CHANGSTAR: Audiophile Headphone Reviews and Early 90s Style BBS
Lobby => Amp and DAC Measurements => Topic started by: Hands on February 21, 2014, 10:05:56 AM
-
(http://schiit.com/public/upload/general/imagecache_thumbnail_product_gallery_large/gungnir_01_09.jpg)
Test Hardware and Setup
-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. Plugged into APC H10 power conditioner. I did run Fidelizer on it for the tests. Using USB 3.0 ports, tested for which ones produce best results and use those two.
-JKSPDIF MK3 - Input with generic USB cable, output with 75ohm RF attenuator (-10dB) and 75ohm BNC to BNC coaxial cable to DAC. Battery powered.
-Schiit Gungnir DAC - Takes JK BNC out to BNC in, RCA out to SB1240. Using the most recommended cables listed on Blue Jeans Cables for both. Plugged into wall socket. Also used USB input (same cable used for JKSPDIF) for a couple tests. USB input is Gen. 2.
-Creative X-FI HD USB (SB1240) - Takes line-out signal from DAC to line-in for recording purposes, powered by USB bus, plugged into desktop. Unit is limited to 48/96KHz input/output.
Software
RightMark RMAA 6.3.0, ARTA (used basic multimeter for calibration), JRiver MC 18
Misc. Notes
I used BNC input on the Gungnir for most of these tests. I did some in USB and decided it was not worth re-doing all of them. You can get a basic idea how they compare based on the RightMark and jitter measurements.
Creative SB1240 has a slight roll-off in the bass.
I do not guarantee 100% accuracy with these measurements. Take these for what they are.
Links
Schiit Gungnir DAC - http://schiit.com/products/gungnir
JKSPDIF MK3 - http://www.johnkenny.biz/home-1/mk3-hiface
Creative SB1240 - http://us.creative.com/p/sound-blaster/sound-blaster-digital-music-premium-hd
RightMark Audio Analyzer - http://audio.rightmark.org/index_new.shtml
ARTA: http://www.artalabs.hr/
-
Gungnir with BNC input, RightMark RMAA 16/48 tests using ASIO.
Frequency response (from 40 Hz to 15 kHz), dB | +0.02, -0.14 | Very good
Noise level, dB (A) | -92.8 | Very good
Dynamic range, dB (A) | 93.2 | Very good
THD, % | 0.0046 | Very good
THD + Noise, dB (A) | -82.7 | Good
IMD + Noise, % | 0.023 | Good
Stereo crosstalk, dB | -93.1 | Excellent
IMD at 10 kHz, % | 0.010 | Very good
General performance | Very good
Attached RMAA pictures will always be in this order unless noted otherwise:
1. Frequency Response
2. Noise
3. Dynamic Range
4. THD
5. IMD
6. IMD Swept
7. Crosstalk
-
Gungnir with BNC input, RightMark RMAA 24/96 tests using ASIO.
Frequency response (from 40 Hz to 15 kHz), dB | +0.02, -0.15 | Very good
Noise level, dB (A) | -94.8 | Very good
Dynamic range, dB (A) | 96.3 | Excellent
THD, % | 0.0046 | Very good
THD + Noise, dB (A) | -83.4 | Good
IMD + Noise, % | 0.023 | Good
Stereo crosstalk, dB | -90.6 | Excellent
IMD at 10 kHz, % | 0.0091 | Very good
General performance | Very good
-
24/96 tests WITH USB instead of BNC. Results show USB is likely a slightly better solution than BNC via JKSPDIF, but this could change with a better converter.
Frequency response (from 40 Hz to 15 kHz), dB | +0.02, -0.15 | Very good
Noise level, dB (A) | -96.9 | Excellent
Dynamic range, dB (A) | 97.0 | Excellent
THD, % | 0.0042 | Very good
THD + Noise, dB (A) | -84.1 | Good
IMD + Noise, % | 0.023 | Good
Stereo crosstalk, dB | -94.6 | Excellent
IMD at 10 kHz, % | 0.0084 | Very good
General performance | Very good
-
Gungnir JTest jitter measurements using a 24/48, 0dB file generated in ARTA and played back through JRMC 18. Tested BNC and USB. Also tested with software upsampling in JRMC, just to compare how this turns out on the Gungnir vs. other DACs I've tested (idea is to see if some DACs benefits from it more than other or not at all, though upsampling could ultimately be messing with results and making them useless).
