REALLY? Other people tested the passives. They get 80-90dB SND... Probably TDH+SND dB are 70's...
Yes, when the THD+SNR is at -70db that it is a problem. You can say that measurements don't count, but the truth is what measure good, has good chances to sound good. And what measure BAD, sounds BAD - no matter what.Still of the view that THD is one of the most important measures for a DAC's sound quality? Good luck with that
Last edited:
We call it "begrijpend lezen". In the link you gave to the document no comparison was made, You stated that opamps were tested by ESS and that they outperform anything that passives can do while passives are not even mentioned in the same document. So you are mixing your opinion/experiences with the document from ESS. This is simply creating your own truth.
I am not saying you are wrong, the way you try to convince is just not right.
"probably"... I have to see a document with measurements to believe something. Till now ESS only advised which opamps to use with their standard/reference opamp DAC output stages. No other possible output stage was mentioned nor criticized.
Last edited:
maybe not, but opc tested passive with the ES9018 and it measured similar to excrement at 90-100db dnr (actually the number is becoming clearer maybe it was only just over 90 but THD+N) its in the development thread, not the build thread. has the most comprehensive independant testing of the sabre ive seen published
that being said, this was an undertaking to find the best mosfets for the NTD1 for sabre and the result was THD+N ~-115db so i dont mind forgoing the opamp in this case and i'm a friend of the opamp
that being said, this was an undertaking to find the best mosfets for the NTD1 for sabre and the result was THD+N ~-115db so i dont mind forgoing the opamp in this case and i'm a friend of the opamp
Last edited:
I (and the rest of professional enginering comunity) say that OpAmps are best to use for the audio DAC conversion and the OpAmps provide the best results that we are aware of. Measurements are published, data sheets reflect that conclusion (all the I/V stages, of all the manufacturers, are OpAmp based)
I saw measurements published for the passive (by DIY-ers) and they are worse. So I feel that my part of QED is done.
You say that passive conversion (with tubes probably) is a better solution. But you don't provide any proof for that and you require me to disprove your affirmation.
In the "normal" engineering/scientific world, you would need to provide those measurements, since the technical ideea that you advance is yours.
Last edited:
No, I did not say (nor write) what you write I would have said. Simple. You are the one that states that opamps are better without any similar proof that passive solutions can be worse....
It is funny to see you say I would need to provide those measurements while you made the statement that passive solutions are worse not backed up by any data (like the ESS document concerning opamps).
Since I already gave the reasons why opamps are the common solution I will not repeat myself. At the same time I wish not to debate your and the professional engineering communities findings. All I can say that I am glad that not everybody goes the beaten track.
It is funny to see you say I would need to provide those measurements while you made the statement that passive solutions are worse not backed up by any data (like the ESS document concerning opamps).
Since I already gave the reasons why opamps are the common solution I will not repeat myself. At the same time I wish not to debate your and the professional engineering communities findings. All I can say that I am glad that not everybody goes the beaten track.
Last edited:
Hi this is interesting. I am on the ADP151 route myself and thought the opamps solution ESS suggest worth trying out. I now see that it is not ideal.
Maybe other things start to play a role with such low noise numbers, I begin to wonder if the human ear can even hear such differences at all.
Last edited:
Depends on the end to end system.
A good stress test is a high efficiency horn combined with an listener whose ears are 10dB better than normal (there are probably folks whose hearing's even better than that, but my guess would be +10dB is already getting into the tail of the distribution). If you figure the horn's at 105dB efficency and the listener is testing the system by putting their ear in the horn throat that implies the threshold of audible input power to the compression driver is about 30fW (assuming the ambient is good and quiet), or roughly 500nV for most drivers. This is obviously an extreme example; with more typical speakers and listening arrangements one can back off by 40dB or more. Unfortunately using a typical power amplifier with about 30dB gain consumes, well, 30dB, that margin and one ends up needing a noise floor below 10uV or so on the amplifier input. That's about the DNR of a typical 19 bit DAC. For normal hearing one can take another 10dB off---the better part of 2 bits---implying a 17 bit DAC's sufficient. That's fairly easy to work with given a decent analog background. Though arguably somewhat pointless unless there's Squeezebox type volume control producing 24 bit output so that bit 17and beyond can be used---my experience with 16 bit data paths is I end up measuring a THD floor around -50dB. Sounds crappy.If you can reduce the voltage gain from the DAC to the driver to unity things get easier; the threshold of audibility with typical ears and typical 90dB SPL/W sensitivity is around 1mV. That translates to about 1W peak per amplifier channel; sufficient for most purposes.
