CHANGSTAR: Audiophile Headphone Reviews and Early 90s Style BBS
Lobby => Amp and DAC Measurements => Topic started by: Hands on December 26, 2013, 04:43:31 PM
-
I wanted to post more measurements of this unit for completeness, since it is currently what I use as my ADC when measuring other gear. As such, any other measurements I do using this as the ADC will be limited by this unit's performance, and these measurements can act as some sort of reference for what those limitations are. And, since this is a loopback that tests the DAC and ADC in the unit, ADC performance by itself should only be better than what is seen in the context of these measurements.
This post also shows some of the new approaches I'm taking to measurements compared to the ones I did for the Metrum (which I hope to revise), mostly when using ARTA. Jitter measurements are now included using JTest files. Spectrum measurements should be much more informative now, and I've also decided to not follow up with further sine wave response measurements on DACs (details below).
Do note that when looking at my measurements, I do NOT always set scales and zooms to match 1:1 across measurements. For example, with impulse responses, I'll scale and zoom to get the best depiction of the information I can.
Refer to the Metrum Quad DAC measurements thread for any other detail I might have missed.
I've attached a ZIP file of what should be all the tests I have of this unit, some of which I did not post on here.
Test Hardware and Setup
1. Acer Aspire 5560G Laptop - Locked at 1.4GHz and 0.9125V via K10Stat, dedicated GPU disabled, internal fan disconnected (low temps and stability verified), most Fidelizer tweaks applied, and other hardware, software, and services disabled.
2. Creative X-FI HD USB (SB1240) - External loopback tests done with BJC 6' RCA cable, powered by USB bus, connected to Acer laptop. Unit is limited to 48/96KHz input/output.
Software
Rightmark 6.3.0 and ARTA
Links
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/
-
JTest 24/48 file with hardware and software set to 24/96 (did this to match my NOS1704 tests, did not see a difference running hardware and software in 24/48). Did my best to match Marv's latest jitter measurements. Used JRiver Media Center for JTest file playback (no software upsampling).
-
1. 1KHz Sine Wave Spectrum at 24/96
2. 100Hz Sine Wave Spectrum at 24/96
3. 60Hz and 7KHz Dual Sine Wave Spectrum at 24/96
4. 25Hz and 66KHz Dual Sine Wave Spectrum at 24/96
5. Noise at 24/96
-
1. Periodic impulse response. Notice the bit of post-ringing relative to a NOS DAC.
2. 1KHz square wave response. Also shows additional ringing vs. a NOS DAC.
3. 20KHz sine wave response. This shows the same characteristics as the Metrum Quad NOS DAC at 20KHz, and I'm not knowledgeable enough to explain what's going on here. Thus, I'm not going to be running sine wave response tests for the foreseeable future on DACs.
4. FR and distortion graph generated by a sweep impulse response. Much like the Metrum DAC, no idea if this is value at all or not but figure it won't hurt to post.
-
For completeness and convenience, I will post the 16/48 and 24/96 RightMark measurements.
Frequency response (from 40 Hz to 15 kHz), dB | +0.02, -0.08 | Excellent
Noise level, dB (A) | -92.2 | Very good
Dynamic range, dB (A) | 87.8 | Good
THD, % | 0.0028 | Excellent
THD + Noise, dB (A) | -82.2 | Good
IMD + Noise, % | 0.011 | Very good
Stereo crosstalk, dB | -89.4 | Excellent
IMD at 10 kHz, % | 0.012 | Very good
General performance | Very good
Files attached:
1. Frequency Response
2. Noise
3. Dynamic Range
4. THD
5. IMD
6. Crosstalk
7. Swept IMD
-
RightMark 24/96
Frequency response (from 40 Hz to 15 kHz), dB | +0.01, -0.08 | Excellent
Noise level, dB (A) | -98.4 | Excellent
Dynamic range, dB (A) | 98.6 | Excellent
THD, % | 0.0016 | Excellent
THD + Noise, dB (A) | -89.5 | Good
IMD + Noise, % | 0.0045 | Excellent
Stereo crosstalk, dB | -90.9 | Excellent
IMD at 10 kHz, % | 0.0049 | Excellent
General performance | Excellent
-
1. Periodic impulse response. Notice the bit of post-ringing relative to a NOS DAC.
2. 1KHz square wave response. Also shows additional ringing vs. a NOS DAC.
