HQ Player

[Miska]

17 hours ago, OzarkMtn said:

I am running HQPlayer 5.1.0 on a M2 Mac Studio, I updated my microRendu v1.5 this morning and now receive a message regarding "broken pipe". I enabled the Log feature, how do I navigate to HQPlayer logs?

 

If you boot microRendu with my NAA OS, does it still work correctly? This could be related to recent update of the stock OS that is not related to HQPlayer or NAA as such.

 

[Kalpesh]

I have been playing with an offline resampler. with and without eQ performed offline

 

With offline eQ, on the tracks tested, peak hold level goes a hair above 0 dB on HQP meter ;

For pure HQP : add -6 dB for PCM vs SDM compensation (Holo DAC), add a fixed secure -18 dB for convolution, add at least -2 dB Volume to avoid inter sample overloads : that's huge difference, that I can of course compensate via my preamp but offline eQ sounds subjectively more dynamic in that context*. 

 

So is there a way to anticipate the risks and maybe not compute as much as (absolute value) 18 for convolution for every track ?

Isn't the -6 handicapping PCM in HQP ? I think I recall you consider digital attenuation as harmless but when it adds up to nearly 30 dB I'm under the impression I can hear dynamic compression in comparisons. 

 

*for the rest, would the fact that an harmonica might sound (almost) like an accordion, albeit with the nice feature of a bigger venue signature/reverb, be accounted for by the notion of truncated transients (paradox, I don't feel the drums and cymbals as wrong) ?

 

[Miska]

2 hours ago, Kalpesh said:

I have been playing with an offline resampler. with and without eQ performed offline

 

With offline eQ, on the tracks tested, peak hold level goes a hair above 0 dB on HQP meter ;

For pure HQP : add -6 dB for PCM vs SDM compensation (Holo DAC), add a fixed secure -18 dB for convolution, add at least -2 dB Volume to avoid inter sample overloads : that's huge difference, that I can of course compensate via my preamp but offline eQ sounds subjectively more dynamic in that context*.

 

Where does that -18 dB come from? From the HQPlayer EQ frequency response plot?

 

If peaks go above 0 dB, then you have some serious problem and Limited counter also starts incrementing to indicate the limiter has been triggered. Peaks should preferably always stay couple of dB shy from 0 dB.

 

2 hours ago, Kalpesh said:

So is there a way to anticipate the risks and maybe not compute as much as (absolute value) 18 for convolution for every track ?

 

Why would it be necessary to to push the levels near 0 dB? You only win increased distortion and risk for triggering the limiter in HQPlayer.

 

2 hours ago, Kalpesh said:

Isn't the -6 handicapping PCM in HQP ?

 

No... Holo has 6 dB level difference in PCM and DSD output modes.

 

You get a bit less distortion from the PCM output at -6 dB though.

 

2 hours ago, Kalpesh said:

I think I recall you consider digital attenuation as harmless but when it adds up to nearly 30 dB I'm under the impression I can hear dynamic compression in comparisons.

 

Do you hear a background hiss when you put HQPlayer in pause (with quick pause disabled)? Like a tape noise? If not, you are not hearing any SNR limitations.

 

There is no dynamic range compression unless you begin to hit the limiter (incrementing Limited counter). So you certainly want to keep it zero at all times.

 

Typically attenuation from pre-amp results in bigger SNR loss than modern DACs. So please remember what I've said about optimizing gain structure!

 

2 hours ago, Kalpesh said:

*for the rest, would the fact that an harmonica might sound (almost) like an accordion, albeit with the nice feature of a bigger venue signature/reverb, be accounted for by the notion of truncated transients (paradox, I don't feel the drums and cymbals as wrong) ?

 

Sorry, but I don't understand what you are talking about here?

 

But harmonica as such doesn't have transients.

 

[Audionumber3]

This question is for Miska...

 

Would you say the DAC 200 is more forward sounding then the Spring 3 KTE?

 

I have a Spring 3 KTE which I use with Hqplayer...and so, I was going to upgrade one day to the May, or the DAC 200. But I heard the 200 can sound brighter, and maybe even has less soundstage depth then the May. What is your opinion on the Spring vs 200 in terms of sound characteristics?