The ESS parts still differentiate on jitter rejection, ASRC design, and stuff like whatever it was Dustin was referring to when he was talking about improved state variable tracking during transients. From a theoretical standpoint I would not expect a playback system where some attention was paid to the configuration of gains in the signal path to exhibit an audible difference between the ES9008/9018 (-119dB THD, 129dB DNR, give or take) and the ES9016 (-110dB THD, 124dB DNR). There might be some audible difference to the ES9006 (-102dB THD, 120dB DNR) but as I have none of these DACs available to listen to that's purely conjecture on my part. At least until I get boards built up to A/B between the 9006 and 9016. That's still some distance in the future, however.
Continued the AVcc discussion over here since power supplies are kind of out of scope to this thread.
Well, a transformer is a great LP filter. It also does other things very well, like common mode noise rejection, RF filtering, ground isolation, balanced to single ended conversion, etc.
But, I'm open to other solutions.... not trying to be narrow minded.
An integrator != a low pass filter. The latter have cut-off frequencies.
In "AC world" an integrator is just a low-pass filter.
As for the affirmation that an OpAmp is "an integrator", that's not mine, I was trying to combat it saying that any low-pass filter is in fact an integrator, so that wold not be something "bad" as it was trying to be portrayed.
As for the affirmation that an OpAmp is "an integrator", that's not mine, I was trying to combat it saying that any low-pass filter is in fact an integrator, so that wold not be something "bad" as it was trying to be portrayed.
The fact it is that a transformer is NOT a LF filter. Is a pass-band complex filter, imposible to adjust and on top of that with neliniarity issues. The worse is that it has an input resistance/impedance that is higher than the one necessary to be seen at the DAC current outputs.
Last edited:
the Gilbert article is pretty much an electronic feedback theory "intelligence test" - most citing it here are simply looking for any support for bashing negative feedback - without in fact understanding it
and yes I will state that from what I've read of Charles Hansen's posts he either doesn't deeply understand feedback theory or is deliberately misrepresenting it for commercial reasons
the example Gilbert uses is basically like using a 741 class op amp for 10x gain at audio with 10 V output - a deliberate bad choice to show the potential problems with misapplication – op amps with insufficient loop gain aren’t completely linear – so what – we can choose better op amps for the task
we can use 50 MHz op amps today, with "more linear" jfet or Analog's new linearized input stage (AD8099, ADA4898 family)
if you understand Gilbert's article you should be able to figure the order of the improvements from greater GBW, with more linear input diff stage, and lower output V swing (with the typical lower consumer line level - most audio DAC circuits use 2 Vrms fullscale)
(hint: Gilbert's undegenerated bjt example input tanh distortion is primarily 3 rd order - proportional to the 3 power of the diff input V - 50x higher GBW/signal ratio gives 2500 x lower input tanh distortion ratio )
DAC I/V is still challenging for op amps due the switching step/glitch edge rate - puting a lead C to give the 1st pole of the anti-image reconstruction analog low pass filter in the I/V feedback is a help
with the feedback C it then becomes important to have low output Z - while flagship audio DAC manufacturer's demo circuits still use 5534 it is poor in this regard - modern isolated junction, complementary processes with faster, more nearly equal PNP, NPN Q are much better
Analog's new linearized input stage op amps treat the input signal linearly for much larger diff input V - and are fast, have high output current (ie could be heavily Class A biased) should be a real advance for op amp I/V
then there is also the possibility of advancing op amp I/V performance with Hawksford's multiloop topology idea
http://www.diyaudio.com/forums/digi...sford-iv-nested-loop-op-amps.html#post2218591
(I've since found that the op amp output Z isn't remotely properly modeled by the manufacturer's models - the circuit would need some bench work with a 300 MHz 'scope before recommending it generally but I think it shows a largely unexplored way forward)
and yes I will state that from what I've read of Charles Hansen's posts he either doesn't deeply understand feedback theory or is deliberately misrepresenting it for commercial reasons
