Too high LF noise for the intended purpose (DACs & ADCs). And the most unreliable IC I have ever worked with. I don't mind the cost but the couple of hours I spent on PCB layout to accommodate these.
Yes. The first 2 I tried failed within 1 hour without any substantial load. No electric shocks or similar happened. Of the dual regulators first one failed and soon after the other. Syn08 has reported similar failures at ASR. After these 2 failures I decided to call it quits with ES9311Q as the noise performance was bad as well. Though much more expensive LT3042 works much better and is very reliable.
LT3042, LT3045, LT3094 and TPS7A types are quite OK (if not the best what can be found today). The LT types dislike heat as was mentioned here. Turned out to be a drawback but it can be solved in a few ways.
LT3032 is on my list of regulators to try out. On paper it seems a good choice for output stages.
LT3032 is on my list of regulators to try out. On paper it seems a good choice for output stages.
Last edited:
The ESS hump is subject of tremendous discussion over at ASR. However, no one ever discusses on that forum the importance of the output stage and the cleanliness (or lack of) of the various voltage supplies necessary to power these D-S devices.
I have an R2R (TDA1541) dac with a hybrid SS and tube buffer OPS. I built that dac from scratch, focusing on the power supplies. I also have a commercial ES9018 which also employs a hybrid OPS (SS (op amp) to grounded grid tube). Both of these dacs sound wonderful and smooth and the TDA1541 very slightly edges it the Sabre dac in overall smoothness and presentation in general. Neither one of these dacs employ inexpensive monolithic voltage devices. Most inexpensive dacs employ textbook / cookbook output stages and monolithic Vregs., which I think are totally inadequate for the AVCC supplies.
The point is, there is a raging discussion about the 'ESS hump" with zero discussion about all of the various supporting circuitry.
I have an R2R (TDA1541) dac with a hybrid SS and tube buffer OPS. I built that dac from scratch, focusing on the power supplies. I also have a commercial ES9018 which also employs a hybrid OPS (SS (op amp) to grounded grid tube). Both of these dacs sound wonderful and smooth and the TDA1541 very slightly edges it the Sabre dac in overall smoothness and presentation in general. Neither one of these dacs employ inexpensive monolithic voltage devices. Most inexpensive dacs employ textbook / cookbook output stages and monolithic Vregs., which I think are totally inadequate for the AVCC supplies.
The point is, there is a raging discussion about the 'ESS hump" with zero discussion about all of the various supporting circuitry.
Well, that's the title of thread. Just saying.... ESS Hump has become a buzz phrase all over the place with little else discussed.
Actually the discussion at ASR has lately been around similar IMD hump in new AK4499EX dacs.
Regarding ESS IMD hump the output stage and voltage supplies do have some impact but as has been discussed in this thread the issue is intrinsic to ESS DAC/ADC architecture. Anyhow the lower the noisefloor the more likely the IMD hump is visible. With tube output stage any signs of the ESS IMD hump will be buried deep below the noisefloor and other IMD artefacts.
Regarding ESS IMD hump the output stage and voltage supplies do have some impact but as has been discussed in this thread the issue is intrinsic to ESS DAC/ADC architecture. Anyhow the lower the noisefloor the more likely the IMD hump is visible. With tube output stage any signs of the ESS IMD hump will be buried deep below the noisefloor and other IMD artefacts.
To support bohrok2610, here are IMD CCIF test of SMSL DO300EX, XLR output with COSMOS ADC, based on ES9822PRO,
and with Nihtila ADC, based on AK5572. Both done with the same settings, only ADC were changed. Of course the two ADCs do not have the same full scale sensitivity, but the x-axes is the generator level and not input level. It is clear the ES9822Pro suffers from IMD hump (and not the AK4499 here).
Soon I will get the cosmos scaler (auto range) that will help to avoid the hump area (or to shift it somewhere else...).
Attachments
Ok, looks compelling. I am amazed that the AK5572 can hold water to the ES9822, espcially in Ivan's implementation. I seem to remember that the AK5572, while it boasted better SNR, was not considered a worthy successor to the venerable AK5394A in the harmonic distortion department, and the ES9822 seems to beat even the AK5394A in harmonic distortion. Of course, we are talking IMD here, so maybe a different story.
Maybe Ivan wants to chime in? He said he changed the front end from Cosmos ADC to ADCiso to improve 10 kHz HD, so maybe this would have an impact on IMD, too?
Maybe Ivan wants to chime in? He said he changed the front end from Cosmos ADC to ADCiso to improve 10 kHz HD, so maybe this would have an impact on IMD, too?
If I will have time, I will send 1khz level seep comparison.
The main problem with AK5572 is its aggressive noise shaping that also starts early.
For harmonics to 1khz it is good enough, but to measure harmonics of 10khz or even 20khz it is useless.
Also, Ivan's cosmos ADC (I have non iso version) has a very low noise floor. For that I am not sure if it only the ADC itself that is better or the low resistance front end that Ivan implemented.
The main problem with AK5572 is its aggressive noise shaping that also starts early.
