[Jake2]

Part of what to look for is encoded in this sentence: "Measurements need to be of wide band type to see differences in ultrasonic noise too to see its level and possible spurious tones, which could intermodulate into audio band.". So it is about noise floor level and spurious tones it contains both in band (audio band) and out of band (ultrasconic area). What is dangerous is any ultrasonic content which changes together with in band audio signal (so called correlated noise), since that creates potential source of distortion in audio band called as intermodulation distortion.
Of course, experts like Jussi can see more thing in more types of measurements. Maybe @jlaako will add a comment.

So in layman speak comparing a wideband measurement (or measurements plural) taken of a DAC fed a DSD256 signal vs another taken at DSD512 to see which is relatively noisier in the areas of interest, noise floor, spurious tones etc?

 

[Currawong]

Like it was already mentioned, it depends on music content. If it frequently contains fast transients, then transient smear is easily audible, at least for me. Such a content sounds much cleaner with short or middle lenght filters than with long ones.

It can be heard on cymbals, or percussions in general and for example acoustic guitar transient attacks also with old content created in analog way and then digitized. But transient smear can be particularly audible with pop music content from digital era containing synthesized sounds (drums, percussions and other synthesized instruments with fast attacks and potentially also fast decays).

It is enough to use a clean chain, good ears and best a DAC with direct DSD capability, so no other oversampling (interpolation) filter appears in the chain than HQPlayer one. So then you know you have only HQPlayer oversampling filter in chain so you can easily compare nature of different kind of filters. HQPlayer Client allows filter switching on the fly without playback interruption, so testing that is easy. Playing a PCM stream into a DAC which oversamples itself is not a way to get rid of transient smear.

An examples of such a content, where for my ears short filters with direct DSD help to clean up the smear:
https://www.head-fi.org/threads/hqp...-settings-rolling-thread.968141/post-18051636
https://www.head-fi.org/threads/hqp...-settings-rolling-thread.968141/post-18051842

With such a content IMO the effect can be easiest audible when you compare sinc-L, sinc-long or sinc-Mx with poly-sinc-xtr-short-mp or poly-sinc-gauss-short.

No, that's not what I'm asking. The original question that was asked was:

If one was to use PGGB to upsample files to 16fs/24 bit, in your opinion, would be the best SDM Nx filter(s) to play those files?

Jussi's reply didn't mention genres. I'm asking if a similar result where the entire track would would have ringing wouldn't result from, say, sinc-l.

 

[bogi]

The original question that was asked was:

If one was to use PGGB to upsample files to 16fs/24 bit, in your opinion, would be the best SDM Nx filter(s) to play those files?

For demonstration purposes, could somebody process the 60 seconds snippet (I already posted a week ago) by PGGB and upload the result somewhere?
EDIT: I have it already, thanks.

 

[dericchan1]

It belong to DACs which don't oversample so it is suitable for use with HQPlayer. With DSD input it provides direct DSD path.



This thread is also about transient smear caused by long filters but not recognized and understood by many people. There is no such equation, particularly generally for all audio content, that hardest to run filters provide the best sound. For classical music or acoustic jazz it can be the case. For electronic pop with fast transients it's hardly the case. But sound you get in headphones or loudspeakers is affected by the whole audio chain used. That then affects preferences one has with his chain. So one can provide a general hint for you, but it's you who decides how much such a hint works for you with your chain.

For me personally it does not bring sense to invest to be able to run sinc-L or sinc-long at DSD512. There are other owners of Cyan 2 who can contribute with their experience. For example I know @dericchan1 prefers DSD256 with Cyan 2.

That is correct, I have tried back and forth between dsd256 and dsd512 for months, dsd256 always came out on top as the cleaner sounding, it is especially true when you have tube gear in the chain that dsd512 turns out to be too much of a good thing in complex tracks. My findings are not only on the Cyan 2 but also on my Ifi idsd pro. Some of my friends who have either May or Spring 3 pretty much have stuck with dsd256 for good as well…

(Dsd1024 with the experimental modulator was out of the race after trying it for a couple of sessions)

I do think if you are building a hqplayer pc now you may want to future proof and have enough juice to run dsd512, like at least you have a choice to hear it for yourself if you prefer 256 or 512… but I wouldn’t waste $$$ and effort to build a pc for 1024 tho

 

[jlaako]

Changing DAC bits to 16 worked btw! Sounds awesome. Holo May showing output of 1.536m.
So am I getting some lesser version by going down to 16 bits?

With for example LNS15 noise-shaper and those rates, you are fine with 16-bits. Noise-shaper at higher rates makes up for the lesser bits. But if you like to optimize for 20-bit output, you would need a suitable NAA to bypass macOS CoreAudio and the limitation.

 

[jlaako]

I wanted to come back to this statement. Does this mean that your sinc-L and similar very long filters will also output music with constant ringing?

No, it would have maximum 0.7 seconds pre- and post-ringing and then it would fade away. This is similar to what you get with Chord M-Scaler.

So if the ringing doesn't get refreshed by another transient within that time, it will fade away in that time.

