CHANGSTAR: Audiophile Headphone Reviews and Early 90s Style BBS

Lobby => Amp and DAC Measurements => Topic started by: Hands on May 01, 2014, 01:16:15 PM

Title: AMB Labs Gamma2 DAC Measurements (y2)
Post by: Hands on May 01, 2014, 01:16:15 PM
Test Hardware and Setup

-AMB Labs Gamma2 DAC (y2) - USB input for data and power from desktop. Has nifty digital filter switcher. Anti-clip mode off. See notes below.

-DIY Gaming Desktop - i5 2500K locked at 4.5GHz, Biostar TZ77XE3 motherboard, GTX 780, Antec EarthWatts 650W Green PSU, Auzentech Bravura sound card, 4 140mm fans, goofy UV cathode lights in case, Fractal Design Define R4 case...overall, most likely not optimal for audio, but does well enough as an output in tests and represents a real-life scenario for many. Plugged into APC H10 power conditioner. I did run Fidelizer on it for the tests. Used USB 3.0 ports from the mobo, tested for which ones produce best results and use those. Generally does not cause issues.

-Creative X-FI HD USB (SB1240) - Takes line-out signal from DAC to line-in for recording purposes, powered by USB bus, plugged into Surface Pro. Unit is limited to 48/96KHz input/output.

-1st gen Microsoft Surface Pro - Records from SB1240 via USB. Running Fidelizer tweaks. Battery powered. Screen is touchable and thus 27% more betterer. Produces cleaner and generally better results than running the SB1240 from the desktop as well.

Software

RightMark RMAA 6.3.0, ARTA

Misc. Notes

This particular Gamma2 does not use hardware upsampling, and it can only run at 16/48 max.

Gamma2 output is ~3dB lower than a "standard" DAC, and I didn't bother compensating for that in measurements. Not really a problem if you keep that in mind.

Could not get reliable results from some tests, like the -90dB low-level resolution test. Will have to invest in other gear in these situations.

Creative SB1240 has a slight roll-off in the bass.

I do not guarantee 100% accuracy with these measurements.

Links

AMB Labs Gamma2 DAC - http://www.amb.org/audio/gamma2/ (http://www.amb.org/audio/gamma2/)
Creative SB1240 - http://us.creative.com/p/sound-blaster/sound-blaster-digital-music-premium-hd (http://us.creative.com/p/sound-blaster/sound-blaster-digital-music-premium-hd)
RightMark Audio Analyzer - http://audio.rightmark.org/index_new.shtml (http://audio.rightmark.org/index_new.shtml)
ARTA: http://www.artalabs.hr/ (http://www.artalabs.hr/)
Title: y2 - Filter A - RightMark 16/48 Measurements
Post by: Hands on May 01, 2014, 01:23:57 PM
y2 set to 16/48 for all tests, as mentioned in first post. ADC set to 24/48 or 24/96, the latter which gives a view of what happens up to 48KHz. The lack of hardware upsampler affects the treble roll-off. This is for filter A. You'll notice it pushes the limits of what the ADC is capable of in some situations.

Frequency response (from 40 Hz to 15 kHz), dB   +0.02, -1.93Average
Noise level, dB (A)   -97.2Excellent
Dynamic range, dB (A)   97.1Excellent
THD, %   0.0013Excellent
THD + Noise, dB (A)   -89.5Good
IMD + Noise, %   0.0044Excellent
Stereo crosstalk, dB   -95.8Excellent
IMD at 10 kHz, %   0.0057Excellent
General performance   Excellent

RMAA attachments will be in following order:
1. Frequency Response
2. Noise
3. Dynamic Range
4. THD
5. IMD
6. IMD swept
7. Crosstalk
Title: y2 - Filter B - RightMark 16/48 Measurements
Post by: Hands on May 01, 2014, 01:28:31 PM
Filter B RMAA measurements.

