Gamma-2 (γ2) DAC Thread - Page 22

greenlee punches.

expensive but sometimes they work well.

My second γ2 lives! This one has AD1896, WM8741, AD8656 and Nichicon KW 470uF output caps in parallel with Wima 1uF MKS2 caps. The companion γ1 full++ has a WM8759 and also the same Nichicon KW output caps. The large caps and high current opamp make it possible to drive low-Z headphones directly from both outputs.

For comparison, my first γ2 has SRC4192, WM8742, OPA2365 and Elna Silmic II 22uF caps in parallel with Wima 1uF MKS2s.

Here are some pics. My γ2 #1 is in the gold enclosure while γ2 #2 is in black.

γ2 #2 and its buddy γ1:


Closer view of γ2 #2:


Cased up (front view):


Cased up (rear view):


Closeup of the front panel:


Closeup of the rear panel:


Two γ2s and an empty silver γ1 full++ case (I used its board to make γ2 #1):


Two "art" shots:

wooooooo you got your front panels!
They look sweet~
forgtlover put our order in so it is only a matter of time!

Very nice Ti. Looks great.

So, here is the big question, how do they sound with the new selection of components?


They look great! Superb packaging!


jk

I'm not big on subjective descriptions, but I'll just say that both of them are superb.

AMB, with all that spare space on the PCB it is easy to forget just how small it is. Someone needs to grab a pic with this DAC hiding under a couple of credit cards to give a sense of scale.

I've tried using a Dremel with a cutting wheel but even on 3mm thick aluminium it's too thick and I end up breaking the wheel. Maybe just use the Dremel to file away at the edges after the round holes have been drilled.

the promised pictures. It'll be caseless until I have some time at the workshop.
This is with the WM8741, the OPA2365AID, the EuroQuartz and small output caps. All went smooth, except for the reclocking chip which was very hard to solder without bridges. The dac with the same package didn't even bridge 2 pins, while soldering. even with my lowly HD595 I appreciate the difference to the gamma1. Also the anticlipping allows me to use the pot on my amp in a better region. Filters make no difference to me, till now .




it is still wet from scrubbing

the switches were surprisingly expensive, so I'd advice non prototypers to hardwire anticlipping and any filter (if the reclocking is used).

Please make sure to include photos with both sides of the credit cards in case we get confused

AMB I take it these great looking panels can be ordered somewhere...

I will offer them when γ2 is released, just like I do for γ1 and Mini³.

The longer I stare at my Mini³, the more amazed I am at how darn small the γ2 is!

RMAA and oscilloscope measurements

I ran RMAA testing on my second γ2, which as mentioned before has AD1896 as the ASRC, WM8741 DAC and AD8656 output filter/converter/buffer. The results are as follows:

- 16 bit, 44.1KHz
- 16 bit, 48KHz
- 32 bit, 96KHz

These tests are all done with γ2's filter switch set to "C" (linear phase). The "32/96" mode is really 24/96 with 24 bits contained in 32 bit words.

The tests reveal a fundamental flaw with RMAA loopback testing on an upsampling DAC such as the γ2 -- The playback and record bit depths and sampling rates cannot be set independently. Since γ2's DAC always runs at 24/96 (with ASRC), there shouldn't be a noise floor or frequency response difference regardless of the input bit depth and sampling rate. Yet these tests show such differences, because the ADC in the sound card that's used to record the DAC's analog output for measurement is set to the same bit depth and sampling rate as the output stream to the DAC. Thus we're limited by the ADC's performance, and the results do not show the true and full performance of the DAC under test. (see note below)

Nevertheless, the results we have here are outstanding, easily the equal of (or exceeds) high-end commercial DACs costing $1000 or more (and I have RMAA'ed many such products).

Regarding the selectable filters, please refer to the following white papers which provides a good overview of what they do and the pros and cons:

- Ayre's white paper on linear phase vs. minimum phase filters
- Wolfson's "Ultra High Performance DAC" whitepaper

Indeed some recent commercial high-end DACs and CD/DVD players are appearing with selectable filters. Many of them use the Wolfson WM8741/42 DACs while others use custom ASICs.

I was able to measure the frequency and time domain effects of γ2's three selectable filters. Filter C is the traditional "linear phase" brickwall filter, filter B is the "minimum phase" slow rolloff filter, and filter A is a linear phase filter with slow rolloff characteristics.

First, here is the comparative frequency response graph showing the frequency domain effect of the three filters.



Here are the time domain impulse response and square wave response oscillograms of the three filters:

Filter C (linear phase brickwall):


Filter B (minimum phase slow rolloff):


Filter A (linear phase slow rolloff):


You can see that the traditional linear phase brickwall filter exhibits the widest passband frequency response, but produces pre-ringing and post-ringing in the impulse response. This is also reflected in the square wave response's Gibbs Phenomenon pre- and post-ringing on both the rising and falling edges. In the real world, musical instruments, voices and naturally-occuring sounds never produce pre-ringing effect (except for something intentionally concocted on a computer to do that).

The minimum phase slow rolloff filter trades some phase response deviation for a complete elimination of the pre-ringing effect. Its post-ringing is more severe than the linear phase filters. The downsides of this filter is said to be less audible and more pleasing to the ear than having pre-ringing.

The linear phase slow rolloff filter is a "compromise" between the two other filters. It has the excellent phase response of the brickwall filter but reduces the pre- and post-ringing to much fewer cycles.

Note that all three filters roll off well above 20KHz on a γ2 with ASRC. On a γ2 not populated with an ASRC, however, the A and B filters will roll off quite severely at 44K and 48K sampling rates (-3dB at ~15KHz and ~16KHz, respectively).

Note about RMAA loopback: One way to work around the ADC bit-depth and sample rate limitation in the recording device, is to run two instances of RMAA, one for recording and one for playback. However in practice this causes both RMAA instances to freeze or crash (at least it does for me on Windows XP). Another way to do this is to run RMAA on two computers and two soundcards, one for playback and one for recording. At the moment I do not have an extra Windows machine to do this.

Thanks for the detail about the different filters it is very interesting to compare them. Seems like filter B is best (B for best ). Maybe Filter C if you don't have the ASRC installed.

I wouldn't worry about capturing more accurate RMAA results unless you get a much more high performance ADC. I'm sure the M-Audio unit is decent but seems like you could do with somthing like an Apogee Mini Me to get accurate results.

BTW is there any real difference between AD1896 and SRC4192 or any other compatible device?

Different people might come to different conclusions about what's "best", perhaps also depends on music program material. Having a switch lets you choose "on the fly".

There shouldn't be any difference between AD1896 and SRC4192 as far as I could tell (except the latter is less expensive). I am not aware of any other part that would be pin-compatible.