3. 20KHz sine wave response. This shows the same characteristics as the Metrum Quad NOS DAC at 20KHz, and I'm not knowledgeable enough to explain what's going on here. Thus, I'm not going to be running sine wave response tests for the foreseeable future on DACs.
4. FR and distortion graph generated by a sweep impulse response. Much like the Metrum DAC, no idea if this is value at all or not but figure it won't hurt to post.
You're plotting the samples with linear interpolation here and you're bandlimiting the output of the DAC with the ADC filter.
1. More post-ringing means that the filter is not a linear phase one but a minimum or mixed phase one. Problem here again is linear interpolation between sample values. The real impulse response of the DAC looks quite a bit different.
Maybe you should plot the spectrum (magnitude and ideally group delay) of the recorded impulse, so people can see what the filter is really doing.
2. Again, maybe show the spectrum. The ringing has to be there to satisfy the sampling theorem.
3. The problem is that you're not looking at the analog waveform the DAC is outputting and again the linear interpolation between sample values.
You'd need a much higher sampling rate with the ADC to see what the DAC is doing.
If the spectrum shows that the 20 kHz sine wave is output at full level then chances are that the DAC isn't doing anything fishy.
4. FR is very important, as is THD.
-
OK, thanks for the explanations. I thought maybe it was something related to that based on our previous discussions, so that's good to finally have a better grasp on. I'll have to figure out how to do spectrum measurements with these. I toyed around with some options but am still learning most of this stuff and couldn't figure it out exactly. If you or anyone else has tips on how to set that up, I'd greatly appreciate it and will try to revise my measurements. Any other suggestions or revisions I should make? I do get that I'm fairly limited with my hardware and software at this point (I'm considering a nicer ADC for the future).
I understand that FR and THD are important, but I can't say I understand exactly how those measurements are done. I just know there are different approaches that can be taken for measurements, hence me not being sure if the sweep impulse method is valid for DAC FR/THD measurements. I'm just a bit ignorant is all. 8)
-
I'm not that knowledgeable in ARTA and RMAA. I had problems calibrating ARTA, and could only work with it at 96k/24bits w the 2i2. RMAA only at 44 & 48k. I'll try to figure it out and do better myself.
However, one other thing that could be happening w the sine wave capture is that 96k is not an integer multiple of 20k. So one might get samples at different phases. I think the LCM of 96 and 20 is 480, which is 24 times 20. So about every 24 samples, things might repeat.
The sawtoothness is indeed likely the result of linear interpolation. A higher order interpolation approach (cubic or splines), or upsampling w interpolation filters to the video resolution might yield better visual results at the expense of complexity. I'm not sure, but don't think ARTA and RMAA offer these interpolation alternatives.
-
will post my set of measurements this weekend of the same unit. It is my main dac till I get something in the mid tier :/
-
However, one other thing that could be happening w the sine wave capture is that 96k is not an integer multiple of 20k. So one might get samples at different phases. I think the LCM of 96 and 20 is 480, which is 24 times 20. So about every 24 samples, things might repeat.
The sawtoothness is indeed likely the result of linear interpolation. A higher order interpolation approach (cubic or splines), or upsampling w interpolation filters to the video resolution might yield better visual results at the expense of complexity. I'm not sure, but don't think ARTA and RMAA offer these interpolation alternatives.
Ah, OK. I'll have to try something that's a multiple of 96KHz to see if that at all improves things. Might fiddle around with other settings.
-
It might be hard to get the phase aligned to the max amplitude. Still, I would give 12kHz and 24kHz a try and see what comes out of that.
What you showed makes sense to me though.
One thing to try in your measurements is to check the voltage levels. What I do is send a tone and adjust the knobs until I measure 1 volt (rms) or whatever level I target. Make sure the tone level is the same as the measurement level.
-
Yeah, that's another step I've considered, checking voltages. I'd have to get a multimeter.
