XXHighEnd

I’ve just gotten around to doing something that I’ve been meaning to do for a while – comparing the effects of QAP on my Esoteric D70 vs. my Pacific Microsonics Model Two.

The D70 uses PCM1704 converters (in dual-differential mode) and still oversamples at 176.4KHz (I believe it’s 2x oversampling, using its FIR filter). The Model Two uses custom, discrete, full-ladder converters and is non-oversampling at 176.4KHz (I believe).

The D70 results first. With no QAP (i.e. straight 16/44.1), XXHE through the D70 sounds great. It’s a beautifully laid back presentation and about as un-digital as you could imagine. If this was the last DAC I ever had, I’d be pretty satisfied with this sound.  However... engaging QAP simply kills the sound. The bass loses its tunefulness – it becomes fat and lumbering. The ambience and twinkle in the top-end is gone. If anything, it sounds like QAP increases smearing. Overall, it reminds me of why, in the past, I’ve never liked the effects of any upsampling.

On to the PM Model Two. The first thing that hits you is the shear visceral impact of the sound – this thing has an output impedance of only 20 Ohms [EDIT: 10 Ohms per leg] and will drive a bloody tank! (If you ever get a chance to hear one of these things... do it!) With 16/44.1 material, the sound is stunning. But... with QAP, it simply gets better! Everything becomes more focused. And the dynamics... well you’d better have equipment that can handle what’s thrown at it.

QAP through the Model Two is the second best sound I can get from digital. The first is from original 24/176.4 files recorded on the Model Two itself (e.g. the 24/176.4 RR files). Interestingly, these 24/176.4 files don’t sound very good on the D70, suffering the same sort of effects as 16/44.1 files with QAP.

In conclusion, I believe QAP needs a 24/176.4 NOS DAC to do its magic...

Mani.

Thanks Mani, this feedback is very useful for me.

Does the Model 2 accept 24/352.8k? :)

No, the Model Two works internally at 24/176.4K (or 24/192K) :(

But even if it worked at 24/352.8K internally, there would be no way of piping data into it at this rate - it only has single- or dual-wire AES/EBU inputs.

I would strongly urge anyone thinking of buying a new DAC to hold off until the NOS1 arrives. I will buy one as soon as one becomes available, and will post my findings - NOS1 vs. Model Two (considered by 'those in the know' to be the best A/D/A converter ever built) - right here.

Mani.

Hi Mani,

Thanks. Indeed this is useful, as you (I think) are amongst the few that can compare the real merits.
Of course it would be true that the 24/176.4 NOS is "needed" because it is made for that (24/352.8 being better again), nothing being in the way to "destroy".

I can't be sure, but the D70 might be engaging the "filtering" in a less intelligent way, meaning it's "harder" - more persistent in applying its filtering than filtering could do at anticipating on what actually (still !) has to be done (this is the hard to explain process about the magic of FIR filtering which makes everything straight just by means of "some" averaging). Of course I can only say this afterwards, knowing how many like the AP even with OS DACs.

Might you run into a native 24/176.4 (or 192) I can try myself for judging, pllllease let me know. It could be very interesting if my conclusions are different (which should not be the case, but seems to be so, thus far).

Thanks,
Peter

very interesting; I am currently running my BADA off my Macbook Pro Toslink, i.e. I don't have support for 176.4.
So I can only use Arc interpolation in double mode (88.2) leading to unpleasant "smearing" as the Alpha DAC adds some "incompatible" interpolation on top.
It might well be that as I am upgrading my SPDIF interface to allow for 176.4 I will come to different conclusions as apparentlt for 176.4 and 192.2 no further interpolation would be done in the DAC.

BTW same is true for my Denon Amp with AL24 interpolation, again I can only do 88.2 and don't like the results compared to 44.1 and letting the Amp do all of the work.

My understanding is that the BADA uses delta-sigma chips, which, by design, need to oversample... even at 176.4/192KHz. (Feel free to correct me if the BADA does not use delta-sigma chips.)

It'll be iinteresting to hear your experiences nevertheless - others on this forum are reporting good results using QAP with delta-sigma DACs...

Mani.

Peter, have you heard of the fluency algorithm that Esoteric uses, called RDOT? Here's some info on it from the Esoteric website:

*************************************
RDOT (Refined Digital Output Technology)

When music data is recorded for CDs, the frequencies that exceed 20 kHz cannot be reproduced due to limitations of the CD standard sampling frequency. Music however, usually consists of frequency information that far exceeds 20 kHz. The effect of not being able to reproduce this additional information is that the listener cannot experience the most accurate and natural sound.

As the solution for recreating sounds above 20 kHz, we use RDOT; a technology that enables interpolation by analogy based on the fluency algorithm. This highly sophisticated RDOT technology uses analogy-based principles to generate all music information likely to occur between data samples. The information is generated from source signals of all sampling frequencies up to 192 kHz, not just the standard CD frequency. As a result, data that exceeds fs/2 is generated. This results in music with natural depth, and listeners can especially feel differences in resolution and reverberation when instrument sounds overlap. The extremely natural timbres and sound fields used to reproduce a live performance are clearly superior.

***********************************************

How similar is Arc Prediction?

Mani.

Not. Or better, it doesn't say much about the process, while the described result would theoretically count for every means of upsampling. But if they made a good job out of it, it should sound similar to A.P.
I guess it depends on how explicit it has been done (like sinx by no means is explicit).