Frequency response (from 40 Hz to 15 kHz), dB   +0.02, -1.62Average
Noise level, dB (A)   -97.2Excellent
Dynamic range, dB (A)   97.1Excellent
THD, %   0.0011Excellent
THD + Noise, dB (A)   -89.8Good
IMD + Noise, %   0.0042Excellent
Stereo crosstalk, dB   -95.7Excellent
IMD at 10 kHz, %   0.0054Excellent
General performance   Excellent
Title: y2 - Filter C - RightMark 16/48 Measurements
Post by: Hands on May 01, 2014, 01:31:37 PM
Filter C RMAA measurements

Frequency response (from 40 Hz to 15 kHz), dB   +0.02, -0.06Excellent
Noise level, dB (A)   -96.9Excellent
Dynamic range, dB (A)   97.0Excellent
THD, %   0.0013Excellent
THD + Noise, dB (A)   -89.3Good
IMD + Noise, %   0.0047Excellent
Stereo crosstalk, dB   -96.3Excellent
IMD at 10 kHz, %   0.0059Excellent
General performance   Excellent
Title: AMB y2 Square Wave and Jitter Measurements
Post by: Hands on May 01, 2014, 01:40:04 PM
Attached are some square wave and jitter measurements. You'll see that filters A and C are both linear-phase, but A is a very mild form of it. B is a mild minimum-phase filter, similar to what Ayre implements in their DAC. Wolfson has good info on this on the DAC chip's white paper. My preferences are B>A>>C.

1. 1KHz Square wave, -3dB, 16/48 - Filter A
2. " Filter B
3. " Filter C
4. JTest 16 48 - Filter A
5. " Filter B
6. " Filter C
7. " Filter C - Upsampled to 16/48 in JRMC

It could be that the lack of hardware upsampler is what causes the jitter measurements to not look super "clean." I'd like to measure a y2 with the upsampler in place, mostly to see if that affects the digital filter ringing characteristics in the end or not (much like I'll show in a later thread how it affects the PWD MK1).
Title: AMB y2 ARTA Measurements
Post by: Hands on May 01, 2014, 01:44:52 PM
13/14KHz and 19KHz sine wave tests, at both 0dB and -3dB. Attached in order A, B, C with the 13/14KHz test first.
Title: AMB y2 White Noise Measurements
Post by: Hands on May 01, 2014, 01:46:00 PM
White noise measurements for the 3 filters, averaged over time.
Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: firev1 on May 01, 2014, 04:50:30 PM
Could you do a high res -3dbFS 1khz sine spectrum with ARTA? I'm surprised to see that the "skirting" does not show up in the RMAA measurements.

Also to acquire CCIF IMD measurement numbers on ARTA(ADC side), set mode to "Two Sine" mode and calibrate the generator to match the twin tone test signals you are playing from the other ARTA(DAC side) window, even if there is supposed to be no signal output in that ARTA(ADC) window.

Below is how I would set an "External" signal CCIF IMD test, except probably with diff windows.




Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: Hands on May 01, 2014, 06:34:06 PM
Good tip. I was only really using it for the spectrum, but those numbers would be good to see. I also need a better multimeter for calibration purposes. I can probably whip up a quick 1KHz test tonight.
Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: mechgamer123 on May 01, 2014, 11:15:49 PM
Could you explain what the different filters are, and why filter C has much better extension than the other two?
Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: OJneg on May 02, 2014, 01:14:14 AM
It's interesting to look at these measurements next to AMB's. Keep in mind his signal path is much different. Coax input and with ASRC chip are at work in his measurements. And he's also using 24/96 on the input, rather than 16/44.1 (which USB is restricted to)

Looks like both the noise floor and crosstalk are a good 10dB higher than what his tests indicate. I thought that might be a USB thing, but it looks like it's the noise floor of your ADC creepin' in. (If your loopback measurements in the other thread are accurate) Actually, the gamma2 even surpassed the measured performance of AMB's M-Audio Firewire audio interface in most areas so it shouldn't come as a terrible surprise. So in actuality, I'd expect most of those results (specifically noise, crosstalk, and THD) to be even lower than your test showed. Hard to be sure unless some very precision gear is used to test.

