9822 needs 4.5V, 9039q2m 3.3V. But the 3.3V can be derived from the 4.5V with a buffered/filtered resistor divider so that they have the same long-term drift and almost the same 1/f-noise. This will reduce loop-back delta (DAC vs ADC) gain drifts and thus will null better in compares.
A common clock is, of course, one most important thing, so that one can use synced averaging and can do compares with DeltaWave straight away.
All -- I'm looking for some collective wisdom here to help with a problem I'm having with a Cosmo E1DA ADC APU.
I've got the device working just fine and I get spectra using various signal sources just as expected -- I'm using REW.
But about half the time I suddenly get a spectrum that appears to show a strong 60Hz signal modulating my test signal. This goes away after a while and a normal spectrum returns, but it makes the unit hard to use.
I've attached an example of what this looks like -- the input signal is a simple 1kHz sine wave at 0.5V.
I'm using a separate USB power supply for the unit, but changing the power supply does not cure the problem. Similarly, changing the signal source gives the same results.
Unplugging the signal source from the unit makes the problem disappear, but it returns as soon as I plug it back in.
The strong interfering 60Hz signal might be consistent with a signal ground connection becoming intermittent, but changing the various analog cables also has no effect.
Any suggestions would be greatly appreciated!
I've got the device working just fine and I get spectra using various signal sources just as expected -- I'm using REW.
But about half the time I suddenly get a spectrum that appears to show a strong 60Hz signal modulating my test signal. This goes away after a while and a normal spectrum returns, but it makes the unit hard to use.
I've attached an example of what this looks like -- the input signal is a simple 1kHz sine wave at 0.5V.
I'm using a separate USB power supply for the unit, but changing the power supply does not cure the problem. Similarly, changing the signal source gives the same results.
Unplugging the signal source from the unit makes the problem disappear, but it returns as soon as I plug it back in.
The strong interfering 60Hz signal might be consistent with a signal ground connection becoming intermittent, but changing the various analog cables also has no effect.
Any suggestions would be greatly appreciated!
What driver setting are you using in REW?
Suggested REW Driver
A large input buffer can help with these problems. Sometimes.
Are you using an ADCiso or an ADC "Classic"? Isolation is often needed on the USB lines to eliminate these kinds of problems. Of course, you could also use a battery powered laptop for isolation purposes.
These bugs can take a while to find and fix.
Suggested REW Driver
A large input buffer can help with these problems. Sometimes.
Are you using an ADCiso or an ADC "Classic"? Isolation is often needed on the USB lines to eliminate these kinds of problems. Of course, you could also use a battery powered laptop for isolation purposes.
These bugs can take a while to find and fix.
Thanks for the suggestions!
I'm using the ASIO4All v2 driver with REW -- it seems to work for this purpose. I usually use Java drivers with REW but I think I understand that Java won't support 24 bit data.
Other details that might have a bearing (or not!) -- I'm using a batter powered laptop pc. I'm also using an ADUM 3160 based USB voltage isolator with a separate linear power supply to power a Topping D10 DAC, as a signal generator. Other signal generators I've used are battery powered, but they give the same result as I've shown above, so the power source of the signal generator doesn't appear to be a factor.
I've not been able to get the voltage isolator to work with either the signal or power supply USB inputs on the Cosmo E1DA, so they are connected to a stand alone USB power supply and directly to the computer, respectively. I think I read somewhere that the Cosmo is not designed to work with a voltage isolator, although why that would be I'm not sure!
I appreciate that bugs like this can be tricky to resolve!
Thanks again,
Mark
I'm using the ASIO4All v2 driver with REW -- it seems to work for this purpose. I usually use Java drivers with REW but I think I understand that Java won't support 24 bit data.
Other details that might have a bearing (or not!) -- I'm using a batter powered laptop pc. I'm also using an ADUM 3160 based USB voltage isolator with a separate linear power supply to power a Topping D10 DAC, as a signal generator. Other signal generators I've used are battery powered, but they give the same result as I've shown above, so the power source of the signal generator doesn't appear to be a factor.
I've not been able to get the voltage isolator to work with either the signal or power supply USB inputs on the Cosmo E1DA, so they are connected to a stand alone USB power supply and directly to the computer, respectively. I think I read somewhere that the Cosmo is not designed to work with a voltage isolator, although why that would be I'm not sure!
I appreciate that bugs like this can be tricky to resolve!
Thanks again,
Mark
Well...
