One thing is relaying other is "dont want to see any basic meas"...
And maybe just maybe it can be OK with common users who buying the finished devices. But I spot that not small number of people who build sometnihg, of audio gear, dont even measure the piece.
The idea is to save FFT measurements until you are done. Its just to double check and make sure you didn't miss anything important. IOW the idea is to judge the dac primarily by listening with your ears. Looking at FFTs first is pretty much the same as any other kind of sighted listening. What you expect to hear will be influenced by what your eyes see. That's the problem with sighted listening in general.
Scope measurements, RF spectrum analyzer measurements, DVM measurements, etc., don't seem to influence what we expect to hear nearly so much. Maybe its because FFTs show very tiny imperfections with a very limited type of test signal, one that is PSS. Music is not PSS, so measuring PSS and assuming it predicts system response to non-PSS signals down to very tiny levels involves a big leap of faith.
Maybe the assumption is not so bad for memoryless power amplifier ('memoryless' means nothing like thermal distortion, or frequency dependent distortion occurs). A sigma delta dac is much different in the greater complexity of its operation and behavior as compared to the model of an near-ideal power amp with a slightly curved, but unchanging curvature over time, transfer function. The dac uses amplifiers too, but in the known presence of RF. Plus there are analog voltage and time references, a sigma delta modulator, noisy USB ground currents, etc.
That's my two cents, anyway.
Scope measurements, RF spectrum analyzer measurements, DVM measurements, etc., don't seem to influence what we expect to hear nearly so much. Maybe its because FFTs show very tiny imperfections with a very limited type of test signal, one that is PSS. Music is not PSS, so measuring PSS and assuming it predicts system response to non-PSS signals down to very tiny levels involves a big leap of faith.
Maybe the assumption is not so bad for memoryless power amplifier ('memoryless' means nothing like thermal distortion, or frequency dependent distortion occurs). A sigma delta dac is much different in the greater complexity of its operation and behavior as compared to the model of an near-ideal power amp with a slightly curved, but unchanging curvature over time, transfer function. The dac uses amplifiers too, but in the known presence of RF. Plus there are analog voltage and time references, a sigma delta modulator, noisy USB ground currents, etc.
That's my two cents, anyway.
Last edited:
If I hear something odd, then I would re-measure but I don't obsess over measurements assuming the first set don't show up any problems. (eg crossover spikes).One thing is relaying other is "dont want to see any basic meas"...
And maybe just maybe it can be OK with common users who buying the finished devices. But I spot that not small number of people who build sometnihg, of audio gear, dont even measure the piece.
Remember that commercial products have competitors and there will always be a battle of the specs between them, even though the best specs don't guarantee the best SQ.
Yes offcourse, I think You are right, but i will try to speak as short as I could about...
Comercial products from digital branch have the competion, but also lately, they have one common thing that made competition close to irrelevant. Designers are not from the analog domain but from the industrial and programming area. They does not have enough experiance for analog modules. That is why they putting the OP amps in the analog end... But they have huge praxis in PCB design, programming FPGAs, displays, etc etc... That is why contemporary digital devices looks all the same... They simple saw the chance and opportunity to efficiently step into the market, and make some money legitimetly from own knowledge. BUT from other area of audio... They ususaly accept all of these audio terms frequently used in the internet and incorporated in the advertizing and presenting the models. Like NOS, Gold plated, R2R, Low esr superbrand capacitors, very complex but good looking PCBs etc etc... Acting in hybrid mode of industrial digital/design and advertizing.
They dont doing much of the tests of analog stages, the main tests are of the software running the device. The goals are, short time from start to finished product, small dimmensions, preferably one PCB for all parts. Inclouding all connectors soldered on the PCB...
They dont interested at all how the sound is. Other on the net will acomplish that job...
Comercial products from digital branch have the competion, but also lately, they have one common thing that made competition close to irrelevant. Designers are not from the analog domain but from the industrial and programming area. They does not have enough experiance for analog modules. That is why they putting the OP amps in the analog end... But they have huge praxis in PCB design, programming FPGAs, displays, etc etc... That is why contemporary digital devices looks all the same... They simple saw the chance and opportunity to efficiently step into the market, and make some money legitimetly from own knowledge. BUT from other area of audio... They ususaly accept all of these audio terms frequently used in the internet and incorporated in the advertizing and presenting the models. Like NOS, Gold plated, R2R, Low esr superbrand capacitors, very complex but good looking PCBs etc etc... Acting in hybrid mode of industrial digital/design and advertizing.
