Discussion in 'Blind Testing and Psychoacoustics' started by ultrabike, Aug 17, 2016.
The steepness of the filter is independent of the whether the filter is linear or minimum phase.
I think this video explains the differences really well:
@Madaboutaudio, I'll take a listen after work... Can't do right now. Thanks for posting
Slow rolloff linear phase for Sabre DACs please. The needles, the needles!
I think I got confused with that linear phase = sharp roll off, and minimal phase = slow roll off. Guess it's not the same thing after all.
I think the Gibbs effect isn't as bad as it seems on well implemented dacs, but that really depends on your ears and equipment on whether the pre-ringing or any other digital artifacts can being noticed easily or not. Sony seem to have done a very remarkable* job on the Zx2 which doesn't seem to have noticeable ringing or needles etc, which I suspect could be due to well thought engineering along with the use of high quality smoothing capacitor.
*they managed to find a good balance on getting the right amount of sharpness to the sound without sounding like needles
Gibbs is good.
We needs it. We likes it. Precious.
When dealing with adaptive equalizer, folks normally never initialize the main tap to the front. It's usually somewhere in the middle or a little to the right or whatever... because there is precursor stuff to deal with.
I could say similar things about Kalman filters and GPS. Sometimes we need accuracy and we can go forward and backwards in time (analysis of the data). Sometimes that's a bit difficult.
One needs information in past and future for awesome-sauce interpolation. We can't have the future. So we delay shit.
Gibbs is indeed a way to look at things from a different perspective...
Neil Young had ESS make him his own custom warmfag filter for the Pono for good reason.
The only Sabre DAC I've heard that ever sounded right was the Audio-GD NFB-7, the hugely expensive one with three separate transformers for left, right, and digital. Every other Sabre sucked but the ones that "softened" the treble like m920 and XSabre were less bad.
"The ODAC-revB was not created to replace the original ODAC created by Yoyodyne Consulting and NwAvGuy with a superior design but to extend the life of the Ojective DAC concept promoted by NwAvGuy. The ODAC-revB was created to deal with unforeseen component issues. The resulting board has the same footprint as the original and is a drop-in replacement"
"From the side by side comparison you can see the combination of SA9023 and PCM5012A has improved numbers for all measured parameters with jitter showing the largest improvement."
So goodbye ESS ES9023A and Tenor. Hello TI and SA9023...
Inaudible meaningless improvement ... so... errr... they did it anyway. Cuz it didn't make a difference (actually it did according to the numbers).
Anyhow. I think the issues had to do with component life or something. But who knows. Things proly do indeed sound better now. I dunno. I don't have the ODAC. Don't own any stuff with ESS.
speaking of ESS, it seems that there's a new/upgraded Mirus Pro dac with ES9028PRO at whopping US$6,000.
Also it seems project86 got his hands on the Mirus Pro:
I probably could have expected the passband ripple there.
I'm not sympathetic to minimum phase filters, but I wasn't convinced that the phase shift is so detrimental to listening. I suppose the non-linear phase might risk added peaking of signal as with headphone equalization, without the benefit of improving the phase accuracy of the response.
Slow roll filters reduce the temporal ringing, but this isn't a matter of shifting the ringing after or symmetrical to the impulse as with minimum or linear phase filter. Some DACs do use minimum phase filters that also have smoother cutoff, but these are not synonymous.
Slow roll filters have a certain issue with them: with insufficient suppression of frequencies above half the sample rate, you allow the digital junk in, images and aliases of the original signal. The more bandwidth the DAC has over Fs/2, the more aliases are made.
Don't ask me what that sounds like, I haven't played around with different filters.
Yep. One does risk that. I remember sometime ago Xnor applied a simple filter to a file that was heavily compressed in dynamic range... and all of the sudden it was no longer compressed in dynamic range... though it was.
The deal was that the filter added some peaking to the signal due to phase non-idealities. Enough to make it seem that there was no dynamic range compression.
