Peter, I used HDCD for the test because I wanted to ensure that the data reaching my DAC via 1) the transport and 2) the soundcard was identical.

I had no choice but to use the mixer pots, in order to get an HDCD signal to be recognized by the DAC.

With the MOTU, there was no attenuation required - I just needed to shift the L/R panpots to their maximum positions. WIth the RME, I needed an exact 1.0dB attenuation - attenuating by 0.9dB or 1.1dB switched the HDCD processing off. When XXHE allows me to attenuate by this level, I'll try it.

BTW, this was done by testing and experimentation...

Mani

Hi Frank,

I've heard improvements from updating drivers before. FYI, I was using 2.8.6 on this occasion.

Mani.

Hi Edward,

I'm sure you're right, though I haven't compared my two PCs... yet. (I might have to wait until my wife's out of town again )

At least we can do something about optimising our PCs, in much the way you describe (thanks BTW). I know that Peter believes the firewire PCI card is really important when using an external firewire soundcard.

But what can we do if our soundcards are manipulating our beloved data?

Mani.

Rolf,

My main worry is that the soundcard (hardware and/or mixer software, I don't know) is detemining the ultimate SQ and that there is nothing we (or XXHE) can do about it. Theoretically, taking the spdif output from the soundcard and slaving it to a high-quality DAC should be all that's required. But I'm hearing differently...

Mani.

Hi Everyone,

As I’ve mentioned in another post, I had a protracted listening session last night (with 0.9u-11) and early this morning (with 0.9u-12). I was using system 1) in my signature (but with the RME FF800).

Here’s what I found:

Double Upsample:

Did not work with Mem box ‘unchecked’ and ‘DAC is 32 bits xxx KHz’.
Worked with Mem box ‘unchecked’ and ‘DAC is 20 bits xxx KHz’.

Very difficult to discern any consistent difference between Mem box ‘checked’ and ‘unchecked’. But if pushed to answer, I’d say I prefer ‘unchecked’ – it seems to sound slightly more ‘solid’ and precise.

My preference for upsampling or not really depended on the configuration:
a) with the soundcard in Master clock mode and DAC in PLL mode, I preferred upsampling - it seemed to smooth out the rough edges, give a blacker background and make things more 3D
b) with the soundcard in Master clock mode and DAC in RAM buffer mode, I preferred no upsampling – way too laid back with upsampling
c) with soundcard slaved to DAC, I definitely preferred no upsampling - upsampling took all the beautiful sparkle and ambience out of the music.

Quad Upsample:

Not tested. My DAC requires a dual AES/EBU input (may need to buy a single spdif to dual AES/EBU converter once RR release their 176.4 KHz material...).

Processor core:

Again, very difficult to discern any consistent differences. But if pushed to answer, I’d say I prefer ‘no appointment’  – it seems to sound slightly more lively (I may have to change my signature if I continue to find this...).

As I was using the RME, I couldn’t resist taking a few nice pics.

Mani.

Hi Everyone,

I'd like to share the findings of a protracted listening session I had last night and early this morning (my wife is away for the weekend, so I just had to take the opportunity).

My intention was two-fold:

1) to identify the difference the Mem box makes
2) to decide whether upsampling helps or not

All listening last night was done with 0.9u-11. This morning, with 0.9u-12. I was using system 1) in my signature.

But things didn't quite work out as I was expecting or hoping... 'The best laid plans of mice and men' and all that...

You see, I wanted to start by making a definitive comparison between XXHE and my Esoteric P70 transport... to convince myself that PC audio really is the way to go. To make the comparison fair, I burned an HDCD-encoded album onto a blank CD (black with bevelled edges) using Yamaha's 'Audio Master Quality' function. This should have eliminated any effects that EAC might have introduced during the extraction process vs. the standard CD.

But I couldn't get my D70 DAC to recognize the HDCD content from XXHE. So, I connected my transport to the soundcard via AES/EBU and started playing around with the soundcard's mixer software. Finally, I got the DAC to recognize HDCD from the transport via the soundcard (I needed to play around with a couple of pots in CueMix, MOTU's mixer software). HDCD from XXHE worked fine also after that.

But I had a real shock when I disconnected the transport from the soundcard and plugged it directly into the DAC again. It sounded better than when fed to the DAC via the soundcard.