Pavel Hofman wrote special Java drivers that are included with REW for both Windows and Linux. They allow you to use WASAPI Exclusive input to your heart's desire. I have found this to be the best solution under Windows 10 - that's also what Ivan (Mr. E1DA Cosmos) recommends. I have the buffer size set to 512k. REW certainly believes it is getting 32 bit data. I had loads of troubles like you describe when using ASIO of any kind. Switching to Exclusive under Java remedied all of them.
The power source is certainly a consideration with using a set-up like this. But, breaking the loop in the return path in the USB chain is important for getting rid of USB junk. You still have that loop if you are using a single laptop run on battery for all the devices. An ADUM3160 won't work with either the D10 or the Cosmos ADC unless you run them at the lower sample rates. The ADUM3160 will not operate above "Full Speed" USB rates. You need an isolator based around the ADUM3166, ADUM4166, or their competitor's equivalent parts that operate up to "High Speed" USB.
I know that you can make a D10 > Cosmos ADC set-up work - I've done it myself. Before adding an ADCiso to the bench, I used a classic Cosmos ADC with an ADUM4166 isolator that I built between the Cosmos and the PC. Worked great.
Or, you could just have gremlins in your workshop.
Pavel Hofman wrote special Java drivers that are included with REW for both Windows and Linux. They allow you to use WASAPI Exclusive input to your heart's desire. I have found this to be the best solution under Windows 10 - that's also what Ivan (Mr. E1DA Cosmos) recommends. I have the buffer size set to 512k. REW certainly believes it is getting 32 bit data. I had loads of troubles like you describe when using ASIO of any kind. Switching to Exclusive under Java remedied all of them.
The power source is certainly a consideration with using a set-up like this. But, breaking the loop in the return path in the USB chain is important for getting rid of USB junk. You still have that loop if you are using a single laptop run on battery for all the devices. An ADUM3160 won't work with either the D10 or the Cosmos ADC unless you run them at the lower sample rates. The ADUM3160 will not operate above "Full Speed" USB rates. You need an isolator based around the ADUM3166, ADUM4166, or their competitor's equivalent parts that operate up to "High Speed" USB.
I know that you can make a D10 > Cosmos ADC set-up work - I've done it myself. Before adding an ADCiso to the bench, I used a classic Cosmos ADC with an ADUM4166 isolator that I built between the Cosmos and the PC. Worked great.
Or, you could just have gremlins in your workshop.
Forgot.
There's more info on the REW drivers starting with a post from Pavel Hofman here:
REW Java Drivers
Some more here:
Mo
There's more info on the REW drivers starting with a post from Pavel Hofman here:
REW Java Drivers
Some more here:
Mo
I've been having similar problems with the Ultralite Mk5 using a battery powered laptop and the driver package supplied by Motu and ARTA. One or several spectra will be off, then it's working nicely, then again. I changed the ADC side to ADCiso which made these problems rarer but not go away competely. I haven't had time to go the bottom of it yet. Since I now have a 9039S, I may go that way.
BTW, this is some resume about how to measure audio with Cosmos devices, made by the user. I hope it will be useful for some newbies.
cosmos adc distortion when bypassing input caps (220u/10v elyt). the 20-150Hz band is affected.
measuring chain:
Topping D10s -3dBFS (1.4Vrms) -> RCA to XLR cable -> Cosmos ADC 1.7Vrms left unbalanced 96k
maybe higher input sensitivities work better
measuring chain:
Topping D10s -3dBFS (1.4Vrms) -> RCA to XLR cable -> Cosmos ADC 1.7Vrms left unbalanced 96k
maybe higher input sensitivities work better
It is not unexpected to see DC blocking caps contribute distortion near and below the corner frequency of the high pass so I believe LKA can obtain a 20 dB improvement by bypassing these caps.
How much DC offset was introduced by this and did this maybe result in additional distortion that somehow cancelled existing distortion?
Where does the residual distortion come from? Does the D10S have output caps?
And the big question: why is unmodified distortion measured by Ivan about the same as modified by LKA? Better signal source? Newer ADCiso with changed frontend?
How much DC offset was introduced by this and did this maybe result in additional distortion that somehow cancelled existing distortion?
Where does the residual distortion come from? Does the D10S have output caps?
And the big question: why is unmodified distortion measured by Ivan about the same as modified by LKA? Better signal source? Newer ADCiso with changed frontend?
IVX, how to properly bypass the dip switches and install a bigger input capacitor ? I want only 1.7Vrms input sensitivity or so.
imho those small dip switches are design flaw. when 1.7Vrms range is used there are 6 contacts (3x IN+ and 3x IN-) in the signal path.
