• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Best spec ADC Chip currently.... ??

HpW matches the AP THD+N values near perfectly. There are some simple reasons for that once you know them. Unfortunately I can not disclose these.

Trade secret, or some contractual obligation with AP for some reason?


For me, the only interesting part is the hardware. All I'd need is a glorified sound card. I can easily write software to do exactly the analysis I want. In fact, I prefer to do that over using some pointy-clicky turkey solution.

Surly you've taken a look at the costs of getting someone, to have knowledge/ability like you + to have the wherewithal to make reliable software that tailors to an entire consumer base, but also a consumer base that requires versatility on a fairly technical level, but not technical to the degree where they would require the knowledge in software programming itself?

So when chris mentioned you're not even remotely considered their customer, that would be an understatement. Made more true by the fact that you possible could whip up half of their product (at least the software) on your own if you desired and had access to the hardware firmware and such.

To their customers, the interesting part is the entire solution. Hardware that speaks for itself, and software that is intelligently made enough not to be completely esoteric to where the customer would have to be as intelligent as the programmer himself in order to make use of it. But at the same time still be a bespoke solution that doesn't leave much for wanting by said consumers. I doubt the companies using AP devices lack the sort of people as AP has employed. They're not looking to buy a product that their R&D department needs to be enchanted with, they're looking for something that streamlines the workflow, so their R&D department wouldn't need to build an in-house AP alternative.

Having the software, or hardware in isolation defeats the solution offered.. Both need to be the entire package, otherwise we would contract you to write the software up for the hypothetical ADC IVX is making at a fraction of the cost of the AP (which would still be fine considering the cost). Or if we had the software, it wouldn't be too useful since you're back at the position of not having the hardware performance/capability only possible from a hardware designer who can come up with the entire package and modularity of the AP while maintaining such performance and usability in tandem.

Though do a collaboration with IVX if hardware is only the problem (or perhaps RME?) you have. Maybe whatever solution you envision for yourself would be cleared with being provided the higher performance you want - could also be something many others may find profound value in?
 
Tks, APx555 uses AK5574 as I remember i.e. lower grade ADC, however, they use the ADC in collaboration with an analog notch filter and variable frequency. APx555 has a bit lower than -120db(some voltage range -117db, some -122db) THD+N resolution. Very cheap analog notch filter at fixed 1kHz frequency lets me to see THD+N -126db(0.00005%, actually, I didn't see less than 0.00006% so far but this is my sine-source issue) from 1 to 10Vrms.
PS: see attached APx555B residual THD+N(red one curve, blue curve SYS2***)

AP555B tear down on AK5594A ADC: Archimago keep in mind the TRACKING NOTCH since SYS-1 :eek: this is not a DIY project ;)

Open about :

. the used clock module and clock distribution(s)
. compare the Jitter/PN using two AP555B, one as DAC and second as ADC and larger FFT than AP supports :p

Hp
 
HpW, correct. AP SYS2*** used almost exactly the same var-notch as described in AP SYS1 service manual. Just better opamps, still single-ended, for sure differential implementation, could improve that for 3-5db, what perhaps happened with 555 ;)
AK5394A not AK5594A yet, it is really old ADC, as I remember, I had some EMU audio-card on that chip in 2005 for class D measurement.
 
I wonder, to those high end ADCs or DACs have significant variation from one chip to the next regarding either noise or HD? Could be possible, especially if they use multibit sigma delta. I wonder if AP preselect their ADCs before they solder them to a beautiful board like this.
 
I wonder, to those high end ADCs or DACs have significant variation from one chip to the next regarding either noise or HD? Could be possible, especially if they use multibit sigma delta. I wonder if AP preselect their ADCs before they solder them to a beautiful board like this.
The chip datasheets usually (with the exception of ESS, of course) specify worst-case and "typical" values, so some variation is presumably expected.
 
Well, the AK5394A data sheet specifies only typical S/(N+D) at 110 dB with a special input circuit and 1000 µF buffer cap on the reference voltage. For the simpler circuit, it specifies 94 dB typ and 87 (!) dB min. From these data, you would not know if this chip is noise or HD limited, and you would never guess that it way outperformed its specs and was (maybe still is) king of the HD hill for nearly 20 years.

Similarly, the product brief of the new kid on the block, ES9822, tells us only - 118 dB THD+N (typ? min?), and Ivan's results are way better. Ivan also said that he sees sample to sample variation in his USB HPAs. Is this the ES9038 that varies or resistor tolerances of the analog circuits after the DAC.
 
HpW, correct. AP SYS2*** used almost exactly the same var-notch as described in AP SYS1 service manual. Just better opamps, still single-ended, for sure differential implementation, could improve that for 3-5db, what perhaps happened with 555 ;)
AK5394A not AK5594A yet, it is really old ADC, as I remember, I had some EMU audio-card on that chip in 2005 for class D measurement.

AK5394A is the best measuring audio ADC actually, aside from this new ESS part. It has better distortion performance than the newer AKM parts, PCM422x, and CS5381.
 
I suppose not, as long as your notch works in all cases. AP also has a AD7760.

I do think they could do a bit better if there were more competition. I have a board I designed at work with 2x LTC2387-18 that would probably do very well in the audio band with a decimation filter.
 
