T+a dac 8 dsd

[pipis2010]

am i the only one one earth using HQPlayer reading straight 16/44k PCM files with no oversampling or dithering and find it very good ( ok i have a R2R pcm dac ) . or am i really missing something with all the DSD stuff ?

 

I think it's really a personal taste. You should try DSD and see for yourself.

 

Personally, at this moment, I don't use HQPlayer, but Daphile, but I do upsample everything to DSD256 with an ifi nano DSD. If the music sounds sooooo bloody good (at least to my ears and with my mid-fi equipment), I can only imagine what the combo of HQPlayer and a better DSD DAC will do in the future!!!

Food for thought...

[pipis2010]

LOL, thanks for taking the time educating me on your absolute opinion and facts, no offense taken. Surely my 30+ yrs as a musician I have no ref point from live, acoustic and studio settings. Yep, it is difficult to deal with for some.

 

So that's one of the reasons why we have to be careful of what we say. We rarely know the background of each other or where they're coming from...

[pipis2010]

I believe that a faster and easier way (in terms of setup) for people who want to try upsampling to DSD256-512 would be Daphile. It's Linux based on LMS-Squeezelite and you get the DSD filters of mansr (he's the same guy who built the DSD filters for latest beta of A+), maybe previous generation and I am pretty sure you wouldn't need a power-hungry system to implement them.

 

I am using Daphile for some months now and successfully upsample to DSD256 with an i5 Haswell.

 

I am only suggesting this, because HQPlayer is more advanced and apparently needs more resources to successfully upsample to DSD512.

 

Just an easier method, because Daphile can be run off a usb disk, so no installation required and you can setup a small library, just to test the waters. Later, you can go back to HQPlayer in all its DSD512 goodness!!! Again that's just my 2c...

[pipis2010]

It also depends a lot on CPU generation. I can also upsample to stereo DSD256 on i5 6600T (Skylake) using HQPlayer without problems, on purely CPU. I can also upsample to 8-channel DSD256 on the same machine with the help of GTX980 GPU.

 

I can also upsample RedBook to DSD256 using -2s filters on my i5 Mac Mini (i5-2415M, 2.3 GHz), but upsampling hires PCM is too much and becomes unreliable.

 

But remember that every time you double the output sampling rate, you at least double the CPU load.

 

Of course it is possible to make things run on lower end machines by trading quality for speed.

 

At the moment playing @DSD512 to the T+A DAC8 DSD from my Linux workstation, using the MacMini as Windows-based NAA:

[ATTACH=CONFIG]26050[/ATTACH]

 

Hey Miska, thanks for your input and I hope that you didn't mind my comment concerning Daphile, nor am I trying to lure people away from HQPlayer.

 

I completely understand what you are saying, but I was just under the impression that your filters are much more resource-hungry than Daphile's. Of course I wouldn't know for sure, because I am not an expert, just based my opinion on what I have been reading in these forums for the past few months.

 

Anyway, sorry to T+A for going off-topic and I am really looking forward to listening to the DAC8-HQP combo one day!!

[pipis2010]

The SQ improvement going from DSD256 to DSD 512 was the big surprise for us. We were used to the level of improvement going from DSD128 to DSD 256 on the Exasound E20/E22. Expecting diminishing returns, we were expecting to have another half step up going from DSD256 to DSD 512. What we find is an improvement Delta 3 times that of going from DSD128 to DSD256, a difference six times greater than we expected.

 

the reasons ?

 

- the two oscillators in the DAC are 22.6 MHz and 24 MHz, so at DSD512, they are operative at "Native" frequency and not being down sampled by a counter circuit or other frequency reduction scheme.

 

- the oscillators operating at their native frequencies are amenable to have their jitter specs further improved by PLL's which T+A have done with German efficiency and quality

 

- with the implementation being discrete instead of being in chip, signal path length and component value tolerances can be tightly controlled to a degree that is not possible in commercially available chips

 

- HQ Player operating at DSD 512, is building a much larger digital filter file, which takes much longer, but also delivering significantly more precision.

 

Some of the front end data manipulation is performed in HQ Player with 80 bit floating point arithmetic so as to get data output with 64 bit floating point precision. Marry this with sigma delta modulation at 22.6 or 24 MHz, then it's not so surprising that the combination of HQP and the T+A DAC 8 DSD is getting stellar results.

 

one other take away is that the bit length of the shift register used in the DAC 8 DSD is significantly shorter than we expected, the beneficial result of this is the very fast reaction to transients

 

I honestly have to ask, because reading the forums one can immediately see that as we go up the ladder of DSD, the better the SQ becomes arriving at the DSD512 and its implementation by T+A and HQP.

But does that mean that potentially DSD1024 would sound even better? Does it even exist yet or is it possible to exist? Sorry for my newbie question, but logic suggests that since DSD512 sounds so good, then most likely DSD1024 will sound stellar with the appropriate implementation, or we will just hit the ceiling in terms of SQ improvement...

[pipis2010]

It is my understanding that the i7 6700T will not have enough horsepower to get the job done (up-sampling all rates to DSD512). It is only 2.80GHz/max3.60Ghz. An i7 6700 overclocked at 4.00GHz might. I am sure someone else would know better.

 

There is a new baby in town, although much more expensive. i7-69xx, 10-core/20 threads, 25MB...when will we see systems based off of this cpu?

[pipis2010]

IMHO, those 8-10 core systems are really of any use for multi-channel upsampling. They also costs more than 3 times a i7-6700K.

 

Agreed, but I would also suggest that they save even more money in any future upgrades as well. So, yes, more money invested now, but better ROI in the future, especially with the arrival of DSD1024 and who knows what else...

[pipis2010]

I'm not sure what that has to do with my post or why you quoted it, I'm also based in the UK and have never ordered from Newegg but yes it looks a good deal, there are some deals on the 980 and 980ti due to the 1070 and 1080 becoming available. Pipis2010 was referring to the 6950x 10 core cpu which is currently £1,400 here.

 

Most deals in the USA in computer parts, are good deals. In Europe, it's another story, at least in some countries anyway...

[pipis2010]

Better return on roi with that 10 core £1,400 cpu? It's also 140w TDP and the price increase from the 5960x (Haswell Extreme, 8 core 20MB cache £875) to the 6950x (Broadwell Extreme) is a big increase. If you ever see it in any audio-based server I'd be surprised.

 

Well I am already seeing people using Xeons at higher prices, so I wouldn't be surprised. Besides, give it 6-12 months and the price will drop, give it 2-3 years and you have a great cpu that will last for many years.