Lots of people have studied distortion in electrolytic capacitors. One of the most thorough investigations was by Cyril Bateman about 20 years ago. Yeah, you can measure better now and caps have changed.
Bateman on Capacitors
Look at section 6, especially. He found that bipolar electrolytics have the lowest distortion, everything else equal. He explains why in other sections.
Nichicon sells a couple series of bipolar aluminum electrolytic capacitors. So do Elna and Audio Note.
The problem is that none that I've found are direct drop-ins for the caps used in the Cosmos ADC. The lead spacing is larger and the cans have a larger diameter. They're taller, too, but that doesn't seem to be a problem in the Cosmos ADC case. I suppose that you could space a larger cap above the board to clear the resistors and all that. The lead lengths would be a bit longer, but that may not matter much. I'm not sure about lead diameter.
About 10 years ago some measurements were posted of distortion of Nichicon ES series bipolar right here on diyAudio.
Nichicon Muse ES Bipolar Caps Measured
Then, there's this:
Selecting Electrolytic Capacitors for Lowest Distortion
I guess every ten years or so somebody has something to say on the subject.
I suspect that none of these tests really are exactly comparable to the conditions for the Cosmos ADC. Except for the Self test, the impedances are all much higher, for one thing.
Before changing caps out, I'd want to try bypassing the internal caps and measuring the performance of external blocking caps. And, get Ivan's opinions. I don't love the idea of desoldering and soldering new caps onto the ADC board for fear of damaging the board from too many soldering cycles. (Maybe this would be a good task for me once I get through the current backlog of unfinished projects.)
BTW... The D10s has some bumps when it comes to the distortion performance, as does the E50. The guess is that it has to do with how the digital distortion cancellation is set up. You can change its performance by fiddling with some settings in REW.
E50 and D10s performance in REW
Bateman on Capacitors
Look at section 6, especially. He found that bipolar electrolytics have the lowest distortion, everything else equal. He explains why in other sections.
Nichicon sells a couple series of bipolar aluminum electrolytic capacitors. So do Elna and Audio Note.
The problem is that none that I've found are direct drop-ins for the caps used in the Cosmos ADC. The lead spacing is larger and the cans have a larger diameter. They're taller, too, but that doesn't seem to be a problem in the Cosmos ADC case. I suppose that you could space a larger cap above the board to clear the resistors and all that. The lead lengths would be a bit longer, but that may not matter much. I'm not sure about lead diameter.
About 10 years ago some measurements were posted of distortion of Nichicon ES series bipolar right here on diyAudio.
Nichicon Muse ES Bipolar Caps Measured
Then, there's this:
Selecting Electrolytic Capacitors for Lowest Distortion
I guess every ten years or so somebody has something to say on the subject.
I suspect that none of these tests really are exactly comparable to the conditions for the Cosmos ADC. Except for the Self test, the impedances are all much higher, for one thing.
Before changing caps out, I'd want to try bypassing the internal caps and measuring the performance of external blocking caps. And, get Ivan's opinions. I don't love the idea of desoldering and soldering new caps onto the ADC board for fear of damaging the board from too many soldering cycles. (Maybe this would be a good task for me once I get through the current backlog of unfinished projects.)
BTW... The D10s has some bumps when it comes to the distortion performance, as does the E50. The guess is that it has to do with how the digital distortion cancellation is set up. You can change its performance by fiddling with some settings in REW.
E50 and D10s performance in REW
CG,I think I made my point clear, these caps are neutral, at least if they are not damaged. I tried different alcaps with Cosmos ADC proto, and found no difference with the best versus which I finally used for the production. Above I posted the THD+N vs freq plot, as you can see there is <-110db@20Hz. Right now I retested that with Cosmos ADCiso, and again, I see no difference between my caps and shorted caps.
with original 220uF/16V:
and shorted capacitors:
with original 220uF/16V:
and shorted capacitors:
Another THD test. The capacitors are not bypassed !
GREEN trace: inserted 1k resistor (small black box on the picture) between D10s and Cosmos
RED trace: direct connection
The output level of the D10s is chosen so that the input level of the Cosmos is -10dBFS in both cases. ( 0dBFS = 1.7Vrms)
It is clearly visible that additional series resistor helps.
GREEN trace: inserted 1k resistor (small black box on the picture) between D10s and Cosmos
RED trace: direct connection
The output level of the D10s is chosen so that the input level of the Cosmos is -10dBFS in both cases. ( 0dBFS = 1.7Vrms)
It is clearly visible that additional series resistor helps.
Last edited: