Discussion in 'Source Measurements' started by purr1n, Jun 4, 2018.
500Hz square wave 44.1kHz.
What kind of power supply is it? Looks like a cheap brick from the product photos. Could you take some measurements with an LPS if that's the case?
It's a cheap SMPS brick. Or "switched PWM" as William Tse put it on Massdrop boards. Can't do $350 with a good LPS.
I'll need to find a 5V LPS.
AC wallwart and on-board linear regulator like Schiit and JDS Labs do couldn't have cost that much more.
On the flipside, this allows people with power supply nervosa to get their upgrade on.
You are asking Airist to do too much. Scouring datasheets and trying different onboard rectifier and regulator chips is beneath those with Ivy League educations.
Would a Wyrd work?
Edit: nevermind, switcher has 3 amp capacity and wyrd is only 500mA
I agree with @Cspirou. If designing any electronics, especially in audio, one should understand power supplies since they are so critical. Most chip regulators already have the documentation on how to implement - they are IC's after all.
This will depend on the frequency of the tone (ie. single frequency sine) and the bandwidth of the critical band that the ear has around that frequency. Also the spectrum of the noise. A simple model of the ear is a series of overlapping passband filters called the critical bands whose bandwidth is frequency dependent, with a rough approximation of its bandwidth being 1/3 octave.
So if the noise is random, then the passband can dig into the noise, so to speak, in the ratio of passband:20kHz. Eg. passband of 200 Hz will be 200/20000Hz, or 1e-2 or 20 dB below the wideband noise. Also that tone has to last long enough for the integration to actually happen. Lots of ifs and caveats as usual.
Yup. That's the thing with FFTs. They are an average of what's going on inside a defined time interval of time domain information (basically the waveform). And successive FFTs are averaged on top of that in order for us not to get a snapshot of random crap, which is always there.
We destroy a lot of information contained in the time domain in order to get a bigger picture more understandable visualization of what might be going on. Analyzer results are not the end all be all. Anyone who claims to be an engineer should know this.
0.05% resistors in a correctly designed sign magnitude R-2R DAC can easily do less than 0.01% THD, my dam1021-05 diy DAC using 0.05% resistors usually have around 0.006 % THD, see FFT plot from my prototype at:
But then there are 0.05% resistors and there are 0.05% resistors.... In RDAC's case it's probably the strange coupling of the two resistors networks though the output filter/buffer that is the problem, or the R and 2R parts are not matched correctly with the bit drivers.....
I was wondering how much the low distortion at low level can be contributed to the sign magnitude approach? Soekris, you are the best guy available to answer that question.
In a non Sign Magnitude DAC with a single R-2R resistor string the distortion is relative to full level, so the distortion goes up as signal level goes down. So the R-2R resistor strings need to be very precise, like the TotalDAC with their ultra precise and very expensive Vishay Bulk Metal Foil resistors, or Schiit with their ultra precise and expensive industrial DAC chips....
In a Sign Magnitude DAC with two R-2R resistor strings the distortion is relative to the signal level, the distortion stays the same when signal level goes down. So the resistors don't need to be that precise, but you need twice as many of them and you need to consider how to connect the two R-2R strings together....
Other design considerations, like clock quality, power, glitches and noise, apply to both types. And then you can do other tricks to improve things, like calibration or dithering, or splitting the R-2R stings in two parts, like what Metrum do or the classic UltraAnalog DACs did.
Where is a Topping Dac when you need one? It seems like this is just another stealth marketing ploy like ASR. Lame.
Thank you Soren. I find the various approaches to R-2R dacs to be very interesting. I know that you believe in sign magnitude and over sampling and I respect that. You certainly get excellent measured performance in addition to good listening reviews.
That said, it would be interesting to compare a sign magnitude dac like yours at NOS in comparison to a Metrum. Or something a Metrum approach over sampled in comparison to your dac. It would be fun to hear expert evaluations and measurements to understand the difference in approach.
Thanks again for explaining the differences.
Why is having ~10.4 ENOB not an issue when playing 16-26 bit FLAC files? How does that translate into the very positive listening impressions? I use a Modi Multibit as my main source which I believe also measured around 10 ENOB or lower. What does having such a low number of bits do to 16-24 bit music? I haven't heard anything discernibly different in comparison with my delta sigma dacs, but I am curious how the math works.
I think it just means that these sorts of measurements and calculations aren't perfect or holistic.
Also, we really need to get past x ENOB. it's just a number. The suite of measurements offer a much more complex picture of the behavior of the RDAC at different output levels. I would suggest taking a look at the measurements and reading in-depth discussions while attempting to get a better understanding.
ENOB of 10.3 for a DAC is like saying a headphone has a THD of 0.2% It doesn't mean much.
ENOB was originally used to describe a specific attribute of A/D converter performance. The way it is now being bandied about for audio DAC performance is misguided and not helpful.
Please take to heart what @purr1n is saying. Forget about ENOB. It contributes little, if anything to the attempt at gaining insights correlating measurements and auditory experiences.
For those willing, please read a fine tutorial MT-003 by Analog Devices here:
I’ll only chime in here to note that the 10 ENOB figure for the Modi Multibit came from Amir and is total bullshit. Marv’s measurements of the Modi Multibit with the Averlab show it around 17.4. Totally suitable for redbook content. Objectively speaking, the Modi Multibit measures much better in essentially all respects than the Airist unit. Subjectively, we have senior members saying the Airist DAC is at least competitive and maybe a little better than Modi Multibit. One of the things SBAF prides itself on is correlating subjective experience with measured data. My comment was mainly pointing out that that is difficult to do in this case. They’re both multibit DACs, but it looks like the measured differences are largely attributable to the RDAC’s discrete topology. Regardless, I wouldn’t touch anything from Airist with a 10 foot pole.
Probably the answer I was looking for?
Q: Does this DAC “upsample” it’s input? Is it “NOS”?
A: The RDAC up-samples the input but the digital-to-analog conversion does not oversample. Thus, it can be considered a “no-oversampling” DAC even though there is “upsampling” at the digital inputs (with an FIR filter to reduce noise through input processing before D-to-A conversion).
Separate names with a comma.