Soekris DAC2541 Review And Measurements

I have measured exactly one DAC that truly has 140 dB of dynamic range, which translates to 23 Effective Number Of Bits (ENOB). Repeat, 23 bits. While chasing bit depth people tend to forget the demands on a power supply design to achieve such a significant level of dynamic range. If a DAC produces 4 Vrms at 0 dBFS, then 140 dB down is 400 nV rms. Power supply noise and the reference for the R2R ladder must be less than this. In technical terms, this is RFH to accomplish.

So 24 bits in the delivery format makes sense. But 32 bits is more convenient given most data transports tend to be either word (16 bit) or double word (32 bits). Rather than go to the extra effort of packing 3 byte data transports, it is easier to send double words. Now who wants to hear that bandwidth and storage are being wasted by 25%? Enter marketing. Turn that negative into a FEATURE. We have MORE bits than competitors. Never mind the science doesn't support it actually being useful.

In the studio during inter-stage operations bit depths of 64 bit and even higher are useful to avoid rounding errors and other mathematical manipulations from creeping into the final product. But at the final product delivery 24 bits are the limit of usefulness.

 

"quite challenging" that's some stiff-upper-lip British nobleman levels of understatement.

On a 24 bit recording, if the levels are set so that the ceiling is somewhere near the loudness you'd feel standing next to an airliner's engines when they're at full thrust, bit 24 would be recording quantistic thermal noise in the atoms in the microphone.

Bit 21/22 are infrasonic noise like water going through pipes somewhere a few rooms over from the recording studio.

Theoretically possible to record meaningful information there if you were to suspend the entire recording room from the kind of heroic isolation systems used to calibrate the kind of scales that are then used for to make the instruments that detect gravitational wave anomalies.

It is, however, theoretically possible to record - or generate - music with 18/19 bits of meaningful data.

Not practically possible, unless you do no mastering, but in theory.

And since our computer systems like multiples of 8 bits, we have 24.

But if anyone here is ever thinking to themselves that they could hear the information in bit 17+ anywhere but in an anechoic chamber, I have some property in Times Square I'm looking to get rid of, for cheap.

Serious callers only.

 

Thanks for taking your time measuring.

Hmm, are you sure you're not getting into slight clipping at the 0dB measurement ? I measured all harmonics to be below -100 dB, typically around -110 dB, except the 2nd and 3rd which stay around -100 dB....

 

Dunno if it's relevant, but the DAC1421 loaner clipped occasionally for me with music when fed S/PDIF and set at 0dB.

Ha, you liked the soft min filter - don't you usually prefer linear phase? Most who tried the 1421 (including me) reported they preferred soft min too. House sound again, I guess?

 

You know what, the distortion/noise seen here actually reminds me a LOT of the Sony TDA1541A. 2541 is on the top of my list now, for sure. It looks like quite the awesome ladder DAC.

 

purr1n,
Yeah, the -1dB FFT looks exactly like it do here in my testing.... Don't know why the 0 dB clips, I test all units at 0 dB (and -30 dB and -70 dB) without issues, and I actually have around -0.8 dB margin....

I'm just worried if somebody see the plot with clipping, and they drop interest in the dac2541....

You might also want to increase the FFT window on the multitone test, then the low freq will also look good...

Did something go wrong with the Linear and Min Phase Mix White Noise test, seems like it hit a floor at -85 dB ?

 

Agree, I usually have 0.1% THD as limit to what you can hear.... I'm just trying to get things perfect, maybe it's just me....

 

Mine is hopefully going to get through customs and down to me in time for the weekend. I'm looking forward to listening to the sound, not the measurements.

Out of curiosity @soekris , is 0 still actually -1dBfs in this firmware?

 

What is your chain? I am always on the look something that improves it but the 1541 was not it.

 

DAC1541-->Genelec 8030B+7050B with DSP Chain: DMG Equilibrium running Sonarworks room compensation curve-->SPL Twintube-->DMG Essence