DacMagicPlus Filter Mirror Images - Short Study

[KSTR]

What do we get for letting some imaging in ? what is worth the tradeoff . I don't think its explained at all in the tread .

When your hearing stops at 20kHz, or much likelier way below 20kHz, that little bit of imaging is irrelevant to most anyone. You'd need pretty bad amps/speakers to run into any IMD issues larger that the normal IMD. The HF noise from aggressive noise-noise shaping for DSD (and derived PCM) is for sure the much bigger problem.

It's been mentioned already, to find out what's going on, given the more like 16kHz top limit for most of us, we need to reduce sampling frequency to say, 32kHz or even lower and test different filters there.

When the input to an ADC is properly band-limited (both with analog and then with the digital filter before decimation), the DAC's reconstruction shall not add to the overall transfer function and it shall not add artefacts (new frequencies).. The only way to get there is the full sinc reconstruction. This reduces the lowpass function (and impulse response / ringing) to that of the ADCs filter. It cannot get any better than that. You need a steep filter here to avoid a lot of aliasing which would really corrupt the signal forever. It will have a lot of ringing, inevitably.

If the digital signal contains 'illegal' content, the full sinc will produce lots of ringing but of couse this is still correct as it reconstructs the signal that would have been needed as "equivalent analog input" after it has gone through all the ADCs anti-aliasing filters.

The whole idea of ADC-->DAC conversion is that -- for a maximised passband -- the unavoidable ringing at (or near) fs/2 is inaudible to begin with, being outside the human hearing range.

 

[chris719]

Ah, sorry, you'd be wrong there again. Perfectly relevant, I'm afraid.

Nope.

So, we got delta sigma, out of band noise, cheap, alignment-free gear and eventually a decent performance. It took a long time and there are still compromises, especially with D/As designed to take everything from 12bit to 24bit (or more) and sampling rates up to daylight. Personally, if I consume my audio in 16/44, I want a system optimised for that, not a jack of all trades, master of none.

Wrong again. I built an AD1862N-J DAC with PMD100 filter around 2001. I have experienced the supposed pinnacle of Redbook via trimmed R2-R and it is nothing special. Today's top audio DACs are more master of all than master of none. I'm not even sure you know history accurately, because the last production top multibit DAC was from TI. Namely, PCM1702/4.

You love to talk history but you still ignore the fact that you didn't even understand the filter cutoff scales with sampling rate and you are cherry picking ancient crap.

Do I have to go back and quote you when you said that no one could tell one of these old machines from a new DAC in a blind test? If so, why do you want some ancient tank. You can keep your 20 lb garbage player with a 13th order analog LPF. I suggest you go back to the other thread attacking Topping for using the built-in filter of a CS43198 like everyone else does. That seems more your speed.

Maybe you can check the datasheet and see if it's optimized for 48 kHz. LOL

Stop spreading falsehoods and fake information or attempting to rewrite history to suit your narrative. It won't work. Ever.

I'm not spreading false information, you have some kind of issue dude. You go into the Topping D30Pro thread and then crap on them for doing the same thing that everyone else does. Then you come here and post dumb shit from the 80s that is irrelevant and can't even admit that every single DAC made since like 1992 has this "problem" you are ranting about, while suggesting a 13th order passive LPF is better (lol, better watch the distortion with those ferrite core inductors, nevermind the phase shift). No one cares about your fetish with analog brickwall filters. I posted you snippets from about 10 datasheets but you seem to be incapable of reading. You should go into the RME thread and tell them they don't know how to design gear either. Also, I recommend you email Cirrus, AKM, TI, AD, and ESS and ask them to un-optimize their filters for 48 kHz . If you tried understanding the basics of how digital filters work before you pretend to be Claude Shannon it might help.

The real bottom line here is that this is basically a non-issue, you can't read, and despite your BS it does affect close to every single CD, DVD, Blu-Ray, DAP, and phone produced after 1990. You know, anything people on a forum about audio measurements in 2021 might care about.