1. BNC
2. BNC - Upsampled
3. USB
4. USB - Upsampled
Jitter is low either way, but USB looks a bit better. Upsampling might actually be slightly detrimental, though still gives good results. IIRC, this is touching on the limitations of the SB1240, so it's entirely possible the Gungnir has less jitter in actuality.
-
-90dB, 1KHz sine wave tests for 16-bit low-level resolution abilities. Posting both time and spectrum results for the 16/48 and 24/96 tests. I can't vouch these tests are accurate, and I'm also not sure if ASIO is forcing 24-bit mode when running these tests (meaning 16-bit tests might actually be 24-bit tests).
1. 16/48 Spectrum
2. 16/48 Time
3. 24/96 Spectrum
4. 24/96 Time
I have confirmed the SB1240 used in these tests can properly play and record this low-level resolution test (i.e. it displays a smooth, clean sine wave when tested for this on its own).
The Gungnir shows that even in 24/96, its noise floor and noise characteristics make it difficult to fully resolve this test. However, the general sine wave pattern is still there. You can detect this background noise if you turn your amp up beyond where you'd normally listen, though it is slight (NOS1704 was noticeably worse). In noise comparison, the Hex is much cleaner. It had no issues resolving this test, but, of course, it output the sine wave as you'd expect a NOS DAC to (i.e. it displayed the 3 quantization steps instead of a smooth wave). The Hex displayed more harmonic distortion in these tests, whereas the Gungnir has more random noise.
I also finally used a multimeter to better calibrate the ARTA measurements. Still hard to get everything exact, though.
You can read more about this test here: http://www.stereophile.com/content/quality-lies-details-page-6
-
Sweep Impulse-based measurements, 24/96 (rest of settings are same as what I did for the Hex). I always got slightly different impulse results each time I ran this, so I tried to pick what looked most repeatable and consistent. Even then, I don't know what to make of the burst decay results (maybe these are right and results like the Hex gave are wrong?). I've seen results like this in other DACs. Maybe it means this set of results aren't actually valid...maybe it tells us something useful about the DAC. I have no idea. So, I won't vouch for the accuracy of these. Take them as they are. If someone knows what's up and knows how to get better results and info from this test, let me know.
1. Sweep impulse response
2. Frequency response and distortion, generated from the sweep impulse
3. Step response (Does anyone know why it droops and doesn't quite come back up to 0?)
4. Impulse response envelope
5. Burst decay
-
Some noise and distortion measurements from ARTA. Done in 24/96. Most of these are dBFS, but not all. Matched as close to 0dB as possible. Note that the graphs do not all start at 0dB.
1. Noise - log graph
2. Noise - linear graph
3. 1KHz sine wave distortion - log
4. " - linear
5. 100Hz distortion - log
6. " - linear
-
Some ARTA 24/96 IMD-based measurements. I believe I matched these up with dBV as close to 0dB as I could, A-weighted.
1. 60Hz and 7KHz, 4:1 ratio
2. 250Hz and 8KHz, 4:1
3. 25Hz and 66Hz, 1:1
4. 13KHz and 14KHz, 1:1
5. 19KHz and 20KHz, 1:1
-
I did a music-based test with the Hex as seen in this linked post, and I thought it would be interesting to see how the Gungnir compared: http://www.changstar.com/index.php/topic,1425.msg37411.html#msg37411
"I took a 1-minute sample of a track I regularly listen to (16/44.1, WAV copy of Opeth's "Ghost of Perdition") and did an exp. average in the spectrum analyzer over the entire sample."
1. 16/44.1 playback in JRMC
2. Upsampled to 24/192 in JRMC
See how it compares to the Hex's results, if you'd like.