Will this increase the risk of clipping or having analog front ends working on non-linear regions for a short period of time and produce nasties? Probably.
If all of that is not an issue, will things sound a bit off? Probably as well. Maybe not as a tone issue, but maybe certain signal combinations will not sound right. And phase in general may through things off depending on source material.
BTW, I do understand where you are coming from.
And this and that may happen. And then this and that may not.
I'm actually not too concerned about minimum phase filters. I just don't see their advantage or how they solve this "pre-ringing" problem that AFAIK is not a problem.
I do see however that they don't perform as well though. So, it seems to me they are solving a problem that does not exist, while creating other problems that do exist, however big or small they are.
Could one not argue that all digital filters are solving a problem that isn't that important with glaring issues? And just to play a little bit more devil's advocate; why hasn't a company designed an all in one system where the speaker or amplifier has a sympathetic phase-shift to their DAC's minimum phase filter as to cancel it out? Where phase shift is audible with headphones/wideband speakers how noticable is it with multi-drivers which are not sufficiently tuned as to have the drivers time coherent?
@RedFuneral, I found a nice article which explains in depth, the technical design issues involved with what you mentioned:
Note: This non-time coherent issue is something I am experiencing* on the Sony XBA-Z5 (Hybrid, 2BA,1Dynamic) Iem). But this issue doesn't happen all time and it is highly dependent on recording, and sometimes the effect can be quite intriguing to listen to as it somehow makes music instruments popup at different position/timing vs what you have heard before on other single driver iems.
*I have also encounter the same issues with Westone 4R.
Even if you argue that NOS does not sound that horrid, sample rate conversions need them.
Proly because there is just too many permutations. There is not just one minimum phase filter. And proly because that's what an equalizer is supposed to do.
Also, producing a speaker or amplifier that is sympathetic phase-shift wise to a DAC seems harder to me than the other way around.
Furthermore, why force the costumer to use your system only and rule out other combinations using other components. Doesn't seem even logical to me from a business perspective in many ways.
Hi-fi multi-driver speakers are usually tuned to be time coherent using the cross-over. The cross-over is more than a cross-over. In many cases, it seems like an analog equalizer.
I think there are also cases where the tweeter and the woofer are located a bit off from eachother to minimize phase shift issues (time alignment).
Why hell, if things are 180 degrees misaligned between drivers at the cross over I think you get all kinds of suck outs cancellations and crap. This is AFAIK pretty critical and the difference between a shit job and a good job, obvious in the results. I think @noaudiophile bitch slap the Micca Club 3 at some point for this:
Doesn't really matter. Phase shift is all over the place when we include reactive elements in the circuit (coils and caps for xover and analog EQ). Caps will delay the signal differently at different frequencies. So will coils. It's basically a huge clusterfuck once you plot the phase out. But xovers do their job and they do work. Generally the simpler the xover, the less phase clusterfuck there is. I doubt anyone here is going to hear any phase weirdness as long as the xover / EQ is not greater than 3rd order or high Q shit is avoided. Basically, this is all worrying about nothing.
In the meantime, I will augment these discussions with some scientific drawings and plots:
LOL! I think it maters a little. I mean one can even tell if the filter is linear or minimum phase by looking at the impulse response out of a DAC after all the clusterfuck.
And some folks seem to be able to produce decent responses out of a speaker system after all shit is accounted for (at some point in 3-D space or so). I know it's kind of zoomed out and shit, but look at the impulse response of the Quarks:
Not sure if I 100% believe the guy, but the Quarks seem to have a following. Here is the kit:
End of the day one likes what one likes. But once in a blue moon one suddenly feels the urge to call out shit... "pre-ringing" my ass.
Here is too "science":
Not sure if those lines represent filters or hair growth...
trying to figure out why I can't "dislike" that post.
the cow pattern on the accent pillows is just wrong. she doesn't appear to be a cow.
Separate names with a comma.