It can be a real PITA to get everything matched for these tests across the DAC and ADC. Getting every test setup properly takes the vast majority of time when measuring, I've found. Depending on which mode I'm testing, setting either of the devices to 100/100 volume in Windows (only method I have for volume/gain control outside of test gain settings in ARTA) creates additional noise or distortion that really isn't going to manifest itself in proper listening conditions. I really doubt either this Creative unit or my NOS1704 DAC have 9% THD, so sometimes you gotta bump that volume slider down on both devices. (For my Auzentech Bravura card in particular, setting Windows volume to 100/100 will actually produce a "hot" signal that causes distortion and clipping on most devices.)
So, I have to find some sort of balance that gives proper levels but without falsely displaying better or worse measurements than in reality. Often times, I will test a large sampling of volume settings across devices that still match up to the same level for measurements in the end, compare results, and tend to go with whatever settings give the most consistent, repeatable, and rational values. Generally I find it works well to match the volume settings on both devices in Windows that give proper levels in measurement software. I know these methods aren't as scientific as they should be, but I do strive for getting the best, most repeatable, most consistent results I can with the methods and technology I'm using.
I think despite all these limitations, there's a lot of good, solid information that can be gathered from these measurements. Plus, I'm always willing to go back and revise my methods and measurements if I learn something new/better. :)p7
-
I will probably skip on posting impulse response measurements in the future. With my lack of understanding and relatively limited hardware/software, it's hard for me to get consistent, repeatable results (i.e. I've seen NOS DACs show ringing with impulse response and OS DACs show no ringing, for example, when playing certain tests a few times back-to-back). If I get the right stuff and knowledge to produce good impulse response measurements that contain valuable information, I'll do so.
However, I am still toying around with spectrums generated from impulses, and my preliminary tests (aka I have no idea what I'm doing, so I just push da buttons and change da numbers) are showing different results from the NOS1704 and SB1240. I just have no idea if they're set up right or provide valuable information, so I'll look further into that.
I could have sworn I saw someone who did something similar with a DAC...can't find it now.
-
xnor or anyone else knowledgeable on the subject, these loopback measurements were done with REW at 48KHz (max I could do). Are any of these tests and results informative, valuable, or valid given the context of DAC measurements? I ran the same tests with the NOS1704 as well (using SB1240 as ADC). Still open to other suggestions to try.
Apologies if the colors make it hard to read...one minute the software used one color scheme in dark mode, and the next it's using another. No idea why.
-
I think the frequency response one might be useful in double checking your results with the other software. The rest might be more useful in speaker characterization.
You could also use REWs RTA spectrum analyzer to double check THD results. Unfortunately AFAIK REW cannot generate arbitrary multiple tones for IMD currently like ARTA.
-
Would like to post more measurements but I'm not sure they are valid, my unit is measuring weird, I'm sure I borked my ADC after amp testing without zener diodes.
(http://i.want.tf/to/that/xpaVjPc.png)
-
I see you're using ARTA, but what steps did you take to get that measurement? I could see how mine checks out compared to yours. Also double check all your settings in software, sound device properties, etc. Make sure you got the bitz and the hertz all matched up. I have gotten weird results before that just happened to be mismatched settings and properties.
-
This one is done using periodic sweep, I have repeated numerous test using Steps(I had weird distortion figures -150db THD @20khz. Used the frequency response and voltage functions of my multimeter(Fluke 8060a) to check my DAC outputs for the correct voltages (which showed extended response to 30khz(did not check further)) . Also done a spectrum using the FR function and measurements in RMAA.
(http://i.want.tf/to/that/5gjMTBJe.png)
Impulse measurements for the lols, mine is the textbook linear phase DAC. Since my ADC is borked(I think) I doubt its valid. (Click below)
(http://i.want.tf/to/that/TERij.th.png) (http://i.want.tf/to/onL)
Need to take a look at my soundcard settings again but I'm at least sure my DAC has nothing wrong(or I need to get a new DAC/ADC)
-
Can the measurements be done at 96kHz sampling rate? For RMAA I had some problems w my 2i2 at that rate.... It would freeze the program for some reason.
Actually, your measurements seem pretty close to Hanns to me. Good low and high end extension. The upper end ripple may be due to a sharp low pass filter somewhere, and not necessarily from the DAC. I used to get it for some reason when measuring w my laptop line-in using the Wanckdose drivers.
