Thanks! I’m still using it for everything and like it a lot. But, I’ll just throw out this quote from @jussi_laako in this thread for you to consider:
“Most of my content is RedBook and most if it will trigger Apod counter incrementing, so I stay with apodizing filters. Especially since there’s no harm running apodizing filters for something that doesn’t need such.”
Honestly, I don’t know what apodizing really means and wouldn’t mind clarification on it’s definition from Jussi. All I know is that I prefer the sound of gauss-halfband to gauss-long, even on tracks that increment the apodizing counter. It would be nice to know what apodizing filters are doing differently and what sounds I should be looking for when evaluating apodizing filters vs non-apodizing ones.
Here is the definition of apodizing from the HQPlayer user manual:
“For PCM source content, HQPlayer, can detect need for an apodizing filter. This is based on detected errors that originate from the recording ACD or mastering tools. Every time such occasion is detected in source content, the counter is incremented. When such content is detected, especially with higher counts, use of apodizing filter is recommended. This detection is not absolute, but can be used as guidance to decide when non-apodizing filter shouldn’t be used. There is no harm in using apodizing filter for content that doesn’t need one. But there is harm using non-apodizing filter for content that would need one.”
What exactly is the harm in using a non-apodizing filter for songs with high APOD counts? Does it introduce/amplify noise or artifacts that would NOT be present if I just let my DAC handle the PCM file without upsampling? If the answer is yes, then I may look for an apodizing filter I like. But if the answer is no, then I don’t think I would mind those errors.
I heard that the best way to evaluate apodizing is by comparing poly-sinc-gauss-xl to poly-sinc-gauss-xla, but I don’t know what I listening for. Could someone send me a Qobuz or Tidal link to a song that has a VERY HIGH apodizing count AND a noticeable sound difference between those two filters?
Listen to highest frequencies, for example cymbals and other instruments that have frequency content spreading beyond 22.05 kHz. High frequency transients are the best. Listen for accentuated glare and unnatural “rough, digital brightness” or extra sheen or shimmer at the top end. Instead of precise, fast, clean and smooth sound you get from apodizing filters.
Sometimes one could initially mistake such extra brightness as extra detail, but it is actually the contrary, it hides the actual details in overwhelming sheen and shimmer that “blurs” the top-end. A bit like scratchy car windscreen in a direct front sunlight.
When Apod counter increments, there are certainly error signals in the source data that don’t belong to the original analog signal. These are digital artifacts that are results of the DSP algorithms in the production chain, either (typically) from the ADC itself, and/or from the production tools, such as ProTools.
Non-apodizing filter will reproduce these artifacts, so you are listening to the original signal plus digital errors etched on top.
Thank you! I think I’ve found a good track to demonstrate the difference. I listened to Bubbles by Yosi Horikawa (from 0:30 to 1:00). It features many high frequency marble sounds that increment the apod counter rapidly. With poly-sinc-gauss-xla (apodizing variant), these high frequency sounds retain more of their actual marble sound. With poly-sinc-gauss-xl, they sound less like marbles and more like indistinct high frequency tones. I will definitely start using one of the apodizing gauss filters for tracks like this.
Sorry for a bit of offtop, @jussi_laako - just wanted to check if Gustard have managed to sort out the DSD Direct mode in the A26 DAC which I know you own? If so, hopefully there will be a Black Friday deal on it as I’d be very keen to finally use HQPlayer with a DAC whose internal processing can be fully bypassed.
P.S. Poly-sinc-gauss (and also the -xl and xla versions for those prioritising extra space over other aspects of sound) and hires-lp with the 7EC-Super modulator are still out-of-this-world-good as a full package. Can’t get enough of the snappiness of transients and the lack of any digital “edges”/glare.
I can absolutely hear a difference between poly-gauss-xl (non-apodising) and the xla (apodising) versions on certain recordings with lots of apodising errors detected. Whilst the extended frequency response of the -xl appears to give some extra “air” to the sound, most recordings do sound more natural played with the -xla filter (where, as I understand, the frequency range containing apodising errors just gets discarded).
If I understand correctly, apodising errors are likely to occur close to Nyquist. If that’s the case, how is it even possible that my ears are able to hear the difference so clearly at such high frequencies? I’m thinking that either those errors also affect the reproduction of lower frequencies or perhaps they are located further from Nyquist than I assumed.
I’m looking to replace my RME ADI 2 dac/preamp with a new one.
To make the best use of HQP/NAA, I first thought of the T+A DAC 200, but the budget is significant.
The Gustard A26 or R26 also seem interesting for a much lower budget, but is their rendering as efficient as the T+A… Is the difference in price justified?
Also, I don’t know between the A26 or the R26 which would suit me best? Reading different forums, some prefer the A26 and others the R26. I think I read that there were firmware issues on these models, but perhaps they are resolved now?
In short, I don’t really know, or just keep my RME ADI 2!
I just now have. Although, I would need the exact parameters for your crossfeed settings to more closely mimic.
The resultant setting minus the crossfeed was a very nice experience no doubt. However, missing spacial presence associated with my reference preference.
Interesting! You push 44.1 to 1024x48!? I normally keep rate settngs on auto so that my SDM and 44,1-ish sources go to 1024 and 48-ish sources to 1024x48. This works for most filters with 7ECv2/7ECv3/7EC-Light and Super but not with 512+fs Light and Super.
Jussi, I am extremely interested in your post. What oversampling filter are you using with your Mac to get DSD512 with ASDM7EC-light?
I am using a Mac Studio M2 Ultra, and I cannot get ASDM7EC-light to run with any oversampling filter that I have tried. It always get skips and pops, or outright stutters.
I went back and more rigorously tested for audible differences between poly-gauss-xl and xla. Any differences that I initially perceived were the result psycho-acoustic bias. I was NOT able to pass a blind ABX test. A comparison of the upsampled PCM files revealed high frequency artifacts in the audible band, however, they were not audible at listening volumes and only became audible to me at ~20dB+ amplification. It is possible that more severe apodizing errors may be present in other tracks, but there was essentially no audible difference in the tracks that I tested, despite each having 100s of apodizing errors.
Yes they have. All new production should include the updated DAC controller firmware. If the unit is older, you need to ask Gustard support for the firmware update. Although I think someone posted a link to the update in the “best native DSD DAC” thread.