-
I thought it would be interesting to see how the Gungnir and Hex compared with basic measurements (since I have the files available), and RightMark makes it easy to generate comparison reports. Tests and order are as follows:
1. Hex - 24/96 - ASIO
2. Hex - 16/48 file, playback upsampled to 16/192 in JRMC - Kernel streaming in JRMC playback, ASIO in RM recording (I know, a 24/96 upsample test would have been a better comparison...)
3. Gungnir -24/96 (BNC) - ASIO
Frequency response (from 40 Hz to 15 kHz), dB: +0.00, -0.29 | +0.02, -0.09 | +0.02, -0.15
Noise level, dB (A): -103.1 | -95.8 | -94.8
Dynamic range, dB (A): 101.7 | 95.6 | 96.3
THD, %: 0.013 | 0.013 | 0.0046
IMD + Noise, %: 0.022 | 0.022 | 0.023
Stereo crosstalk, dB: -96.2 | -96.1 | -90.6
Files are attached in regular order (IMD swept is missing). I know the graphs can be a bit hard to read.
As for that nasty crap in the far end of the spectrum on the upsampled test, I'm inclined to believe those are RightMark artifacts due to the nature of the test and its attempts at interpreting a 16/48-based test for 24/96 results and not the actual output. Also note that the Hex does show oddities in 24-bit RMAA tests, as noted in that original thread.
-
I'm sure it sounds good but this is not mind-blowing performance for a dac in 2014. I realize it's fairly affordable but let's see if Yggy can be more competitive. Expecting a DAC that plays 24-bit files to have greater than 16 bits of resolution is reasonable.
-
Meh, I kind of gave up on relying too much on measurements. They tell us a lot of useful information, but they can make some products look worse on paper than they sound and some sound worse than they look on paper. I think the Gungnir is at bare minimum good enough on all levels (mostly an understatement) that it's reasonable to see why many would like/love it.
I do agree that my main "complaint" would be its noise characteristics and associated resolution abilities (on an objective basis). But it does still sound quite good.
I've also seen measurements of an upgraded Bifrost that looked better than what results I got with the Gungnir, so don't take my work as gospel.
-
I'm sure it sounds good but this is not mind-blowing performance for a dac in 2014. I realize it's fairly affordable but let's see if Yggy can be more competitive. Expecting a DAC that plays 24-bit files to have greater than 16 bits of resolution is reasonable.
Something does not look right (there's a certain way the sine ways should look, and any kind of noise could disrupt that) with the 16 bit / 24 bit -90db tests. As Hans says, take things with a grain of salt. Highly precise equipment is needed for such tests. Here are the Stereophile tests of Bifrost:
(http://www.stereophile.com/images/713Schiitfig07.jpg)
(http://www.stereophile.com/images/713Schiitfig08.jpg)
-
Yup, that's the set of measurements I was referring to.
firev1 had a couple good points, one being that it could be something as simple as a ground loop. Still, I do find it odd that I can get the test to give what seems to be good/correct results from the Hex or SB1240 itself (loopback) but not other DACs. I don't understand electronics well enough to explain such things, unfortunately. On the other hand, I can get the Gungnir to produce background noise on my setup (volume well beyond normal listening levels), so there's still the possibility these measurements have some ounce of truth to them.
Tonight/tomorrow morning, I will experiment with these measurements in particular and see if I can't get some better results.
-
Yep, something is funky with the setup--the Gungnir betters the Stereophile tests of the Bifrost on our Stanfords and on an AP-2722.
-
1khz test and PN noise IR would be cool to see, its on the shoutbox but I mention I do have a problem with measuring the Schiit Modi as well with the SB1240. It maybe some impedance/grounding/Chaos problems affecting the SB1240 due to its non-isolated inputs.
RMAA test on THD do show something though, really low higher order stuff which stuff like the Herus or lesser DACs cannot match.
On another note, I think the sweep impulse do show little compared to what is already there with the other graphs. The stepped sine method would be more revealing in a way though it requires that only one clock be used(a problem for me). Will be exploring what more we can do within our means with methods.
-
Tonight/tomorrow morning, I will experiment with these measurements in particular and see if I can't get some better results.
That will pretty much be impossible with a soundcard setup using the A-D on it. -90db is very very very low level. Consider that even small signals below -90 are actually very high relative to -90db.