Except for the roll-off on the very top, I'd say the gamma2 surpasses your interface as well. Note the wee bit of roll-off on the low end must be from your device's coupling cap (although the gamma2 is cap coupled as well).

Jitter is an interesting thing. AMB specifies .5ps (1ps max) of jitter, but that's only with the ASRC chip being used. Somehow (maybe someone else could chime in), use of that ASRC chip greatly reduces measured jitter. Not sure with how to interpret those results though; Is there a way to get some sort of direct read-out with JTest?

Also worth noting the nature of the gamma2's distortion character. Higher order components are absent, and the level of each component is decreasing with order (which is often something to look for when designing audio gear). Look at the SB1240's or M-Audio's distortion character in comparison and you'll see higher-order components sprinkled throughout.

With regard to the digital filters, I think it's important how both A and B are better at rejecting ultrasonic stuff. I'd wager that this (combined with the distortion character) is part of the reason that the gamma2 has such a smooth character to it. Very digititus free compared to other cheap DACs IMO.

The last thing I would mention is that the early roll-off in the treble isn't as bad as it looks. While Mode C does give more "air" in a loose sense, it doesn't make the DAC any more resolving or detailed really. Just a bit more presence when it comes to cymbals and stuff. OTOH, both A and B's presentation of the rest of the spectrum was superior IMO, making it a small price to pay for playback and listening.
Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: firev1 on May 02, 2014, 03:45:24 AM
I would hold my breath till a 24 bit test could be done as only then can THD test reveal distortion below the 16 bit noise floor, the THD for the price is still friggin good though. Stereo Crosstalk is indeed a limitation of the 1240, max out at 96db.

As for jitter to my knowledge, there is no way to get a direct read out, the only way is to guesstimate and sum the products manually. I would say though that jitter here looks like adaptive isochronous usb, not as good as the usb implementation on the ODAC and async implementations. There is also a bunch of low frequency jitter causing a spread in the base of the fundamental signal. I wonder if these characteristics might improve with 24 bit data(I have not done 16 bit Jtest ever). I think the low frequency jitter is inherent of the Gamma 2 though usb or not, take a look at the 1khz graphs on AMB's site, some spreading in the base of the signal which is not as tight as the ODAC or SB1240.

For measuring the local clock, I reckon they are quoting spec sheet or have a high resolution jitter analyser(less likely).
Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: OJneg on May 02, 2014, 04:07:31 AM
Adaptive isochronous is correct. Possible benefits to be had in terms of jitter with coax source, or better USB input.

24 bit tests aren't going to happen. As I said, the USB input is limited to 16/44.
Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: Hands on May 02, 2014, 07:00:29 AM
Quote from: mechgamer123 on May 01, 2014, 11:15:49 PM
Could you explain what the different filters are, and why filter C has much better extension than the other two?

Different digital filters for the oversampling process that takes place in the DAC. My understanding is not good enough to explain well, but the type of oversampling/filtering used affects the frequency response and other characteristics (i.e. the frequency domain). On the other hand, certain filters introduce pre and/or post-ringing into the equation, thus affecting the time domain. I've commonly been told that either A) this ringing isn't audible and/or B) pre-ringing is usually shown on impulses that aren't natural to music, but I think there is a clear, but subtle, audible difference between different digital filter methods.

I have found that I don't particularly care for linear-phase filters, or those with pre and post-ringing. Likewise, I find it hard to differentiate linear-phase with a strong minimum-phase filter, that being one with a lot of post-ringing. I really like the minimum-phase filter implemented in this scenario. Very short post-ringing. That said, I don't think the DAC sounds bad at all with filter C (strongest linear-phase filter to pick from), but I just prefer B. A is pretty close to B as well. However, some people have the opposite preference as me, and that's fine. You can read more about my subjective thoughts on the filters in my DAC impressions thread.