 

Thanks

 

[Miska]

26 minutes ago, Audionumber3 said:

Would you say the DAC 200 is more forward sounding then the Spring 3 KTE?

 

Possibly yes. Most of the time mine have been on different systems, thus not so much direct comparisons between the two.

 

My DAC 200 is on my living room system, together with Marantz SA-12SE, Ferrum Wandla and Accuphase DAC-60 card.

 

Spring 3 is on my office system together with TEAC UD-701N, Gustard A26 and Musical Fidelity M3xDAC.

 

(then I also have two separate headphone systems, one with HA 200 and another one with Spring 2 + Schiit Jotunheim - these are my primary development / monitoring systems)

 

26 minutes ago, Audionumber3 said:

I have a Spring 3 KTE which I use with Hqplayer...and so, I was going to upgrade one day to the May, or the DAC 200. But I heard the 200 can sound brighter, and maybe even has less soundstage depth then the May. What is your opinion on the Spring vs 200 in terms of sound characteristics?

 

Brighter maybe yes, but I don't think it has less soundstage depth.

 

I assume these descriptions refer to running the DACs in DSD256 or DSD512.

 

[Audionumber3]

1 minute ago, Miska said:

I assume these descriptions refer to running the DACs in DSD256 or DSD512.

Yeah actually, I'm not even sure. I remember reading it a while ago.... But honestly, I thought I would just ask you. I'm definitely using mine with DSD512 though. Being more forward-sounding admittedly causes me reservations about going for the 200. I listen to a lot of rock music, and I feel like I'm on my limit in terms of forwardness. 

 

Anyway, thanks for your input.

[Miska]

7 minutes ago, Audionumber3 said:

I listen to a lot of rock music, and I feel like I'm on my limit in terms of forwardness. 

 

Just curious which settings do you use mostly? Have you tried poly-sinc-short-mp(-2s)? "The old prog-rock filter"?

 

[Audionumber3]

10 minutes ago, Miska said:

Just curious which settings do you use mostly? Have you tried poly-sinc-short-mp(-2s)? "The old prog-rock filter"?

I feel like I've used almost every filter for a period. LOL. I do gravitate towards the poly-sinc-long-mp more then the short though for rock.

[Kalpesh]

6 hours ago, Miska said:

 

Where does that -18 dB come from? From the HQPlayer EQ frequency response plot?

Yes, to keep it below 0

 

If peaks go above 0 dB, then you have some serious problem and Limited counter also starts incrementing to indicate the limiter has been triggered. Peaks should preferably always stay couple of dB shy from 0 dB.

No incrementation but the offline resampler sure aims at 0

 

6 hours ago, Miska said:

 

Do you hear a background hiss when you put HQPlayer in pause (with quick pause disabled)? Like a tape noise? If not, you are not hearing any SNR limitations.

Can't reproduce it at the moment but sure noticed a surprising (CD says DDD) tape hiss at end of tracks. I'm testing with 1991 Shirley Horn's You won't forget me. Had me thinking there was still a tape deck in the chain. So SNR limitation then ?

 

 

 

Typically attenuation from pre-amp results in bigger SNR loss than modern DACs. So please remember what I've said about optimizing gain structure!

I believe I have a nice analog gain structure : I listen mostly at Unity gain on my preamp (combo of gain and volume settings) with HQP displaying -2 dB volume. If I push volume preamp to the max, that it 8.5 dB above Unity gain in my settings, I get 83 dB per channel with band limited PN

 

6 hours ago, Miska said:

 

 

Sorry, but I don't understand what you are talking about here?

 

But harmonica as such doesn't have transients.

