen this argument from pro-DSD people a lot that most DAC chips are essentially a type of DSD. Except for most multibit SDM chip, if you're summing up 16 different streams of DSD signal, you end up with a 4-bit/2.822MHz signal that's is very unlike DSD. And most of these chips don't run at 2.822MHz but at much higher frequencies.

 

That is not what happens. For multibit PDM (Pulse Density Modulation) (DSD is 1 bit PDM) a single SDM (Sigma Delta Modulator) is used but the quantization is multivalue. This is very different than having 16 SDMs with 1 bit quantization, which is what it sounds like you are talking about.

 

The multilevel output then goes to a DAC that converts these levels to analog voltage. Various different value to bit configurations have been used over the years. The "thermometer code" is popular these days, but is certainly not the only one possible.

 

The multiple values do not even have to be linear, as long as the what is done in the DAC part matches what the quantizer does.

 

 

John S.

[Bob Stern]

My guess is that a good starting point would be to set Audirvana to upsample everything to 176 or 192 and turn off both upsampling and the digital filter in the DAC. iZotope performs these functions with 64-bit arithmetic and up to 2 million filter coefficients, far surpassing the precision of the digital filter in the DAC chip.

 

Regarding your iZotope settings, my ears are tin rather than golden, so take the following with a grain of salt:

 

You should increase "max filter length" to the maximum possible value, i.e., 2 million. That maximizes the accuracy of the filter. Lower settings are offered only for slower computers.

 

You might try a lower steepness value to reduce ringing. I use 50. You also might try pre-ringing = 0, since many people find pre-ringing more audible than post-ringing.

 

You might try a lower cutoff frequency to attenuate high frequency distortion in early digital recordings made with insufficiently steep anti-aliasing filters. I believe this is what Meridian calls an apodizing filter. Because my tin ears lack response to 22 KHz, I set the cutoff frequency to 0.8, i.e., 17.6 KHz for 44.1 sample rate. I do not notice any treble rolloff on any recordings with this setting. However, I realize that is heresy on this forum.

 

Lastly, an even more heretical idea: After you get used to the sound under those conditions, try setting Aurdirvana to upsample to 96 KHz instead of 192. 12 years ago, Dan Lavry asserted that most DAC chips sounded better at 96 than 192 KHz because the chips at that time required more settling time to achieve their lowest distortion. Newer chips may have overcome this problem, but it's worth experimenting with.

 

http://lavryengineering.com/pdfs/lavry-sampling-theory.pdf

 

p.s. Given that you've never heard a difference between DACs, please let us know: (1) whether you hear a difference among these settings, and (2) whether optimizing the Teac this way enables it to surpass your Peachtree.

[mkrzych]

You might try a lower cutoff frequency to attenuate high frequency distortion in early digital recordings made with insufficiently steep anti-aliasing filters. I believe this is what Meridian calls an apodizing filter.

 

Does it mean that apodising filters have sharper roll-off at lower frequencies than the regular ones like minimum/linear phase/ half-band or linear phase soft-knee filter? It might be hard to say and depends on the DAC implementation I guess or those are rather standard definitions of the FIR filters? I have found for instance that for high resolution files from old recordings like jazz from 50s/60s apodising sounds slightly better than minimum/linear phase soft-knee. For redbook, I think linear phase half-band.

[Bob Stern]

Does it mean that apodising filters have sharper roll-off at lower frequencies than the regular ones like minimum/linear phase/ half-band or linear phase soft-knee filter?

Yes, I believe that's what Meridian means by "apodising".

 

But I may be over-simplifying, and it's been a long time since I read the Meridian promotional material. Eloise probably can elaborate.

[mkrzych]

Yes, I believe that's what Meridian means by "apodising".

 

But I may be over-simplifying, and it's been a long time since I read the Meridian promotional material. Eloise probably can elaborate.

 

Possibly, plus eliminating pre-ringing AFAIK. I've found nice explanation from user Ken Newton here:

 

I think there are two main issues which are affecting the success of so-called 'apodizing' reconstruction filters. One issue has been the confusion over what, exactly, defines an apodizing filter. As I recall, Peter Craven, who proposed this filter in an AES paper, defined apodizing as a minimum-phase reconstruction filter that featured a stop-band frequency that not only fully rejects the ultrasonic alias image band (of which, the ubiquitous half-band reconstruction filters do not fully), but also rejects the upper band edge of the original ADC used to for recording. To my thinking, that latter part was the innovative part of what Craven proposed, not so much the use of minimum-phase filters, which, while pretty much ignored for digital audio up to then, were also quite well known to DSP in general. In addition, I find the the name apodizing less than an accurate description. The name itself (which loosely means, 'to remove a foot') actually speaks to what's called filter 'windowing'. Such windowing is about lessening the abrupt transition effects of having a finite length filter kernel, when sampling theory calls for an infinitely long filter kernel.

 

The other main issue affecting the wide adoption of apodizing, IMHO, is that it's founded on an unproven assumption. Which is that linear phase digital filter 'pre-ringing' is responsible for the negative subjective response many of we audiophiles have to 'the sound' of digital audio. Therefore, the use of a minimum-phase reconstruction filter would eliminate such pre-ringing at the playback end of the chain (by greatly increasing the post-ringing, it should be pointed out). However, the pre-ringing stemming from the ADC linear-phase brickwall anti-alias filter is already encoded on the CD. Can anything be done about it? Craven realized that this too could be removed if the playback filter stop-band was lowered enough to fully reject the Nyquist frequency, perhaps just a little bit below even that. Down to around 21.5KHz, or so.