Quote from: OJneg on May 02, 2014, 01:14:14 AM
What OJ said, because my formatting gets messed up every time I post something...took out all his new lines.

Yup, that's exactly right. We're touching on the limits of what my ADC can do and what you can get from 16-bit audio (my ADC is kept at 24/96 for most of these tests). I can't remember if I updated my loopback measurements of the SB1240, but DACs like this and the X-Sabre tend to max it out. If my ADC gets maxed out, the DAC probably performs well enough for me to not care objectively and, thus, really only focus on subjective traits. The bass roll-off is from my ADC, as mentioned in my first posts. But, overall, the Gamma2 performs very well, especially with distortion.

The treble roll-off really isn't that bad. Filter C brings a bit more detail and air, but, personally, I think it gives a poorer timbre and less body to the sound (personal preference). I really, really like how this DAC sounds with filter B, despite its limitations to USB and 16/44.1. It is a touch dark, but very smooth, on A and B. It is a damn good listen, and I am sad to see it go.
Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: mechgamer123 on June 07, 2014, 09:07:44 AM
@hans, I couldn't find your impressions post in the DAC impressions thread, mind linking me?

I'd just like to post my impressions here (unless I should just post them in the DAC impressions thread instead?). Huge thanks to OJ for letting me listen to this little beast!

Quoted from PM to Oj: "I'm sold on this DAC. I think I'll definitely build one once I get the spare cash over the summer. Like you said, it's a very smooth sounding DAC. I didn't really bother listening for detail levels (since I don't feel confident enough in my ability to discern actual differences and BS I'm just making up in my head) but the instrument separation sounded just a touch worse than the CS4398 I was comparing it to. However, with my STAX Lambda SR-202 and SRM-212 amp, which sound a little bit harsh being run out of the CS4398, and frankly a lot of other DACs I've heard, the smooth sound signature really tames the harshness down to where I can't really even detect it, but it doesn't sound so shitty and rolled off as other "warm" sounding DACs I've tried that just sound dull and boring. "
Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: Hands on June 07, 2014, 11:01:24 AM
No problem! Here's where I started my Gamma2 impressions: http://www.changstar.com/index.php/topic,1535.msg40729.html#msg40729
Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: OJneg on August 14, 2014, 01:29:53 AM
Here's my own run through of the gamma2's measured performance. This is the same unit that hans measured but there's a different signal path here. See attached file. Unzip the .zip file first and make sure the .htm file and the unzipped image folder are in the same directory.

* FR: Slow roll-off filters (A & B) exhibit early roll-off

* Noise and DR: The readouts are limited by 16bit word length. But as can be seen on the noise level plot, this is a very quiet DAC. Keep in mind that while is very much a non-ideal testing scenario the gamma2 performs well; the USB input is driven from the rear of a messy desktop with plenty of other devices on the bus and the line-in is being run across a few mains lines and near an EI transformer. Possible benefits to be gained with independent LPS yet.

* THD and IMD: Comparing these plots vs. Ti's or Hans' is revealing. I'd speculate that the gamma2's analog output performance usurps that of most line-in stages. Other test systems just aren't revealing it. Very impressive.

I suspect any weaknesses of the gamma2 would lie on the digital side, specifically the gamma1 USB input. Hans' jitter measurements hinted at this. Might be interesting to compare the performance after I install the ASRC.
Title: Re: AMB Labs Gamma2 DAC Measurements (y2)
Post by: Hands on August 14, 2014, 12:53:10 PM
Such an awesome, little DAC. One of my favorites out of the DACs I tried in that time period. I thought it sounded great without the ASRC, but, given my tastes, I probably would prefer the sound without it myself.