They dont doing much of the tests of analog stages, the main tests are of the software running the device. The goals are, short time from start to finished product, small dimmensions, preferably one PCB for all parts. Inclouding all connectors soldered on the PCB...
They dont interested at all how the sound is. Other on the net will acomplish that job...
I agree partly because the traditional analogue audio manufacturers don't have the in-house digital skills and as you say, the outsourced designers are not audio engineers and focused entirely on their digital domain designs.
However, where I disagree is that the audio manufacturer must surely listen to and test the digital protoypes and then ask for changes. Or are they so mystified & in awe , that they just don't question what the digital team has given them?
However, where I disagree is that the audio manufacturer must surely listen to and test the digital protoypes and then ask for changes. Or are they so mystified & in awe , that they just don't question what the digital team has given them?
Some other categories I could add to this thread is from my other thread on Types of DACs (posted elsewhen on DIYA, a few weeks back).
http://www.dddac.de/ma_dac33.htm
- FPGA dac (Chord, etc.)
- Multiple parallel (stacking or paralleling single-IC dacs, like TDA1543) .
http://www.dddac.de/ma_dac33.htm
FPGA can mean most anything. AKM prototyped AK4191 in FPGA. IIRC Scott Wurcer said ESS prototyped their dacs in FPGA (well, maybe not the output resistor array - don't know the specifics). In particular, PS Audio as make a point of classifying dacs as Chip DACs, or FPGA DACs, etc. maybe because their top of the line dac uses an FPGA? Anyway, what an FPGA does may pften be converted to dedicated silicon if it is anticipated that enough units will be sold to justify the development costs.
Multiple parallel may tend to work better with older chips that have higher output impedances. Something like ES9038PRO already has many dacs in parallel inside and its output impedance is low enough to make not so trivial to design around. A lot of opamps can't handle it very well.
Multiple parallel may tend to work better with older chips that have higher output impedances. Something like ES9038PRO already has many dacs in parallel inside and its output impedance is low enough to make not so trivial to design around. A lot of opamps can't handle it very well.
Give each DAC its own op-amp and average their output signals then.
Instead of a parallel connection, you can also delay the input signals to make a FIRDAC and reduce the sensitivity to the phase noise floor, or let them make steps in turn such that you make an n + log2(N) bit DAC out of N DACs with n bits.
Instead of a parallel connection, you can also delay the input signals to make a FIRDAC and reduce the sensitivity to the phase noise floor, or let them make steps in turn such that you make an n + log2(N) bit DAC out of N DACs with n bits.
"Give each DAC its own op-amp and average their output signals then."
Yes, that should be possible. However, IIUC the usual rational for paralleling old style dac chips was to improve SQ. It also may have been possible to avoid opamp I/V altogether. Some people believe not using opamps can also help produce improved subjective SQ. Not sure it would be trivial to accrue the old style dac chip perceptual goals using multiple ES9038PRO along with parallel opamps. Haven't tried it though.
Also occurred to me that someone wanting to reduce noise as a goal might want to try multiple ES9038PRO and parallel opamps, but then what about the differential MFB filter stage? Also parallel that? Going down more or less that type of path may be conceptually trivial, but in practice it might not end up being not so trivial to implement well. Should ASRC be disabled? HD compensation? Less masking of hump distortion by noise a possible unwanted side effect?
Guess I might have been trying to keep may original reply a little more simplistic than it needed to be?
Yes, that should be possible. However, IIUC the usual rational for paralleling old style dac chips was to improve SQ. It also may have been possible to avoid opamp I/V altogether. Some people believe not using opamps can also help produce improved subjective SQ. Not sure it would be trivial to accrue the old style dac chip perceptual goals using multiple ES9038PRO along with parallel opamps. Haven't tried it though.
Also occurred to me that someone wanting to reduce noise as a goal might want to try multiple ES9038PRO and parallel opamps, but then what about the differential MFB filter stage? Also parallel that? Going down more or less that type of path may be conceptually trivial, but in practice it might not end up being not so trivial to implement well. Should ASRC be disabled? HD compensation? Less masking of hump distortion by noise a possible unwanted side effect?
Guess I might have been trying to keep may original reply a little more simplistic than it needed to be?