Peter

I've been experimenting with noise isolation and reclocking devices to increase the SQ of my Lynx AES16/BADAC system.  As good as the stock BADAC is I've always felt its SQ was being comprimised by noise and jitter coming into the DAC on the AES/EBU input. 

I started off with an Antelope DA which turns out not to reclock/dejitter the AES pass through, but rather just serves to do noise isolation. Well, there was an immediate improvement in SQ using the DA in between the Lynx AES16 and the BADAC.

Today I took delivery of a Grimm Audio CC1 (Peter - you probably know these guys...).  The CC1 is a master clock and reclocker in one.  Well......this device is quite nice.  Just using the AES reclocker further improved SQ over the DA....not dramatic, but the kind of improvement that once you hear it you can't go back.

But, then using the CC1 to send word clock to the Lynx AES16 card in addition.....further big improvement in SQ.  BIG. 

But the CC1 is just under $3k and the BADAC is $5K (and the AES16 is $700) so I'm going to have to rethink things.  Maybe I'll have to get a NOS1....will it have an analog output stage to directly drive amps?  Clearly the Berkeley guys in order to hit their $5K price point couldn't afford better noise isolation and reclocking technology.

With the CC1 in the audio chain the differences of one player versus another are greatly reduced AND the EMI/RFI noise from the PC is taken out of the equation.  I still prefer XXHE.

In terms of Quad Arc I've come to the conclusion that with the BADAC it doesn't give better SQ and if one does use it you have to use the shorter 2.24 filter and not the 1.24.

I see that Antelope is coming out with their own Zodiak DAC line which will include asynch USB (over 192Khz) and AES and it will include their dejitter/reclocking technology in the DAC together with DC power and a Class A discrete analog output stage.  Should be interesting.  Has anyone heard a MH ULN-8?  There are those that say it outperforms the Model 2.

Yes, and all of 'these' people have an ulterior motive: either they were on the MH developing/testing team, or they are dealers carrying MH products.

If you own a Veyron, you don't need to brag about it... or show boy racers in their Opel Mantas how inadequate they are  ;)

Mani.

PS. I think only people from Germany would really get the Opel Manta comment... and especially seeing as my name is 'Mani'.

You bet !
(I use 3m (10') interlinks here).

But didn't you ask that before ? :) :)

Well, now that I'm using an RME card, I couldn't help having a look at the difference between QAI, QAP and native 24/192 material. I used two versions of the same track downloaded from Linn Records; one at 16/44.1 and the other at 24/192.

As you can see, QAP does a really good job of re-creating the HF information in the 24/192 file. But it isn't quite all there and (perhaps) as a result, it doesn't sound quite as good as the native 24/192 file.

FWIW, the missing HF information in the QAI file is easily audible - the music sounds flat compared to QAP and native 24/192. I'm pretty sure the spectrum would have been identical if I was just using the 16/44.1 file without QAI, but I wanted to be consistent. It's worth noting that 99.99% of people are feeding their DACs with 16/44.1 files with all information above 22.5KHz chopped off!

Note 1: I couldn't get QAI to play for more than 30 seconds or so, hence why the spectrum lower down isn't quite the same as the other two.
Note 2: The measurement at 81KHz for QAP was due to a 'click'/'pop' in the sound, which I still get from time to time with QAP.

Mani - what are you using to capture and plot these measurements?  What is the significance of the ordinates from -70 to -20?

Mani, this thread is very, very interesting. I have thought a lot around this matters myself. Personally, I think that reducing the pre-ringing of every impulse is more imortant than the +20kHz amplitude contents.
The Red Book pre-ringing degrades the microdynamics of all instruments, including bass instruments.

I have started to collect hi-rez titles, and when listening to 24/96, the SQ improvements are more noticeable in the lower octaves, than the very high treble.
When going from 24/96 to 24/192, I don't hear more treble information, but everything else becomes more "real" microdynamic-wise. Especially instruments with high contents of DC. Like drums, guitars and brass.

(Of course, I hear that cymbals too improves at higher rates. I think 20% of its energi specter is above 20kHz).

---------

I suspect QAP is doing "something" with the pulses, because applying QAP to 16/44 I get a similar improvement in microdynamics, as with 24/96.

I recommend all PC-audio users to purchase the dbpoweramp program. It's very good for riping and (batch) converting. Or other DSP operations.  It permits de-coding HDCD (creating a 20bit /44 file out of the 16/44). Also you can do your own downsampling for experimentel use. Purchase a 24/192 recording and downsample it to 24/96, 24/44, 16/44, 16/96 or whatever. Then listen to the differences.

I'm using the RME 'Digicheck' software. The y-axis denotes the signal level (in dB below 0dB full scale), and the x-axis the frequency. So, as you can see, most of the music lies between 60Hz and 1Khz (for this classical piece at least). But if you look at 20KHz and beyond, you will see different amounts of information that each file contains. Although a sine wave at this frequency would be inaudible (certainly to me), the effects of having this extra HF information are clearly audible. Everything has more 'life'... even bass notes are more tuneful and better rounded.


You may well be right. But I can't measure pre- and post-ringing with the Digicheck software :cry:

Mani.

I didn't know that. But then these 20/44.1 files wouldn't work with QAP, and that just wouldn't do!

Mani.

True. Also with the PeakExtension feature, XX can simulate HDCD-decoding somewhat.

By the way; I have several HDCD titles, but they are coded differently. Only a few of them have this Dynamic range compression.

That's because 'HDCD-encoding' includes a number of options, of which Peak Extension is just one. See my 'HDCD 101' thread if you're interested.

Mani.