 

I did not and don't wan't to trigger a battle or bash a competitor who seems very nice on your thread as well as his aficionados but I wish to understand. I'm pretty sure you once wrote the upsampling in said offline upsampler truncates transients. Maybe so but right or wrong it doesn't sound bad to my ears when I try to hear that negative effect on hit cymbals or snare : maybe too rich but fine sounding. But some tonality issues put me off, most notably harmonica's in test CD. You say harmonica has no transients so maybe so but what I hear with offline resampler is disappearance of micro vibrations of the reeds that I thought could be called transients ; to my ears they get turned into an homogenised air flow with said offline resampling

 

 

[Kalpesh]

https://www.researchgate.net/publication/357166455_Transient_response_of_a_half-tone_bend_on_channel_3_of_a_blues_harmonica

 

[Jud]

59 minutes ago, Kalpesh said:

https://www.researchgate.net/publication/357166455_Transient_response_of_a_half-tone_bend_on_channel_3_of_a_blues_harmonica

 

 

Sorry Kalpesh, it's not really the types of rapid transients we're concerned with (the frequency range doesn't go much above 500Hz if I've read the article correctly).

[Miska]

3 hours ago, Kalpesh said:

Can't reproduce it at the moment but sure noticed a surprising (CD says DDD) tape hiss at end of tracks. I'm testing with 1991 Shirley Horn's You won't forget me. Had me thinking there was still a tape deck in the chain. So SNR limitation then ?

 

Likely there's a lot of SNR limited source content, but that is unrelated. When you put HQPlayer in pause, HQPlayer plays silence. If you don't hear background hiss, then you don't have dynamic range limitations due to digital volume control.

 

3 hours ago, Kalpesh said:

No incrementation but the offline resampler sure aims at 0

 

Bad idea...

 

3 hours ago, Kalpesh said:

I believe I have a nice analog gain structure : I listen mostly at Unity gain on my preamp (combo of gain and volume settings) with HQP displaying -2 dB volume. If I push volume preamp to the max, that it 8.5 dB above Unity gain in my settings, I get 83 dB per channel with band limited PN

 

Unity gain on preamp is fine. But then you have low gain power amp?

 

On my loudspeaker setups, net volume typically ranges from -40 dB to -20 dB. Given power amp gain of 36 dB this means roughly unity gain through. And no SNR limitations.

 

3 hours ago, Kalpesh said:

I'm pretty sure you once wrote the upsampling in said offline upsampler truncates transients.

 

Yes, on edges and for rest the transients ring constantly throughout the track. So all transients on the track spread over every sample throughout the track, before and after the actual transient. That is the issue with too long filters.

 

3 hours ago, Kalpesh said:

Maybe so but right or wrong it doesn't sound bad to my ears when I try to hear that negative effect on hit cymbals or snare : maybe too rich but fine sounding.

 

It adds long standing artificial "sheen" around the transients that causes "blur" (in MQA language) of transients. Like sharp ray of light shattering through a scratched glass. On the other hand, too short filter (like MQA) also causes errors, just different type.

 

So you need a filter that is both short and long at the same time. So my goal has been to develop filter that is as compact as possible in time domain while being as good as possible in frequency domain. So getting as close as possible to the impossible, since the two domains are related though 1/x relationship. So the goal is that transients don't begin before they actually happen and the tails actually go off at some point. Minimum phase filters are optimal in that sense, since they don't have any pre-echo. And the longer post-echo is masked by natural decay and reverb of the transient. But they have the modified phase response as result.

 

Most challenging transients are high frequency clicks recorded in anechoic chamber (no reverb at all). Least challenging is organ music recorded in church! Also instruments like piano are relatively easy in this respect since they are very bandwidth limited and thus the "transients" are very slow.

 

[Kalpesh]

4 hours ago, Jud said:

 

Sorry Kalpesh, it's not really the types of rapid transients we're concerned with (the frequency range doesn't go much above 500Hz if I've read the article correctly).

Thank you for providing me the opportunity to extend the question to perception and long filters

fact : listening to a well known presumed good recording (on Bob Katz's Honor roll) upsampled with an extremely (insanely) long fIlter I did not recognise the harmonica.

Hypothesis 1 : momentary lapse of reason/I could not process the cues

Hypothesis 2 : truncated transients (but I'd qualify the rich, rather than snappy, sound of hit snare and cymbals a palatable trade-off rather than a negative)

Hypothesis 3 : ???

