That gamma is on my list. Once I get my amps sorted out, I plan on moving to that as my only DAC.
What is the best DIY DAC up to $500 - Page 3
Gamma 2? I see AMB claims 139db SNR for it but Wolfson (who make the WM8741 used in the gamma 2) only claim 128dB A-weighted mono for their chip. Strange.
It's best really not to look too deeply into these things (DACs, particularly DIY ones and their specifications), because an honest person could easily become depressed.
Edited by wakibaki - 8/27/13 at 6:05pm
Come on wakibaki you know the answer to this. It's 128 dB monophonic, but this is a balanced DAC chip. The opamp afterwards acts as an unbalanced-balanced converter, giving you better SNR than just taking half the output alone.
Duh. Think again. You have no idea what you're talking about, you can't even proof-read what you've written. If 'this is a balanced DAC chip.' what then would be the point of 'an unbalanced-balanced converter'. It's a balanced-unbalanced converter.
If you combine the 2 outputs to use the DAC in mono instead of stereo, then the SNR improves by 3dB. Not 6, or 12 or 14. That hasn't been done here
If you read this Wikipedia page:- http://en.wikipedia.org/wiki/Balanced_audio you will find that 'In critical applications, a 100% differential balanced circuit design can offer better signal integrity by avoiding the extra amplifier stages or transformers required for front-end unbalancing and back-end rebalancing. Fully balanced internal circuitry has been promoted as yielding 3 dB better dynamic range, as explained above.'
If you look at the schematic you'll see that there is a single chip configured as stereo, not mono, giving 125dB according to Wolfson. Which means that there is 14dB gone adrift somewhere. There's just no way to reconcile the arithmetic. Plus AMB's stated test gear won't achieve a noise floor adequate to measure 139dB.
Edited by wakibaki - 8/28/13 at 6:58am
Pupdac uses the PCM1794, not a Wolfson chip. A hell of a little DAC btw.
*facepalm for not checking* The grub uses the wolfson, and I assumed the pup does too. My bad guys. I will just show myself out.
- 4,926 Posts. Joined 4/2004
- Location: Sunnyvale, CA. USA
- Select All Posts By This User
If you look at the RMAA noise graph at the γ2 website specifications page, it shows that the noise floor is indeed at -139dB at 1KHz. But if you look at the numbers below the graph, the 20-20KHz aggregated A-weighted figure is appromximatey -106dB.
All of these measurements are limited by what I could measure with the M-Audio Firewire Audiophile, and this is clearly stated in the text on the γ2 website specifications page.
Perhaps you might like to look at this article from Audio Precision, acknowledged as probably the leading manufacturer of audio test equipment.
It's an introduction to SNR measurement.
Quote:- 'Bandpass filtering and weighting are often used with noise measurements, to ensure that the measurements are both appropriate and comparable.*
Bandpass filtering limits both the low and high frequency range being measured, defining a measurement bandpass. A typical measurement bandpass for audio devices is 20 hertz to 20 kilohertz.
A weighting filter may also be used, either alone or in conjunction with a bandpass filter. Weighting filters apply one of a number of standard weighting equalization curves to the measurement, typically representing the response of human hearing.'
20Hz to 20kHz.
When you state specifications for your equipment, AMB, it's incumbent on you to state those specifications in terms that are likely to be recognised by a broad spectrum of readers, not pick an obscure definition designed to enhance the impression given, as when manufacturers quote 'Music Power' for their amplifier's output instead of RMS or continuous sinewave output power.
This is the standard employed by Wolfson. If you want to be viewed as a reputable designer, you will adopt the same standards as those adopted by the the DAC manufacturers.
Oh, and you can't apply A-weighting at a single frequency. Didn't you know that?
Edited by wakibaki - 8/29/13 at 9:45am
Thanks for posting that W. I was wondering when I read the response from AMB... most people can hear 20Hz to 20kHz, what is the purpose of the 1kHz value? why is it more important/valuable than anywhere else in the spectrum? Is it because of the log scale? Or did you simply find the frequency with the best response and site that?