There is another layer of abstraction, as you mentioned. Further, other than Intel controllers, many have driver issues. Go look at Synopsys / DWC in many ARM SoCs.
Anyway the rest of this is Linux specific and most people don’t use Linux. My EMU Windows drivers from 2008 written for Vista still work fine in Windows 10 because of the stable ABI. I have written Linux drivers and I am fully aware of the advantages IF your driver is part of the tree.
I don’t need a USB device taking PCIe lanes in my system for a crappy ASMedia controller. Agree to disagree 🙂.
Anyway the rest of this is Linux specific and most people don’t use Linux. My EMU Windows drivers from 2008 written for Vista still work fine in Windows 10 because of the stable ABI. I have written Linux drivers and I am fully aware of the advantages IF your driver is part of the tree.
I don’t need a USB device taking PCIe lanes in my system for a crappy ASMedia controller. Agree to disagree 🙂.
The linux description is for showing how the hardware works. It is hard to learn about hardware functionality on windows without any low-level documentation and no access to source code.
The point is that every USB controller uses DMA on any OS. Just like the USB subsystem is almost same on any OS - the hardware dictates the needs.
For measurements being the topic of this thread (i.e. no low-latency urge) I personally would have no problem with a PCI-e standard USB-audio class 2 device.
But everyone uses whatever fits him best, as it should be 🙂
The point is that every USB controller uses DMA on any OS. Just like the USB subsystem is almost same on any OS - the hardware dictates the needs.
For measurements being the topic of this thread (i.e. no low-latency urge) I personally would have no problem with a PCI-e standard USB-audio class 2 device.
But everyone uses whatever fits him best, as it should be 🙂
The main disadvantage of AK5572 (and many other ADC) - is that it is unusable for the measurement purposes at 192kHz, also 96kHz is not so good because 5572 has the huge noise floor rising after ~40-50kHz.
P.S. Unfortunately, the market of the Audio ADC chips, suitable for measurements, is collapsed 🙁
Arda AT1201, AK5394 are obsolete and are Not Recommended for New Designs, AK557x - see above, AK5397 and PCM4220 - pieces of *****, very bad THD performance.
CS5381, PCM4802/1804 - not enough performance.
To go over 192kHz - just with the industrial ADC.
Last edited:
That's the reason I don't use DSM-ADC. Now you have many excellent SAR-ADCs which have better S/N than DSM-ADCs. The disadvantages are a little bit less THD and LVDS interface. You can digitally compensate THD. LVDS forces you to have FPGA on a PCB, which ends up less popularity.
Have you tested LTC2380-24? The THD looks pretty good according to the datasheet anyway. This may be worth a look as well: http://www.ti.com/lit/ds/symlink/ads127l01.pdf
But, many DSM-ADC does not have this feature (or have it at much higher frequencies, which are out of interested band).
For me THD is a major problem then SNR, and I do not think it can be really compensated. The better way is to use notch filter, but this solution have other problems.
LTC, ADS, this it what I said before - we have to go away from audio ADC to industrial ones.
Last edited:
Thank you for your information. My experience is AD9760. The 1st pic is FFT of EV-board by AD. The 2nd is my designed one. The input is 1kHz oscillator by Frex. Both are pretty good, but some harmonics are above the noise floor.
I thought SAR-ADC has capacitor ladder for its architecture, where physical voltage divider results in some distortion like R2R DAC, though the datasheet has no such harmonics.
The 3rd is with the input short. This is almost the same value as the datasheet, very excellent. ADS127L01 has nearly the same value as AD7960 at 20kHz bandwidth from table1 on page 25. But it is with a 2.5V reference, where a 6dB loss in S/N.
The 4th is a 3dB improvement by a mono operation. SAR-ADC has a less individual difference. You can easily improve their performance by paralleling.
Attachments
I also thought compensation of THD wasn't possible. But I'm sure now it's possible both ADC and DAC side separately. Phofman is a specialist whose success is just around the corner. I'm also planning to upgrade my ADC-board to have compensation ability. I need a little hardware mod to assist the compensation.🙂
But, what you gain in this frequency band with $50 chip ($100 for 2ch) in terms of THD?