 

I have been happy for a few weeks with Sinc Long and thus (with a Mac Studio) PCM 16 fs. And then Jussi recently posted on smearing blurring by long filters and PCM distorsion : I have new Graal search crisis in a context where I have to take into account that I might misjudge

[Kalpesh]

3 hours ago, Miska said:

 

 

 

 

Unity gain on preamp is fine. But then you have low gain power amp?

 

On my loudspeaker setups, net volume typically ranges from -40 dB to -20 dB. Given power amp gain of 36 dB this means roughly unity gain through. And no SNR limitations.

 

add (absolute value) 18.5 +2 and you have my -20 dB max volume for better mastering far from loudness war ; 18.5 + 2 +8.5 = -30 being the setting for the bulk of what I listen to and sometimes I lower the volume further : I think we are par

 

3 hours ago, Miska said:

 

 

 

[Kalpesh]

3 hours ago, Miska said:

 

 

So you need a filter that is both short and long at the same time. So my goal has been to develop filter that is as compact as possible in time domain while being as good as possible in frequency domain. So getting as close as possible to the impossible, since the two domains are related though 1/x relationship. So the goal is that transients don't begin before they actually happen and the tails actually go off at some point. Minimum phase filters are optimal in that sense, since they don't have any pre-echo. And the longer post-echo is masked by natural decay and reverb of the transient. But they have the modified phase response as result.

 

 

Wish so much you would extend what you said about Meddle to the all the filters in your go to list !

 

[Miska]

1 minute ago, Kalpesh said:

Thank you for providing me the opportunity to extend the question to perception and long filters

fact : listening to a well known presumed good recording (on Bob Katz's Honor roll) upsampled with an extremely (insanely) long fIlter I did not recognise the harmonica.

Hypothesis 1 : momentary lapse of reason/I could not process the cues

Hypothesis 2 : truncated transients (but I'd qualify the rich, rather than snappy, sound of hit snare and cymbals a palatable trade-off rather than a negative)

Hypothesis 3 : ???

 

I cannot find the recording you are talking about...

 

But the ongoing shimmer makes the sound more congested/busy. This shimmer could also explain the rich character rather than snappy.

 

When I listen for cymbal hits when developing filters, I focus on snappiness and cleanness of the sound. It must be fast and without "halo" around it. When you have a lot of simultaneous sounds in a busy mix, the added "halo" around each will make it congested and less clear instrument separation. This also tends to be weak point around PCM converters. Most of these issues primarily affect 1x rate content, and sometimes 2x rate content. On my own test recordings, this was apparent for example with metal claves.

 

1 minute ago, Kalpesh said:

I have been happy for a few weeks with Sinc Long and thus (with a Mac Studio) PCM 16 fs. And then Jussi recently posted on smearing blurring by long filters and PCM distorsion : I have new Graal search crisis in a context where I have to take into account that I might misjudge

 

I'm personally not a huge fan of those non-apodizing relatively low attenuation super long filters. But I don't mind if someone else likes those, or something else. I just explain my take on what I prefer from the technical standpoint, and how I perceive effects of different filters.

 

Next release makes sinc-short/medium/long filters two stage for SDM outputs. So I was able to run sinc-long to DSD512 output without any issues on my 5800X development machine. So those who enjoy that family of filters will find it much more feasible for SDM outputs.

 

[Miska]

53 minutes ago, Kalpesh said:

Wish so much you would extend what you said about Meddle to the all the filters in your go to list !

 

Particularly huge effort went first into poly-sinc-ext2 and then later to poly-sinc-gauss-long to make these as good all-rounder filters as possible, for wide array of different music genres. Even though latter is called "long" it is still pretty much middle grounds in terms of length. These are what one could still call "compact" without trading off anything really.

 

Note that given otherwise same parameters, only thing that filter length affects is roll-off steepness. Longer the filter gets, steeper it becomes. All other properties stay the same.

 

While on Chord-style filters sinc-L group and sinc-short/medium/long the length also affects stop-band attenuation. But such are pretty much exception to the rule.

 

[Jud]

2 hours ago, Kalpesh said:

Hypothesis 2 : truncated transients (but I'd qualify the rich, rather than snappy, sound of hit snare and cymbals a palatable trade-off rather than a negative)

 

Snare and cymbals, yes, those can have the sorts of transients Miska's particularly concerned with reproducing correctly, I would think cymbals especially.