 

The minimum-phase filters provided within DAC chips appear to meet the first half of what Craven proposed, which is the elimination of pre-ringing by the reconstruction filter, but not the other half, which is the elimination of pre-ringing encoded within the music by the ADC's linear-phase anti-alias via inclusion of the Nyquist frequency within the filter's stop-band. In either case, this elimination strategy is based on the assumption that pre-ringing is bad for the sound, which, while it may seem to make some intuitive sense, I don't believe has yet been proven correct.

 

Audio Engineering Society titled Anti-Alias Filters and System Transient Response at High-Sample Rates and Controlled Pre-Response Anti-Alias Filters for Use at 96kHz and 192kHz. The name “apodising” is owned by Meridian (in the past available here: http://www.avguide.com/forums/apodizing-filters)

[Bob Stern]

Thanks, Krzysztof.

 

Unfortunately, Craven published his 2004 paper defining his apodizing filter in a journal (AES) that's costs money to download, so there are many summaries of his design by people, including me, who didn't read the paper!

 

If eliminating pre-ringing is the main objective, then all minimum phase filters accomplish that, as stated by Ken Newton.

 

I believe Ken's references to ringing at the Nyquist frequency are almost correct but not exactly. By definition, digital audio cannot reproduce a signal at the Nyquist frequency, so there is no ringing at the Nyquist frequency. On the other hand, the analog low-pass filter used in the A/D at the time of the recording can produce ringing at its corner frequency, which should be somewhat below the Nyquist frequency — e.g., 19 KHz. This ringing would be encoded in the CD and could be attenuated during playback by any lowpass filter having a cutoff below the ringing frequency (19 KHz in this example), such as a preamp treble control or a digital or analog filter in the DAC.

 

The ringing produced by the analog low-pass filter of the A/D is NOT pre-ringing because an analog filter cannot produce pre-ringing. Pre-ringing, if any, is introduced by the D/A if its digital filter is not minimum phase.

[Bob Stern]

Regarding Meridian: When they use the term apodizing in their product descriptions and marketing, they are not bound by Craven's 2004 paper. They can let apodizing mean anything they want, with different meanings at different times, like calling a food "healthy".

[mkrzych]

Regarding Meridian: When they use the term apodizing in their product descriptions and marketing, they are not bound by Craven's 2004 paper. They can let apodizing mean anything they want, with different meanings at different times, like calling a food "healthy".

 

Indeed. I have read another quote about it:

 

companies tend to use that name for all filters of the “minimum-phase” type, which remove the oscillations before the impulse, but the Meridian filter does not only take care of the signal it processes, but also each oscillation up the recording chain, also in the studio. Because it works not only on the current signal, but it also improves the recorded one. And this is quite unique.

[mkrzych]

Additional read: https://mrapodizer.wordpress.com/2011/08/16/technical-analysis-of-the-meridian-apodizing-filter/

 

[...]So the absence of the pre-echo comes at the cost of an increasing phase shift across the entire frequency range. People may argue that the shift is only small, and this may be good enough for normal audio applications, but for high-end it seems a serious issue. For one the human brain uses phase information of frequencies up to 3kHz for spacial localization. The inner ear itself has an amplification around 3kHz, which makes the ear most sensitive in the 2 to 5 kHz range (for human speech). And the most important one, instruments and voices are not just limited to a single frequency, they operate on a range. For instance drums even operate on the entire range from low to high. Every frequency component of the sound will get a different shift in time with this filter, it is like smearing the sound over the frequency range, with lows first and highs last. This might also explain why the lows are more pronounced.

[Jud]

Bob - I am supposing there are also digital filters in ADCs.

 

Krzysztof - What you refer to in your latest post is "dispersion" or "group delay," where the time a signal takes to go through the filter depends on its frequency. Keith Johnson of Spectral has said group delay is the primary source of bad sound quality in RedBook playback.

 

At least in my system, where the speakers are designed to be phase correct and to have all frequencies arrive at the listening position simultaneously (see Vandersteen Audio High End Speakers | Vandersteen Audio), this group delay with minimum phase filters is noticeable (I've run A/B comparisons for myself at home and confirmed this) and is not a good thing. However, with other systems/speakers, pre-ringing might sound more evil. Again, digital filter design always involves compromises between opposing distortions, and depending on your system and your ears, a particular type of distortion may bother you, while it may not affect another listener so strongly or at all. Thus we get different DACs with different filter designs, or software that allows a whole universe of such designs. (Though there will be a more restricted set of the possible filter designs that are of good enough quality to use in a signal chain.)

[mkrzych]

 

Krzysztof - What you refer to in your latest post is "dispersion" or "group delay," where the time a signal takes to go through the filter depends on its frequency. Keith Johnson of Spectral has said group delay is the primary source of bad sound quality in RedBook playback.

 

Is it something that could be approach to be truly fixed using digital filtering in the DAC?

[Jud]

I believe Ken's references to ringing at the Nyquist frequency are almost correct but not exactly. By definition, digital audio cannot reproduce a signal at the Nyquist frequency, so there is no ringing at the Nyquist frequency. On the other hand, the analog low-pass filter used in the A/D at the time of the recording can produce ringing at its corner frequency, which should be somewhat below the Nyquist frequency — e.g., 19 KHz. This ringing would be encoded in the CD and could be attenuated during playback by any lowpass filter having a cutoff below the ringing frequency (19 KHz in this example), such as a preamp treble control or a digital or analog filter in the DAC.

 

 

See https://en.wikipedia.org/wiki/Gibbs_phenomenon

[mkrzych]

See https://en.wikipedia.org/wiki/Gibbs_phenomenon

 

Does it have some cure using Lanczos factor in delta sigma if used at all in the DAC?