 

[KSTR]

Still it remains a bit puzzling why all DAC chips, and more importantly all ADC chips, continue to use the 0.4535/0.5465 passband/stopband spec that strictly only makes sense at 44.1kHz, as a sort of established best compromise.
Different filter sets eat up a lot of precious chip real-estate and processing power but this 2021 and resource limits should not be a real deal breaker anymore. As is off-loading the ADC & DAC chip from the hardware digital filter tasks... for those who feel it makes a difference, that is.

 

[chris719]

The Marantz concerned uses a switched capacitor, dynamic element matching delta-sigma D/A and also does not use any such "ridiculous brickwall analog LPF". Tell me, what makes you think otherwise? An internet chat site, or a social media lynch mob?

I forgot to address this part. The reason I said it must use a steep analog LPF is based on your claims. If the player uses CS4396 as you say, which I already posted a snippet out of the datasheet from, it DOES NOT attenuate fully by 22.050 kHz. It's the same f'ing filter cutoff as everything else and the CS4396 doesn't have an external filter input mode that I can see. So, if there is attenuation like you claim, it must be another filter providing that.

Wow, <=10 dB of rejection at 0.5*Fs. Either you performed your test incorrectly or there is another filter like I said. Pick one. That graph is of the combined digital and switched-capacitor filter response, btw. The passband is only -3dB at 0.492*Fs.

BTW I am so impressed you can copy and paste words out of a Crystal datasheet as if the DEM is relevant here.

 

[chris719]

Still it remains a bit puzzling why all DAC chips, and more importantly all ADC chips, continue to use the 0.4535/0.5465 passband/stopband spec that strictly only makes sense at 44.1kHz, as a sort of established best compromise.
Different filter sets eat up a lot of precious chip real-estate and processing power but this 2021 and resource limits should not be a real deal breaker anymore. As is off-loading the ADC & DAC chip from the hardware digital filter tasks... for those who feel it makes a difference, that is.

I can see a logical argument for the standard spec being optimal for a DAC depending on what criteria you value. The ADC though, yes, it's weird. I see aliasing allowed outside of audio all the time, but usually in the analog filter. In my experience, you rarely see a high speed SAR or pipeline ADC with a filter that actually attenuates much at Fs/2.

 

[PierreV]

This, from the man who designed the first digital tape recorder for NHK and wrote the (actual) entire book on digital. Heitaro Nakajima.

Sure, IIR filters, hand adjusted and expensive.

Yeah, but isn't that a bit like old style Swiss watchmaking? Horovitz, who also wrote a book, doesn't seem to be fond of Chebyshev either in theory or in implementation. Assuming the difference in end results is audible (the constant question here), maxing out the result on one characteristic and sacrificing others probably isn't the best strategy.

As much as the concept of the ideal meticulous Japanese engineer cherry-picking and pairing components by hand appeals to me, it doesn't seem optimal or even relevant today.

 

[AudioSceptic]

Yeah, J_J wanted 64 khz sampling with bandwidth to 28 khz or so. Lavry made/make their gear to 96 khz, but filter at I think 30 or 35 khz.

Please forgive my ignorance, but who's J_J?

 

[AudioSceptic]

A 13th order passive Chebyshev filter with ferrite core inductors is probably more damaging to the audio than allowing some imaging. Just because he wrote “the book” doesn’t really mean much. It’s not like he invented sampling theorem.

It is very clear to me that given that 99.9% of all DACs have filters allowing some images through, there is no real issue here. The few products that do it correctly aren’t even universally preferred among subjectivists.

I think we can ignore your last sentence, for obvious reasons. ;-)

 

[chris719]

I think we can ignore your last sentence, for obvious reasons. ;-)

Yes, we all know that audio is full of BS. I just meant to imply that if this was such an obviously offensive thing, you would think people would prefer products that did it "right".

I've been accused of a lot of things over the years, but being a subjectivist is not one .