I can't emphasize more that stuff like this needs to be made clear. Measuring headphones which have gross errors on home equipment is one thing. DAC require another level of measurement instrument precision.
-
I agree that the SB1240 is not an extremely precise device in this scenario, but loopback tests do show the SB1240's ADC should be able to decently resolve this test in at least 24/96 mode (it shows a somewhat clean/smooth sine wave...firev1 posted his results, and my unpublished ones are pretty close). And I think the Hex results are correct...? I'm sure there are a lot of factors at play here that make some devices look poorer than they are if measured on the SB1240. Until I have a better understanding, just keep my disclaimers in mind.
-
Hmm, fishski mentioned some noise with the AD2000s which I didn't hear. I do wonder if something got damaged in shipping or if I left something loose when I changed out the USB card.
Check to see why the rising noise floor past 15k with a peak at 21k... I didn't see that before with my own measurements (if I can dig them up or if I even saved them.)
-
I'm with Jason and Purrvey on this. Something looks amiss.
When I and the local DIY'ers use soundcard ADC's for measurements we just use them to look for obvious malfunctions. These shouldn't be used to determine absolute performance figures. Also you need to know what to look for when using these kinds of solutions.
-
Hmm, fishski mentioned some noise with the AD2000s which I didn't hear. I do wonder if something got damaged in shipping or if I left something loose when I changed out the USB card.
Check to see why the rising noise floor past 15k with a peak at 21k... I didn't see that before with my own measurements (if I can dig them up or if I even saved them.)
that was desktop computer noise with volume knob wide open on the QRV-08. dead silent from laptop.
-
OK, sorry for the confusion, guys. Bear with me, as I've obviously still got some kinks to work out in my methods. I got too comfortable with the setup I was using since my other tests were seemingly coming out just fine. I was able to get some improved results in some regards (not all) by slightly altering my setup. Instead of plugging the SB1240 into my desktop, I used my Surface Pro (on battery). Most everything else was the same, including having the Gungnir's USB input fed from my desktop. I did a lot of experimentation with the gear and software I have available, and this was about the best I could do. As for that noise I heard when turning up my amp? Probably just a fluke and not from the Gungnir.
You will notice the results are much cleaner looking overall compared to what I posted originally. Firev1 had brought up the possibility of a ground loop, and it could have been something like that in my original tests. Either way, I've attached the revised RightMark, 24/96 results of the noise, DR, THD, and IMD tests. Basic, but it gets the point across (you can assume improvements seen here would apply to nearly all other tests I did).
As for the peak in the noise around 21KHz, I am really not sure. Someone with more knowledge in this area might know. This is most likely an artifact due to my setup, though it is possible something is wrong with this Gungnir (unlikely). Ultimately, this would need to be verified on better equipment, or, if it's just a matter of how the Gungnir and SB1240 (don't) work together, possibly even just a different ADC. I guess you can't expect every type of DAC and analog output stage to work with the SB1240's line-in. Considering firev1 had issues getting correct-looking results from some Modi measurements using the SB1240, I think it's most reasonable to assume this noise rise and peak is an artifact.
Despite getting much cleaner results, I still did not have luck getting better -90dB, 1KHz sine wave test results. Again, this is likely just an artifact, considering the Bifrost did a great job with it (see info in prior posts) and the Gungnir does show good results otherwise. However, I think the SB1240's ADC is capable of resolving this test to a decent degree, considering firev1's SB1240 loopback results (24/96 in particular, which I've been able to replicate) and my Hex results (I'm assuming the quantization steps are always expected to be seen on a NOS DAC in those tests, right?).
In the future, I'll do a better job double checking my results with a couple different setups to avoid producing "dirty" results again. I'll also leave out what appear to be erroneous results (like these -90dB tests), though something minor that can be noted and overlooked due to other redeeming qualities of the results (like many of these with that rise/peak in the noise) will likely get posted. I'll probably stop doing the sweep impulse-based tests as well. Interesting to experiment with, but I don't think I was providing useful info. But, again, apologies for the confusion and less-than-stellar work. (I'd go back and edit my posts if I had the ability to.)
