Article: High End Munich 2016 In Words

[EuroDriver]

Chris,

 

I hope you can secure the Sound Galleries Music Server for a review. The sound at the special event at BMW Welt had blown me away ... especially considering the price of the DAC and amps.

 

When running a DSD signal therough the T+A DSD8 DAC, there is no D/A conversion as we know it (with computation) taking place. The signal from USB input passes only a low pass filter before reaching the anolog output stage. That is why it sounds so good. This is the concept pioneered by Łukasz Fikus in the Lampizator DACs many years ago.

 

Hi Chris,

 

I can fully understand your being somewhat dismissive about the concept of format conversion and upsampling. Lothar Wiemann, Chief Designer of T+A told me only two weeks ago that he fundamentally believed that PCM should stay PCM and DSD should stay DSD. After the event last week Lothar and his colleagues are being open minded.

 

A good dealer friend of mine in Germany with a previous IT background asked me " how is it possible with no additional information, to improve the sound ". We can think of a video analogy with the new 4 K flat screen sets which all have built in upscalers which work pretty well.

 

What Jussi is doing in his software is

- original A>D artifact detection and removal

- intelligent interpolation for new images at a higher sample rate

- very well done creation of a 1-bit stream with a user selected choice of 5 th and 7th order modulators

- 80 bit and 64 bit floating point precision, I don't think any DAC to date is running routines with this level of precision.

 

Chris, if you have the time, we will get you a SGM 2015 to review together with the T+A DAC 8 DSD. The same for you Ted :-)

[EuroDriver]

MSB says they do 80bit processing for a few years now.

 

MSB is running their digital filter with 80 bit depth

 

HQP is running routines with 80 bit floating point math, and 64 bit floating point math, more complicated than doing 80 bit and 64 bit fixed point and certainly with an order of more magnitude precision than fixed point math.

 

Is there an audible difference ?

 

Here is a very interesting take away data point. Many HQ Player users have tried the computationally less intensive -2s family of filters which have a much lower CPU load. The filtering is done in a 2 stage process in contrast to the single stage Polysinc Filters, but the sound quality is significantly lower. Nearly Al l HQ Player users confirm this SQ difference.

 

I asked Jussi during the Tech session after dinner why there was such a large sound quality difference between the two filter families. Jussi replied that he is surprised by the SQ difference and can not explain why it is so large as the numerical difference in the results of the two filters is quite small

 

So we have the consistent observation that very small differences in processing precision can have a surprisingly large influence on sound quality.

 

We also hear this precision benefit with the Chord Dave where Rob Watts has implemented a digital filter with something like 168,000 taps. Amazing detail retrieval, but the sense of space we hear with the T+A at DSD 512 leaves the DAVE in another sonic playground.

 

I suspect the short bit length of the shift register in the T+A and the clever way Lothar has implemented the initial stages of the high frequency low pass filter has a lot to do with this.