-
Oh wait, yeah, thats just my ADC operating at 48khz facepalm , will post the results now, sorry about the scaling :X. I still fine the ringing in my unit a bit funny compared to Hanns, could it be the difference in a Fidelized setup or can it be disregarded? (his has more post and mine has quite a bit more pre-ringing) The IR I posted had the gain upped quite a bit to emphasise the ringing though. Jitter measurements later, this unit is good at that though, using a form of async usb before it was cool.
Rig: Fujitsu laptop, no Fidelizer or whatever, some services disabled for performance as per Sweetwater guide. Running DAC at 24/96 with ADC at 24/48. Output voltage referenced to 2.05Vrms@1khz
(http://i.want.tf/to/that/qwvNVtApIB.png)
FR:
(http://i.want.tf/to/that/2sKDhmj.png)
Noise:
(http://i.want.tf/to/that/Hu8yM9.png)
Dynamics:
(http://i.want.tf/to/that/XyOilpGK.png)
THD + Noise (at -3 dB FS)
(http://i.want.tf/to/that/bCAiZ.png)
Intermodulation distortion
(http://i.want.tf/to/that/1T4Frej9N.png)
Stereo crosstalk
(http://i.want.tf/to/that/cEB6TJMG.png)
IMD (swept tones)
(http://i.want.tf/to/that/X7ONuR.png)
For round off, Stepped Sine distortion testing, ref 2.05V@1khz (volume 94 in windoze)
(http://i.want.tf/to/that/j8e24NB.png)
-
I wouldn't worry about the impulse response or ringing. I'm not sure the way we're testing with ARTA is good enough to get accurate results. I ran all of the IR tests several times, and they'd always vary in the amount and type of ringing I got. Even the NOS DACs would show ringing occasionally, just much less often than the SB1240. I'm sure some of it is improper settings, but most is just that we'd need better hardware and software to really get a good look at them. Of course, I could be wrong...perhaps we have the tools but just our methods and settings aren't right.
So far, I've had better luck checking ringing with 1KHz square waves. Even those can be hit and miss, though.
Your results look good...better than mine in many ways. Your laptop could be better, and I also didn't use a multimeter to check my output voltage. I could probably get better results that what I've posted if I toyed around more with the volume settings, which I've already done with the Metrum and NOS1704.
If your ADC was also at 96KHz, your FR would smooth out and look more like mine measured. I've seen that before when my sampling rates didn't match.
This truly is a solid little unit. Performs well and has some good features for the price, which can go as low as $60 at times. I've heard the HPO is a bit weird and might have a high output impedance, so it's not quite perfect. Built well, looks nice, easy to setup and use. If only it could do 44.1 and multiples of that and go above 96KHz. I think you'd be looking at the EMU-0404 for that.
I'll need to start putting my RMAA results in table format. Sorry for hurting your eyeballs, everyone! I might start expanding more tests as well...the THD tests are pretty interesting, the ones broken down by D2, 3, 4, etc. I'm not sure how accurate they are, but it's more useful than just THD by itself. I'm curious if any of my time-based measurements (like spectrogram) provide any valuable information...
-
Thanks for your help Hanns! Figured out the spectrum analyser functions. So here is the Jtest, don't know much about windowing functions though. Will be measuring the Modi I have on hand and probably the humble Ipod 4g given the time. Test as usual ref to 2.05V, my test are of 3db for some reason though. Test signal is generated by ARTA. Weird, I have plenty of sidebands. This time I checked for matched sampling rates though(24/48 DAC and ADC).
(http://i.want.tf/to/that/w2tdg5WNb.png)
EDIT: test signal used https://drive.google.com/file/d/0B4AK9zritapcZFNKM040NnBiT0k/edit?usp=sharing
-
You said you generated this in ARTA? I had a separate program that can also generate JTest files, but I was lazy and used pre-made FLAC files I found somewhere on Head-Fi. I'm curious if that would at all have anything to do with our differences, either the different generation methods (shouldn't assuming both are JTest) or the file type.
I always try to keep my tests at or slightly below 0dB in tests. Going over that was more likely to produce ill effects. If your output is at 2V you could try lowering the A/DC volume.