I assumed that the MOTU was affecting the sound, so I brought down my RME FF800. Again, I had to play around with FF800’s mixer software to get HDCD to work (I had to decrease the output by exactly 1.0 dB in Fireface Mixer). But when I did, the result was exactly the same. The transport sounded better plugged directly into the DAC. (I was careful to reduce the DAC output by 1.0 dB when listening to the transport to make the comparison fair.)

Via either of my two soundcards, the transport sounded flat and lacking in dynamics compared to being connected directly to the DAC. Theoretically, with the soundcard slaved to the DAC, this should not have been the case.

The only conclusion I can make is that the soundcard’s hardware and/or mixer software is/are having a detrimental effect on the sound. I mean, could anyone explain to me why I should have to play around with the soundcard’s mixer to get HDCD to work? It should work with everything set to zero, no?... unless the hardware and/or software are interfering with the bits.

And even worse...

... whatever the soundcard’s hardware and/or mixer software is/are doing to an AES/EBU feed from the transport, it/they seem to be doing to XXHE (or any other software player I tried). I heard the same distinctive ‘veil’ over the music as I did playing the transport through the soundcard. And this is worrying.

I’d love to hear your thoughts on this.

(FWIW, I did manage to get some answers to my original questions, which I will post separately.)

Mani.

I feel like I'm the bearer of bad news all the time... 

Peter, I think something is wrong!

Have a look at this genio of a 1KHz sine and squarewave spot frequency quad upsampled...

Mani.

PS. I've been working abroad this week and haven't had a chance to listen to 0.9u-8.

Yes, that was my understanding.

But I don't understand why more RAM is being taken up now... if the code remains the same.

Mani.

Hi Peter,

Thanks for getting 0.9u-7 out - I suspect the pre-processing code must have been a bit of a nightmare.

I'd rather not comment on the SQ just yet - need more quality listening time.

But I have a quick question: why does the Mem box not change the RAM used for 24/96 FLAC files? 0.9u-6 used less RAM for these files and I expected checking the Mem box to revert to this.

Is this what you would expect?

Mani.

PS. FWIW, with quad upsampling of 16/44.1 wav files (Mem unckecked), I can only play tracks of max length 15 mins or so. These take my RAM usage up to 2.5 GB. Playing larger files gives me the 'XXEngine3.exe has stopped working' error message.

Me neither...

Hi Peter,

Firstly, thanks for your kind words earlier in this thread. I really appreciate that. I just hope something worthwhile will come of any new insights you've had...


For me, this is priceless advice. The problem, of course, is in finding a reliable reference... and then being able to compare 'apples with apples' once you have it (what with so many different factors at play).

Can't wait for 0.9u-7...

Mani.

Just to be clear, these are my settings:

24/96 FLAC: 'DAC is 32 bits 192KHz'
Double/Quad Upsampled: 'DAC is 32 bits 192KHz'
Non-upsampled: 'DAC is 16 bits 44.1KHz'

'DAC is 16 bits 44.1KHz' does not allow the playback of 24/96 FLAC or double/quad upsampled 16/44.1 files.

Mani.

Sorry, I don't have much time right to reply right now...


I'm not sure how many whiskeys you had before you checked the code, but could you double check?... actually make that quad check

Mani.

Hi y'all,

I've just got back from a really hard day's work and thought I'd relax to a bit of music... through XX, of course.

I started with Claire Martin's 'Too Damn Hot!' album on 24/96 FLAC files. I couldn't stop listening - it just sounds amazing through XX...

I then compared this to my 16/44.1 wav files of the same album.

There is absolutely no question that no upsampling sounds closer to the 24/96 FLAC files than double or quad upsampling. In fact, the non-upsampled files are remarkably close to the FLACs. However, the 24/96 FLACs sound more musical - in short, they sound simply stunning.

... BUT...

... I'm not necessarily pointing the finger at upsampling.

I am now convinced that selecting 'DAC is 16 bits 44.1 KHz' is making the difference.

As Peter has already explained, in this mode the data is pre-processed.

My hypothesis is that for 24/96 FLAC files, the data is also pre-processed. The sound is simply too similar to 16/44.1 wav files with 'DAC is 16 bits 44.1 KHz'.

Peter, am I right?

If so, I think all of you upsampling guys are going to have a treat when Peter forces this to be pre-processed also.

Mani.