 

[Louisiana]

I wrote this in the Roon forum, but maybe someone here has the same problem:
When I play my Qobuz albums via the HQPlayer client, only the first track is ever played, you can also see this on the left of the program, track 1/1.
When I play local albums, it works fine, it also shows track 1/12 as an example.
What could be the reason?

Qobuz:

local:

settings:

 

[camrector]

Cheers, was just checking whether it was a particular measurement or noise characteristic to keep a look out for. In the absence of such I’ll continue to trust my ears as per usual.

There are measurements that for sure will show you the best dsd implementation. They are device specific.

 

[jlaako]

Jussi's reply didn't mention genres. I'm asking if a similar result where the entire track would would have ringing wouldn't result from, say, sinc-l.

Since due to prior upsampling the ringing is already baked in the source data far below the new Nyquist frequency, you can use for example poly-sinc-gauss-hires-lp for such content.

You can use HQPlayer's 20 kHz filter option though to remove the ringing and replace impulse response of the earlier filter with much shorter impulse response. This would undo some of the transient damage. But it also extra work for trying to undo the earlier conversion.

 

[jlaako]

When I play my Qobuz albums via the HQPlayer client, only the first track is ever played, you can also see this on the left of the program, track 1/1.
When I play local albums, it works fine, it also shows track 1/12 as an example.
What could be the reason?

Since you are on iPad, you need to disable automatic screen lock and keep HQPlayer Client in foreground without locking the device. Otherwise iOS will suspend Client and then it cannot anymore progress to subsequent tracks.

For local library this is not an issue, because HQPlayer server can play those standalone. I notice you have the older app version too, there's now newer 5.6.0 release for iOS.

 

[Louisiana]

@jlaako
Thanks for the quick reply.
Works exactly as you described and I have updated the app too.
Is there a way to get around the fact that the client always has to be in the foreground?
I could then not use the iPad while I'm listening to music.

 

[EMINENT]

I am using the holo audio cyan 2.



I also love using a few ASDM7EC modulators at DSD256. However, with my Mac, I am unable to utilize all of the modulators and filters. Only certain modulators and filters are capable of playing music smoothly without any lag. Furthermore, I am interested in purchasing a more powerful PC because, in addition to DSD512, I would like to be able to use various modulators and filters.

Haulo.

I remark.

I am envious of Cyan 2 users with access to such an affordable price and capabilities. It is up to the individual how far they wish to take it.
But, if you have the capability at hand, you owe it to yourself to hear for yourself what you wish. Like enjoying your steak fried atop a space rocket may be delectable. What if you never tried butter basted, reverse seared, how would you know?

It would be ill advised to misconceive over generalizations from anecdotal experiences without proper context and correlations of each comparison at hand.
As stated by others, too many variables at play from the shape of your ear to cables, power delivery, etc. The differences we are on about could be down to ultrasonics and fractions of a decibel, even placebo and bad gear even though measured with the usual measurements do well. Any piece of gear and anatomy can lead one on a different trajectory.
If one finds the means to build and are within confines of their budget, one should not leave anything on the table. It is possible to enjoy one vs another regardless of measurements. They are good for reference to know nothing is bad or wrong, but rest is subjective preferences. Had I stopped at PCM with my X26 Pro, I surely would be on a different path today.

Some chose NOS sound. Is this objectively measured better? No. Why?
Could be their transducer and the roll off helps treble. Or, one could have a muddy, bass presentation and they need a certain filter with PCM that alleviates. This works for their chain and is unlikely anyone here has identical setups down to interconnects and anatomy.
Many have passed on a piece of gear without having tested every iteration possible and this is fine if one has the time and bank account for such experiments. However, I imagine a bad time for their wallet greater than some have endured.

 

[Currawong]

No, it would have maximum 0.7 seconds pre- and post-ringing and then it would fade away. This is similar to what you get with Chord M-Scaler.

So if the ringing doesn't get refreshed by another transient within that time, it will fade away in that time.

Just to make sure I get this straight, when you talk about a transient, are you talking about something that is still within the bandwidth limit? If I understand what you're saying, you're referring to transients that have energy near FS/2 that trigger the "apodising" counter in HQPlayer, such as a cymbal hit.

 

[Failed Engineer]

Is there a image of NAA 5 that works with Pi3? I had an old Audio Alchemy DMP-1 with massively overkill LPS sitting in a closet that works with Roon and I assumed was Pi based and finally decided to crack it open a few days ago and yup, Pi3b. No NAA image I tried worked but then I found RopieeXL does have a Pi3 compatible version so I put that in the DMP-1 and its NAA works great. I don't know what NAA it has though, and does it really matter much?

 

[bogi]

Is there a image of NAA 5 that works with Pi3?

Pi3 is not supported as NAA since Ethernet and USB port share the same USB bus, so that USB port is noisy and causes issues at higher rates.
https://audiophilestyle.com/forums/topic/13649-hqplayers-network-audio-adapter/page/102/#comments
https://audiophilestyle.com/forums/topic/19715-hq-player/?do=findComment&comment=1096881

If you can, use Pi4 instead.