For harmonics to 1khz it is good enough, but to measure harmonics of 10khz or even 20khz it is useless.
Also, Ivan's cosmos ADC (I have non iso version) has a very low noise floor. For that I am not sure if it only the ADC itself that is better or the low resistance front end that Ivan implemented.
You're right, that was the main gripe with AK5572 as well as most other audio ADCs other than AK5394A an ES9822, that their noise sampling corner starts too early.
Now that graph is quite something! It is the first time I have seen the IMD hump play out in 1 kHz THD for either ADCs or DACs. It is also much longer (down to -50 dB) and wider (16 dB relative to AK5572) than the IMD humps that are usually found in Amir's testing. Amir's HD and SINAD is usually near full scale (at 4.0 V). It wouldn't hurt to for Amir to add a -20 dB test or even better a sweep of 1 kHz THD.
What kind of software are you using to do those sweeps and combine them into a graph?
As for comparing the two ADCs, this mirrors what's been said about AKM and ESS DACs: AKM has excellent THD up to near full scale and falls apart in the last few dB whereas ESS excels near fs.
More subtly, the AK5572 seems to get even better than the extrapolation of the straight line between -40 and -20 dB (inverse hump). Maybe this is an artefact of the fitting algorithm, i.e. less noise getting binned with the harmonics at higher level? It is also interesting to note that the ES9822 falls about 10 dB short of the extrapolated line near fs.
Lastly, with the hump even for harmonics being most pronounce around -35 dB, that kind of discredits the idea that the hump is about glitching in the switches getting worst near fs.
Now that graph is quite something! It is the first time I have seen the IMD hump play out in 1 kHz THD for either ADCs or DACs. It is also much longer (down to -50 dB) and wider (16 dB relative to AK5572) than the IMD humps that are usually found in Amir's testing. Amir's HD and SINAD is usually near full scale (at 4.0 V). It wouldn't hurt to for Amir to add a -20 dB test or even better a sweep of 1 kHz THD.
What kind of software are you using to do those sweeps and combine them into a graph?
As for comparing the two ADCs, this mirrors what's been said about AKM and ESS DACs: AKM has excellent THD up to near full scale and falls apart in the last few dB whereas ESS excels near fs.
More subtly, the AK5572 seems to get even better than the extrapolation of the straight line between -40 and -20 dB (inverse hump). Maybe this is an artefact of the fitting algorithm, i.e. less noise getting binned with the harmonics at higher level? It is also interesting to note that the ES9822 falls about 10 dB short of the extrapolated line near fs.
Lastly, with the hump even for harmonics being most pronounce around -35 dB, that kind of discredits the idea that the hump is about glitching in the switches getting worst near fs.
Hi,
First thing: as I understand AP uses auto ranger at least for one tone (1khz). By this you bypass the interval of hump and maybe that the reason with Amir you can see IMD hump but not THD hump.
The software is REW.
What you said about near full scale degradation of the AKM is true but the sensitivity I have used for the cosmos is 6.7V scale while for the Nihtila (AKM) about 4V. The SMSL full scale is 5V. So, the cosmos is not as close to it's full scale as the Nihtila.
About the extrapolation, I will send the shots of AKM and ESS ADCs separately and with respect to their input level to ADC.
First thing: as I understand AP uses auto ranger at least for one tone (1khz). By this you bypass the interval of hump and maybe that the reason with Amir you can see IMD hump but not THD hump.
The software is REW.
What you said about near full scale degradation of the AKM is true but the sensitivity I have used for the cosmos is 6.7V scale while for the Nihtila (AKM) about 4V. The SMSL full scale is 5V. So, the cosmos is not as close to it's full scale as the Nihtila.
About the extrapolation, I will send the shots of AKM and ESS ADCs separately and with respect to their input level to ADC.
I don't think I've ever seen a THD sweep in Amirs DAC measurements.
Just a thought, inspired by Ivans post in the other hump thread that got hijacked by the subjectivists: https://www.diyaudio.com/community/threads/return-of-the-ess-hump.391267/post-7205775
What if your DAC produces copious amounts of this hump and the AKM ADC is blind to this because of its aggressive noise shaping and the ES ADC detects it, maybe as aliasing? Using a switched attenuator would be one way, especially if it is extremely low distortoion as the scaler. The other idea would be to use a brick wall low pass or an inverted notch on the DAC output. Or a regular notch on the ADC which would filter out the fundamental. If the hump remains, it is not due to IMD in the ADC.
Just a thought, inspired by Ivans post in the other hump thread that got hijacked by the subjectivists: https://www.diyaudio.com/community/threads/return-of-the-ess-hump.391267/post-7205775
What if your DAC produces copious amounts of this hump and the AKM ADC is blind to this because of its aggressive noise shaping and the ES ADC detects it, maybe as aliasing? Using a switched attenuator would be one way, especially if it is extremely low distortoion as the scaler. The other idea would be to use a brick wall low pass or an inverted notch on the DAC output. Or a regular notch on the ADC which would filter out the fundamental. If the hump remains, it is not due to IMD in the ADC.