AK5394A is the best measuring audio ADC actually, aside from this new ESS part. It has better distortion performance than the newer AKM parts, PCM422x, and CS5381.

As Jens @DIYAudio compared AK5394A & AK5397 performance wise, AK5394A has a temperature performance issue.

So we see some nice fan's, not a good idea to shake crystal oscillators :D Also on the tear down, no master clock & distribution (TXCO or OXCO ???) seen so far.

While this is a very critical part, so I am very interesting to get any PN performance test using two AP555B (one as DAC one as ADC).
 
As Jens @DIYAudio compared AK5394A & AK5397 performance wise, AK5394A has a temperature performance issue.

So we see some nice fan's, not a good idea to shake crystal oscillators :D Also on the tear down, no master clock & distribution (TXCO or OXCO ???) seen so far.

While this is a very critical part, so I am very interesting to get any PN performance test using two AP555B (one as DAC one as ADC).


Like poorer performance when warm? How come we never see a heat sink on those?
 
AK5394A is the best measuring audio ADC actually, aside from this new ESS part. It has better distortion performance than the newer AKM parts, PCM422x, and CS5381.
The AK5394A really was a milestone with surprising longevity and has
been deployed not only by AP in top-notch audio measurement equipment.
They must have had their reasons to continue using it in their latest
flagship (and hopefully stocked enough with their last-buy-order)!

But it is limited to 216 kHz SR while the ES98xx is capable of at least 4 x more.
In 2-channel mode (hardwired in the ES9822, configurable in the
ES9842), the ESS also achieves slightly higher DR and boasts
considerably lower THD near fullscale.

BUT, getting acquainted better with the part, I noticed (based on some
measurements suggested by another forum member present in this thread)
that the harmonic distortion in the middle range (-20 dBFS to -40 dBFS)
is higher than with the AK5394A and AKM´s newer ADCs, like the AK557x.

With an AKM ADC, from -15 dBFS downwards, the overtones essentially vanish
in the noise floor for middle-sized FFTs, whereas the ESS keeps a grid
of harmonics (albeit at very low level) with level-dependant composition.

This seems to be something typical also for their DACs - extremely high DR and
low THD near fullscale, but still visible distortion in the mid level range.
This may leave the ES98x2 less attractive for measurement purposes,
unless used in conjunction with a tracking notch-filter - something
certainly not trivial to design with the required high performance.
 
So an ESS hump also for the ADC? Seems the trick for the DAC hump was to have exactly the right resistance on the output filter opamp. Not sure how to do this for an ADC.
 
I always suspected that the "ESS hump" in reality has to do with this property of non-totally-vanishing non-linearity at mid levels. I have never used a resistor is series with the feedback-C in the I/U stages - and continue to consider this a bad idea because this way, you´re flattening out the 1st order low pass filter response at higher frequencies, which is not desirable.

Reading about the "ESS hump" and how to presumably remedy it, I was of course curious and tried various resistor values in the 0 to 50 Ohm range in series with the capacitor - without noticing big differences. The composition of the harmonics changes, but overall, they stay at a comparable (very low, but visible) level.
 
Reading about the "ESS hump" and how to presumably remedy it, I was of course curious and tried various resistor values in the 0 to 50 Ohm range in series with the capacitor - without noticing big differences. The composition of the harmonics changes, but overall, they stay at a comparable (very low, but visible) level.

The resistor in series with the cap could be something that made it into their circuit since AD797 was recommended originally. I believe the AD797 datasheet mentions it’s required with a larger cap.
 
Lightrock, so what kind of harmonics have you achieved in with the ES9822 in the -20 to -40 dB range?

What kind of source have you used to measure this? The AK5394A should be capable of -135 dB harmonics at -15 dB full scale which is beyond most generators.

I wonder what the source of midlevel nonlinearity could be. If it's a true 1 bit design, this is hard to imagine. Now if they internally used on of their DAC designs and a comparator, then maybe yes, and there it might be a weird efect between the DAC and the output amp.

I wonder if this is a problem in measurements. To measure SINAD or HD, you would use it near full scale. I wouldn't trust IMD measurements, though.
 
The resistor in series with the cap could be something that made it into their circuit since AD797 was recommended originally. I believe the AD797 datasheet mentions it’s required with a larger cap.
Very likely, indeed! The AD797 can be a bit picky with purely capacitive feedback. I wouldn´t use it as I/U stage for a current output DAC - far too expensive and just one opamp per package. I have used OPA1602, OPA1612, OPA1678 and OPA1692 with the ESS DACs - all yielding satisfactory results in terms of distortion.

The AD797, on the other hand, is extremely well suited as ultra low noise / low distortion preamp, for example as first stage with variable gain in an instrumentation amplifier - not by accident, they are also located in the AFE of the AP analyzers.
 
Lightrock, -20-30db with AK5394A will hide -125db harmonics in the noise floor, isn't it?

I believe it will with my RTX6001. It's been a long time since I checked where the noise floor was at, though.

Obviously, with some setups and techniques (ex. coherent sampling + averaging) you can see below the noise floor.
 
Last edited:
Back
Top Bottom