If I had to guess, my measurements probably show less sidebands because of all the tweaks I did to my laptop and because I run it off the battery for tests and listening. OR I messed mine up. :)p6 I could retry everything set to 24/48 instead of matching hardware/software at 24/96 and just playing the 24/48 file. I didn't see any differences the first time I tried, but it can't hurt.
Perhaps I'll see how the SB1240 does with jitter and other measurements when fed through my gaming rig (without audio tweaks at all). I think I will do tests in the future with the Hanning and Kaiser windows. I think other sources of jitter measurements I've seen use Hanning or maybe flat-top, but I do like the additional "clarity" in details you get from the Kaiser windows.
All said, though, those sidebands are still below -120dB, so jitter is still quite low.
-
I think flat top is the most representative, some of the other windows like Hanning sometimes can smooth out the pickets. I tried lowering the volume a little to no avail, somewhat reduction in sidebands though. The result is still pretty darn nice for a 100 buck thingy back then.
-
Well, I just discovered something must be up with my laptop that I am not aware of. I just did a 24/48 RMAA test running it from my gaming rig. No sort of audio tweak applied.
It measured noticeably better in all regards except for THD and IMD, which were a bit higher numerically and weren't as clean. Ahhh, WTF.
Also, firev1, I tried your JTest file, and my SB1240 also now shows jitter sidebands. The FLAC file I have does not. I'm not sure if the conversion from WAV to FLAC caused this, if it is something about the playback of FLAC, or if the file I had is just junk.
Also tried various software with the WAV file, and you can see noticeable differences in jitter results among the software.
Rather than re-do my measurements, just take everything I've done before and assume the noise floor is at least 5-10dB lower. Maybe cleaner looking in some spots, but no worse. For example, I was able to squeeze out -107dB for noise like you were running it from a front-panel USB port. I think my other tests didn't break high 90s.
I think this also shows that I definitely need to test the JKSPDIF with different computers.
-
You're going to get unreliable data if you use directsound in windows. ASIO is the only acceptable format with WASAPI coming up second. The differences are not small especially in the realm of DAC testing.
-
Yes. Windows drivers can eff up measurements... and sound quality. I use ASIO whenever I can. Wished Audacity for Dirtdoze supported ASIO w out having to recompile it. Heard good things about Linux ALSA.
-
Indeed, all test were done with WASAPI though. I get repeatable and good results with lower FFT samples at the expense of being unable to fully resolve the 24 bit noise floor, decent for most measurements since skirting and pickets with still be observable.
-
I've used ASIO, WASAPI, or kernel streaming when listening for a long while now, which is the most important part in the end. Let me see what I can do about ASIO and RightMark...it was not working properly on my laptop. I did run Kernel Streaming measurements when I could, like with the JK in the chain. Only spent maybe 5 minutes with the SB1240 measurements on the desktop. I will also double check how I had ARTA setup.
That said, I'm not sure DirectSound would account for the difference between the laptop and desktop in regards to, say, noise measurements, but it could be possible. I'd love to learn more about how that might happen.
I seem to get consistent, repeatable results when running the same sound mode. Moving to ASIO or KS generally gives quite similar looking measuements overall but with better values. I think for a basic spectrum view it's not a big deal unless huge differences become apparent, but something fine grained like jitter measurements would probably see more benefits with ASIO, if I had to guess.
At least, so far this is what I'm thinking and seeing with measurements, and I do a lot of testing and experimenting outside of what I post (though clearly not as much as I could be, but time is limited).
I will re-do my measurements with that JTest WAV file since that FLAC copy I had wasn't doing me many favors. We'll see if it affects the other DACs the same way it did the SB1240.
Edit: firev1 posted that while I was still typing. Did you also mean your RMAA tests were done with WASAPI?
-
Oops, RMAA ones were done with MME, windows, only the jitter ones were with WASAPI :3
Anyways, update: I have some measurements for Modi but I'm really uncomfortable with how it sound and measured. I suspect the one I have to be broken as it deviates much away from the SR analyzer measurements.
-
Update to the thread, new measurements done or revised measurements with the correct units or scales etc. Using mainly ARTA as RMAA is barely usable for the way I'm doing my test. The 16 bit sine looks kinda funny though I admit.
-
Had to use separate ARTA windows to get this out as ARTA uses 16 bit for the recording side when doing wave settings as well which is why I'm getting the weird 16 bit graphs.