Still, with these revised measurements, and taking the likely artifacts and other points into consideration, there's still some good information to be gathered:
- Frequency response results seem mostly accurate, tell us a bit about the DAC, and validate Schiit's published specs on their site. (The SB1240 will slightly alter the response, with one example being showing more bass roll-off than in actuality.)
- Noise floor is clean and low. I bet those small spikes and wiggles in the 60-300Hz range would be mitigated with a better environment and setup (i.e. USB not from a gaming desktop). Fluke or not, that noise spike around 21KHz still sits below -110dB. Overall, I think this validates Schiit's posted SNR specs. Dynamic range looks similarly great.
- THD measurements hit the limit of what the SB1240 can do, around 0.004%. Anything that hits this limit is already doing a pretty good job, and I see no reason to doubt the Gungnir performing to Schiit's specifications of <0.002% THD in actuality.
- JTest jitter measurements show a low, clean noise floor and sidebands that hit around -130dB. That's around the SB1240's limits, so the Gungnir should do an excellent job in terms of having low jitter (tested BNC and USB).
- RightMark pegs IMD at around 0.02% using the SMPTE test, but do note that Schiit lists IMD specifications in CCIR. I am honestly not sure how CCIF and CCIR differ. But, what I do know about CCIF tests is that I can get close to Schiit's posted IMD specification if I use the DFD2 CCIF results (a common method for CCIF measurements, I've read). At least, testing in that manner gets to the limits of the SB1240. But that might not mean much. Either way, I just wanted to point out why my IMD results will have discrepancies from Schiit's specs.
I knew full well with the SB1240 that excellent DACs would exceed its ADC capabilities and that some gear just might not measure 100% correctly on it. And that's OK when you keep everything in context, plus the SB1240 already performs quite well (much better than its price would indicate!). In other words, if it's maxing out the SB1240's limits, that gear probably already performs at a level of excellence to the point you need not be concerned if you like what you hear.
With these revisions and my noted points, I think these measurements overall validate Schiit's specs and prove it is an excellent performing DAC. I thought it had a lot of really good qualities to it in my listening tests! I think Marv and other pirates have done a good job describing its characteristics.
-
I'm sure it sounds good but this is not mind-blowing performance for a dac in 2014. I realize it's fairly affordable but let's see if Yggy can be more competitive. Expecting a DAC that plays 24-bit files to have greater than 16 bits of resolution is reasonable.
hard to know without seeing what the A/D is contributing to the results, in this case the Creative X-FI HD USB (SB1240). I would guess the results would look significantly different depending on which sound card is used for the test.
-
I need to re-do my SB1240 measurements, but there's some decent documentation on its capabilities in another thread here (combine the best results from myself and firev1, you'll get a decent idea...we were still working on our methods when we posted those results). Basically, it has a flat response with a slight bass roll-off, noise and DR numerically measure around -107-108dB (looking at the graph, though, you'll see the noise floor is clean and sits somewhere between -120-150dB depending on the setup and test), THD at around 0.0045-0.0035%, IMD around 0.004%, crosstalk around -98-100dB, can do a decent job resolving the -90dB 1KHz sine wave test, etc. One major limitation is that it can only do 48/96KHz sampling rates, at least on the RCA inputs and outputs. The SB1240's line-in might also be "picky" about input voltages, impedances, etc. Not sure...
Clearly, truly good DACs will hit the SB1240's limits, but I've made that clear from the start (I posted the SB1240 measurements as a reference point, though they need updating.). Once I got cleaner results, it was pretty clear the Gungnir was maxing out what the SB1240 could do in nearly all regards. Sweet, the DAC has objectively excellent performance and is likely operating as specified. That's good! Plus, we can learn a bit more about its performance characteristics.
When potential artifacts come up, it's good to pull from other resources. In this case, the Bifrost easily passed the tests on Stereophile, so it's almost guaranteed this is just one scenario where the SB1240 is producing some funky stuff that screws up a few, but not all, results and information. Best we can do is note that, keep the context in mind, and move on.
That said, I've still got a few ideas in mind to see if I can get better looking results. I doubt I'll make much more progress, but the Gungnir has been a good learning experience for me in terms of measuring DACs.