the example Gilbert uses is basically like using a 741 class op amp for 10x gain at audio with 10 V output - a deliberate bad choice to show the potential problems with misapplication – op amps with insufficient loop gain aren’t completely linear – so what – we can choose better op amps for the task
we can use 50 MHz op amps today, with "more linear" jfet or Analog's new linearized input stage (AD8099, ADA4898 family)
if you understand Gilbert's article you should be able to figure the order of the improvements from greater GBW, with more linear input diff stage, and lower output V swing (with the typical lower consumer line level - most audio DAC circuits use 2 Vrms fullscale)
(hint: Gilbert's undegenerated bjt example input tanh distortion is primarily 3 rd order - proportional to the 3 power of the diff input V - 50x higher GBW/signal ratio gives 2500 x lower input tanh distortion ratio )
DAC I/V is still challenging for op amps due the switching step/glitch edge rate - puting a lead C to give the 1st pole of the anti-image reconstruction analog low pass filter in the I/V feedback is a help
with the feedback C it then becomes important to have low output Z - while flagship audio DAC manufacturer's demo circuits still use 5534 it is poor in this regard - modern isolated junction, complementary processes with faster, more nearly equal PNP, NPN Q are much better
Analog's new linearized input stage op amps treat the input signal linearly for much larger diff input V - and are fast, have high output current (ie could be heavily Class A biased) should be a real advance for op amp I/V
then there is also the possibility of advancing op amp I/V performance with Hawksford's multiloop topology idea
http://www.diyaudio.com/forums/digi...sford-iv-nested-loop-op-amps.html#post2218591
(I've since found that the op amp output Z isn't remotely properly modeled by the manufacturer's models - the circuit would need some bench work with a 300 MHz 'scope before recommending it generally but I think it shows a largely unexplored way forward)
Last edited:
Yes; IMO this is one of the reasons for choosing an output buffer topology which inserts an RC lowpass between the DAC and the output buffer, such as second order MFB. ESS gets pretty good results without an RC between the DAC and the op amps in their current mode topologies though. The onboard IIR filter in the ES900x and ES901x parts seems to do its job.
Because of the Nature of cfb opamps, they are the only choice when dealing with highspeed glitches
To me, newer OpAmps like the AD8099 or LM4562 are plenty sufficient for the today DAC's I/V stages. DAC manufacturers agree also with that.
There is no "improvement" that a passive I/V conversion can add. Personally I saw only degradation of parameters in the few measurements posted. If anybody claims different results, it should back that claim with some data so we can compare it with the data published for OmAps.
There is no "improvement" that a passive I/V conversion can add. Personally I saw only degradation of parameters in the few measurements posted. If anybody claims different results, it should back that claim with some data so we can compare it with the data published for OmAps.
Just a comment about what DAC vendors believe constitutes sufficient performance. The AD1865 DAC chip is usually used in current output mode, but it also contains an on-board I/V op-amp. The Analog Devices datasheet takes the time to specifically mention that this on-board op-amp was designed to handle the high slew rates it would be subjected to, even at an x16 oversampling rate. They even go as far as to mention that this internal op-amp features an all NPN output stage.
Now, AD knows as much about high-performance op-amp design as anyone. So, after experimenting with the AD1865, non-oversampled, in both it's current (simple resistor i/v), and voltage output modes (internal op-amp), I can report that, sonically, there's no comparison. The current output mode via simple resistor i/v is so much more transparent, you'd think that you were listening to two entirely different converter chips.
Last edited:
Yep, AD engineers are a bunch of monkeys. They got the plans for their 1990 DAC from martians and they just slapped a stinky OpAmp inside.
That by the way was absolutelly appropiate for the 16 bit signals of the era.
The use external OpAmps AD797 in the datasheet of their today flagship AD1955 is just another conspiration...
That by the way was absolutelly appropiate for the 16 bit signals of the era.
The use external OpAmps AD797 in the datasheet of their today flagship AD1955 is just another conspiration...
Last edited: