John Curl's Blowtorch preamplifier

Status
Not open for further replies.
Jump to Latest
#11,761
I contacted Bruno Putzeys by email a few months ago about this. He confirmed the earlier post but he didn't have the exact values at hand; at worst, the fast mode would start at a "few tens of ns" (but it could be higher values of jitter). That would include many transports indeed.

Bruno had another interesting post here : http://recforums.prosoundweb.com/index.php/mv/msg/22379/325768/0/ (at one point, I tried to track everything he had on the asrc 😉 )
 
#11,762
Oops, I might be creating some confusion here. My question concerned the SRC4192, which the Benchmark doesn't use. My main interest would be looking at other DACs that might have even better jitter rejection than the Benchmark for my setup, which uses a computer with an E-MU 0404 PCI card as the S/PDIF source.

If what Bruno is saying can be taken literally, namely that the SRC4192 will be in narrow mode for virtually all sources, then it really is a superior solution to the AD1896 - contrary to the claims of the Benchmark people. IOW, it could be that John Siau's objection to the SRC4192 concerning the too-wideband PLL would be for a mode that is almost never used in practice.

Edit: I wrote this before I saw Ben's most recent post. I will definitely check out Bruno's other posts too. Thanks again for the info Ben.
 
#11,764
andy_c said:
SRC4192 will be in narrow mode for virtually all sources, then it really is a superior solution to the AD1896
Click to expand...


Used both chips. Wouldn't lose any sleep over the differences. All this is so old that it's amazing it has to be discussed again in this thread. Especially the DAC1 - cheaply made but very well marketed - the antithesis of the Blowtorch.
 
#11,768
Joshua_G said:
To some up my inquiry into the DAC part of my audio system:

Provided the clock's and DAC chip's power supplies are very clean and stable – and provided all other important measures, like PCB layout, are taken:

Will a crystal oscillators having accuracy of +/- 25 ppm, Jitter RMS of 1 psec Max and phase noise of about -105 dBc/Hz at 100 Hz and -131 dBc/Hz at I KHz, costing $29.08 be the reasonable choice?

Or is there a crystal oscillator costing up to $200 with a better noise floor that will justify the extra cost in improving the sound quality?
Click to expand...

The ink is still drying on my 'pat pending' situation with a clocking device. However, for audio use, it is not likely to happen soon. Too damn big, too universal. And that's about all I can say on the subject.
It was big enough that I sat on it and contemplated it.....for 15 years.

As stated, try fleabay and type in rubidium. Usually about 20 or so come up, but 99.99% of them are telecom MHz frequency designs.
 
#11,769
janneman said:
Hi Alex,

Variac and I talked to the benchmark guy at some length, and indeed it seems that they put a lot of effort in getting the timong from the USB right. Apparently (but I'm no expert) it is possible to determine the timing from the external adapter rather than letting Windows determine it. That supposedly gives the Benchmark the edge.

Yes, it is theoritically possible that the benchmark is so bad, adds so much jitter of itself, that a 30 $ transport sounds like a 20000$ transprt. But I don't believe that. Simple measurements show that the benchmark has extremely low jitter, so the only plausible conclusion is that it reduces jitter from any source so much that the differences between the sources all but disappear.

'Bits is bits' in the sense that if a correct bitstream is received, the only way transports or cables can sound different is either by jitter that they introduce and that bleeds through the DAC or maybe things like noise and hum that the physical connection introduces in the analog signal. Logically, if your DAC kills the jitter bleedthrough, and your connection is noise/hum free, there can be no sound difference. Which seems to be comfirmed by the tests.

@Andy_c: We'll see you next year!

Jan Didden
Click to expand...

Jan,

Problem with the Benchmark is not the jitter but the fact that it is using ASRC to convert everything to 96/24 timed from an internal clock. Only when it is fed from USB stream at 96/24 the data on the DAC is unchanged and still clocked by a local clock. It is as simple as that, and my friend's observations confirm the loss of sound quality when ASRC is in use. Logically, if there is a difference in the actual data fed to the DAC there could be a sound difference as well... . And ASRC do change the data 😀 .

Alex
 
#11,771
00940 said:
x-pro: you're sure about that ? The benchmark dac1 resamples at 110khz afaik. :dodgy:
Click to expand...

from the manufacturers website:

The DAC1 USB features Benchmarks AdvancedUSB technology to ensure bit-transparent, high-resolution playback directly from your computer.

http://www.benchmarkmedia.com/system1/digital-analog-converter/dac1-usb

Here is the manual:

http://www.benchmarkmedia.com/system1/files/documents/DAC1_USB_Manual_Rev_E.pdf

As you may see the manual does not provide any audio bandwidth related information regarding the DAC1 performance at more than 96 kHz input. And if they provide a bit-perfect feed from 96/24 USB stream one can only assume that the internal D-A conversion is always done at 96 kHz.

Alex
 
#11,773
In the AudioExpress review of the Benchmark DAC1 by Gary Galo in the January 2009 issue he writes on page 27:

"In the Benchmark DAC1 USB, the AD1896 is set up to convert all inputs to a sampling frequency of 110kHz, which Benchmark has found to be optimum."

I think I've seen something about this elsewhere as well.

Lindsay
 
#11,774
I am talking only about DAC1 USB. It is possible that for DAC1 without USB things are different 😀 .

Alex

P.S. - my observation is confirmed by this difference in two manuals:

From DAC1 manual (page 4):

"The DAC1 is phase accurate between channels at all sample rates, and is phase accurate between other DAC1 boxes at sample rates up to 110 kHz."

http://www.benchmarkmedia.com/system1/files/documents/DAC1_-_Manual_-_Rev_L.pdf

From DAC1 USB manual (page 8):

The DAC1 is phase-accurate between
channels at all sample rates, and is phase
accurate between any combination of DAC1
and DAC1 USB converters at sample rates up
to 96 kHz.

http://www.benchmarkmedia.com/system1/files/documents/DAC1_USB_Manual_Rev_E.pdf
 
#11,775
Actually the processing power of the latest generation of NVIDIA chips is so far beyond all of these ASRC's it isn't funny. People are already running high end scientific FEM software on PS3's.

You could probably run any dream upsampling algorithm real time on one. Some benchmarks are 1000x a high end PC.
 
#11,777
jcx said:
one problem with graphic chips for audio processing is that the performance comes in part from deep pipelining - for less than ~ million pt fft most pc cpu's will beat similar generation GPU implementations
Click to expand...

Not my expertise, the SDK and programming model is a big hurdle, but people I know say a seventh order double precision real time 192K interpolation would be a piece of cake. Maybe it could be done on a GP CPU too, NBD. Also, NVIDIA was showing off some audio specific chips recently maybe that was what they were refering to.
 
Status
Not open for further replies.
Top Bottom