The same results can be obtained with $6 PCM1804 ($6 for 2ch):
Attachments
While I do appreciate your trust I would suggest to wait for confirmation by other users too. It is still in quite early stages. Plus the computational demands are not negligible. All calculations in floating point, double precision (in my case in octave).
My top priority is S/N. Pcm1804 isn't excellent in S/N from your Pic. Total PCB cost including medium-size FPGA, which must have at least 512k buffer for SD write interface, is about $350. It's not overkilling for me to pay 2x$50 for AD-chips. I would say the audio band is a little bit narrow for AD7960 though.
Yes, sure.
But for audio measurements I don't know how big SNR can help if THD is not good.
why build a usb device that can be purchased eg M-Audio inexpensively?
Your older Iphone or Samsung Android ( Linux ) phone will outperform it.
hits all the specs 115 db s/n 24/96 dac and battery power
Your older Iphone or Samsung Android ( Linux ) phone will outperform it.
hits all the specs 115 db s/n 24/96 dac and battery power
ASIO Arrives – QuantAsylum
Note that the Clarett 2Pre *USB* has poor latency (..it's a USB 2 device (despite having a USB-C mini connector) with less than excellent software/drivers). I'm not sure that matters in this instance, but it's something to be aware of. It's "spec.s" are otherwise excellent, and you can still get the Thunderbolt version direct from Focusrite (that doesn't have the latency issues), but you will need a Thunderbolt port.
The SoundBlaster G6 has poor IMD, and should be avoided because of this in this instance.
The Presonus Studio 26c might be a good compromise overall.. it doesn't have the best spec.s, but they aren't "shabby" either: Studio 26c | Tech Specs | PreSonus
Note that the Clarett 2Pre *USB* has poor latency (..it's a USB 2 device (despite having a USB-C mini connector) with less than excellent software/drivers). I'm not sure that matters in this instance, but it's something to be aware of. It's "spec.s" are otherwise excellent, and you can still get the Thunderbolt version direct from Focusrite (that doesn't have the latency issues), but you will need a Thunderbolt port.
The SoundBlaster G6 has poor IMD, and should be avoided because of this in this instance.
The Presonus Studio 26c might be a good compromise overall.. it doesn't have the best spec.s, but they aren't "shabby" either: Studio 26c | Tech Specs | PreSonus
Last edited:
It depend of what you want to measure.
For the loudspeakers frequency response, THD, IMD measurements - smartphone or an old soundcard is possible to use. Or for the SE Tube amplifier.
But it is not enough for the modern amplifiers, pre-amplifiers, DAC's, etc., which have the noise floor (this is not SNR!!!) about -130-140dB and THD/IMD harmonics level below -100-120dB.
I still think you will need a notch filter if you want to measure the absolute state-of-the-art. The top ADCs and top DACs are all very close in performance. It's hard to have any certainty at those levels.
Sure, I've mentioned the notch filter above, but also I want to have ADC's THD as low as possible.
The best audio ADC is still the AK5394A. If carefully implemented the harmonics are around -130 dB and the noise floor is almost flat across the band. You could parallel several and get even lower noise and possibly less distortion. unfortunately its not recommended for new designs and its expensive.
There are industrial delta sigma ADC's that have high performance e.g. http://www.ti.com/product/ADS1672 . SAR is not the only option. Or, if you are really nuts this could be used: 3458A Digital Multimeter, 8 1/2 Digit | Keysight (formerly Agilent’s Electronic Measurement) It samples at up to 100K/second and is the most linear ADC known.
There are industrial delta sigma ADC's that have high performance e.g. http://www.ti.com/product/ADS1672 . SAR is not the only option. Or, if you are really nuts this could be used: 3458A Digital Multimeter, 8 1/2 Digit | Keysight (formerly Agilent’s Electronic Measurement) It samples at up to 100K/second and is the most linear ADC known.
Do you know if AKM plans to make a replacement ADC anytime soon? A 5397 with better distortion, anyway. It's weird that the 557x has better distortion than their "flagship".
Anyone know what happened to the ESS ADCs? They seem to have been quietly killed.
Anyone know what happened to the ESS ADCs? They seem to have been quietly killed.
along with the RTX 6001, I believe the SpectraDAQ-200 has this: SpectraPLUS Hardware Recommendations - it also has AISO drivers AND IEPE-current drive for Piezo Mic.s..
I think the QA401 uses the "newer" AK5397.