 

Going back to the harmonica, the paper you cited is concerned with the following process: There are two sets of reeds in a harmonica, one for exhale, one for inhale, and each normally resonates at different frequencies. Talented blues harmonica players can "bend" notes by exciting both inhale and exhale reed at the same time. The reeds start out beating at their normal resonant frequencies, but eventually wind up resonating together at the same frequency halfway between the individual resonant frequencies. Moving between these two states (individual frequencies, beating together at the same frequency) they are in a *transitional* or *transient* state, so named because it is temporary, not especially brief or high frequency.

[cpcat]

It’s interesting that with all of the filters to choose from and all the work done I keep gravitating to one of the first on the list-poly sinc mp.  There is a cue up time of about one minute but once played, okay is then immediate for subsequent tracks assuming same base rate.  
 

I assume the “older” filters have been tweaked/updated over time.

[Kalpesh]

5 hours ago, Miska said:

 

I cannot find the recording you are talking about...

 

But the ongoing shimmer makes the sound more congested/busy. This shimmer could also explain the rich character rather than snappy.

 

When I listen for cymbal hits when developing filters, I focus on snappiness and cleanness of the sound. It must be fast and without "halo" around it. When you have a lot of simultaneous sounds in a busy mix, the added "halo" around each will make it congested and less clear instrument separation. This also tends to be weak point around PCM converters. Most of these issues primarily affect 1x rate content, and sometimes 2x rate content. On my own test recordings, this was apparent for example with metal claves.

 

 

I'm personally not a huge fan of those non-apodizing relatively low attenuation super long filters. But I don't mind if someone else likes those, or something else. I just explain my take on what I prefer from the technical standpoint, and how I perceive effects of different filters.

 

Next release makes sinc-short/medium/long filters two stage for SDM outputs. So I was able to run sinc-long to DSD512 output without any issues on my 5800X development machine. So those who enjoy that family of filters will find it much more feasible for SDM outputs.

 

Thank you very much ! here are links to Shirley Horn's tracks with harmonica : https://open.qobuz.com/track/40554705

https://open.qobuz.com/track/40554700

[Kalpesh]

4 hours ago, Jud said:

 

Snare and cymbals, yes, those can have the sorts of transients Miska's particularly concerned with reproducing correctly, I would think cymbals especially.

 

Going back to the harmonica, the paper you cited is concerned with the following process: There are two sets of reeds in a harmonica, one for exhale, one for inhale, and each normally resonates at different frequencies. Talented blues harmonica players can "bend" notes by exciting both inhale and exhale reed at the same time. The reeds start out beating at their normal resonant frequencies, but eventually wind up resonating together at the same frequency halfway between the individual resonant frequencies. Moving between these two states (individual frequencies, beating together at the same frequency) they are in a *transitional* or *transient* state, so named because it is temporary, not especially brief or high frequency.

excellent summary

[Schafheide]

Hi @Miska Given that you recently tuned the EC light modulators, is there a possibility that you could tune the EC super modulators - to make them, possibly, usable for more of us?  I am certain that this would please many!  I understand that this may not be possible. TIA.

[Miska]

34 minutes ago, Schafheide said:

Hi @Miska Given that you recently tuned the EC light modulators, is there a possibility that you could tune the EC super modulators - to make them, possibly, usable for more of us?  I am certain that this would please many!  I understand that this may not be possible. TIA.

 

That tuning involves adjusting some modulator functional parameters, but doesn't affect the CPU load. I have not seen reason to adjust super modulator parameters.

 

[Schafheide]

OK - just wishful thinking!

[OzarkMtn]

15 hours ago, Schafheide said:

Hi @Miska Given that you recently tuned the EC light modulators, is there a possibility that you could tune the EC super modulators - to make them, possibly, usable for more of us?  I am certain that this would please many!  I understand that this may not be possible. TIA.

I am right on the edge of being able to use the EC super modulators @ 48 x512 on my Apple M2 Max Studio.