 

[chris719]

Still it remains a bit puzzling why all DAC chips, and more importantly all ADC chips, continue to use the 0.4535/0.5465 passband/stopband spec that strictly only makes sense at 44.1kHz, as a sort of established best compromise.
Different filter sets eat up a lot of precious chip real-estate and processing power but this 2021 and resource limits should not be a real deal breaker anymore. As is off-loading the ADC & DAC chip from the hardware digital filter tasks... for those who feel it makes a difference, that is.

I thought about it a little more and another factor might be that the new spec war is on power consumption. You can see that the stopband rejection and passband ripple have actually been getting a bit worse over the past 10 years with a few minor exceptions. Since the filters should only be getting cheaper, I'm not sure what the alternate explanation is. The CS43198 consumes a paltry 26 mW in normal operation aside from the load, and it's a headline spec.

 

[BDWoody]

I don't have time for someone as clueless as you.

You need to work on that tone... This isn't how it's going to be here.
You're done in this thread.

 

[Blumlein 88]

Please forgive my ignorance, but who's J_J?

James Johnston. Researcher for Bell labs. Expert in perceptual coding among other areas of psychoacoustics. One of the people involved in MP3 development.

 

[AudioSceptic]

Yes, we all know that audio is full of BS. I just meant to imply that if this was such an obviously offensive thing, you would think people would prefer products that did it "right".

I've been accused of a lot of things over the years, but being a subjectivist is not one .

I'm not doing that at all, just surprised that you would think their opinion was worth anything. They can't even agree among themselves, except that overpriced boutique gear always sounds better than cheaper stuff, regardless of measurements. ;-)

 

[AudioSceptic]

James Johnston. Researcher for Bell labs. Expert in perceptual coding among other areas of psychoacoustics. One of the people involved in MP3 development.

Thanks. For some reason that name is new to me. (I thought MP3 was created by the Fraunhofer Institute, but have just found that others were also involved.)

 

[NTK]

Still it remains a bit puzzling why all DAC chips, and more importantly all ADC chips, continue to use the 0.4535/0.5465 passband/stopband spec that strictly only makes sense at 44.1kHz, as a sort of established best compromise.
Different filter sets eat up a lot of precious chip real-estate and processing power but this 2021 and resource limits should not be a real deal breaker anymore. As is off-loading the ADC & DAC chip from the hardware digital filter tasks... for those who feel it makes a difference, that is.

Nowhere close to being an expert in this area. This is what I found in Chapter 3 (page 52) of Analog Devices' The Scientist & Engineer's Guide to Digital Signal Processing. May be that's why 0.546 Fs is chosen as the stopband (at least one of the reasons). I think the DAC designers just don't think images in that area is that important. I don't think the chosen stopband of 0.546 Fs is limited to Fs=44.1 kHz.

 

[pjug]

Is it just wanting round numbers? Flat 0-20KHz instead of flat 0-19.XKHz?

Still it remains a bit puzzling why all DAC chips, and more importantly all ADC chips, continue to use the 0.4535/0.5465 passband/stopband spec that strictly only makes sense at 44.1kHz, as a sort of established best compromise.
Different filter sets eat up a lot of precious chip real-estate and processing power but this 2021 and resource limits should not be a real deal breaker anymore. As is off-loading the ADC & DAC chip from the hardware digital filter tasks... for those who feel it makes a difference, that is.

It seems that strict passband spec, to couple hundredths of dB to 20KHz is just elevated above the other considerations. Maybe there is just fear of having a spec of +/- 0.02 dB to only 19.X KHz?
edit: correct order of magnitude error!

 

[ninetylol]

This is all very overwhelming for me. What is the best Filter to use here?

 

[NTK]

This is all very overwhelming for me. What is the best Filter to use here?

For me: L fast 2

 

[mansr]

This is all very overwhelming for me. What is the best Filter to use here?

That depends on your requirements, which is why there are choices in the first place.

 

[ninetylol]

I was thinking Red one because it Cuts Fs/2 off. But then im wondering why purple is standard.