Power source is disconnected, tried every configuration, still has a high noise floor.
Got no problems with the MOTU Ultralite MK5 ADC, so I suspected the driver, or some settings that can't be configured inside REW or mac.
Here's ADCiso connected to a macbook air m1 (battery mode), the source is SMSL DL200 from another battery mode windows laptop. (no dithering)
Here's on an unpowered windows laptop with FlexASIO. (DL200 + ADCiso)
Problem is that my windows laptop's battery is not that good.
Last edited:
Pardon me for asking the really dumb questions...
Are the settings the same for both REW setups? Same versions of REW?
Is the Input volume control slider set properly so that both set-ups are in mono mode? (Or stereo - your choice.)
Is the Mac internal microphone turned off?
I haven't run the Cosmos ADCiso for a while on macOS REW, but every time I did I got very similar results to when I ran it through Windows 10. That was true across a couple computers, too. Normally I use an Intel iMac for the lab bench, running Windows 10 through Boot Camp. (That's the only Windows computer in the house.) With that set-up, the hardware is exactly the same for either macOS Sonoma or Windows 10. Both were really, really similar in performance.
That certainly is a large discrepancy that you're seeing.
It's also puzzling that the HD2 is noticeably better on the Mac than on the Windows computer. Of course, most of the other harmonics are close to being buried in the noise on the Mac, so there's no way to tell about them.
BTW, that icon of a camera in the upper left of REW allows you to capture the screen and settings you might choose without using a screen capture utility.
Are the settings the same for both REW setups? Same versions of REW?
Is the Input volume control slider set properly so that both set-ups are in mono mode? (Or stereo - your choice.)
Is the Mac internal microphone turned off?
I haven't run the Cosmos ADCiso for a while on macOS REW, but every time I did I got very similar results to when I ran it through Windows 10. That was true across a couple computers, too. Normally I use an Intel iMac for the lab bench, running Windows 10 through Boot Camp. (That's the only Windows computer in the house.) With that set-up, the hardware is exactly the same for either macOS Sonoma or Windows 10. Both were really, really similar in performance.
That certainly is a large discrepancy that you're seeing.
It's also puzzling that the HD2 is noticeably better on the Mac than on the Windows computer. Of course, most of the other harmonics are close to being buried in the noise on the Mac, so there's no way to tell about them.
BTW, that icon of a camera in the upper left of REW allows you to capture the screen and settings you might choose without using a screen capture utility.
Mac REW Version: 5.30.9, Windows REW: 5.30 Beta 1
Yes, both was set to mono mode via the volume slider (yes it works on mac as well, Cosmos ADCiso is the "record interface")
DAC and ADC harmonics sometimes behave differently on selected Sampling Rate, though both are running the same 88.2kHz. Also maybe because of the slightly different REW versions.
I'm just lazy to use the camera icon and configure the capture image graph settings. Using snipping tool / screenshot is easier on my workflow.
No "add dither" on the checkbox of Generator interface. I also used another laptop as the generator just to make sure that mac isn't changing any generator settings.
This is very weird behavior since I'm getting no problems when I use the Motu Ultralite MK5 ADC on my macbook.
Here's Motu Ultralite MK5 ADC @96kHz, using the exact generator from another laptop (no changes, just unplugging ADCiso and plugging the MOTU UL MK5 ADC.)
Motu Ultralite MK5 ADC @88.2kHz, notice the difference in harmonics that I mentioned above. (although H2 to H5 are almost the same)
Last edited:
Did you guys see the review? https://www.audiosciencereview.com/forum/index.php?threads/e1da-cosmos-adciso-review.54020/
I would love to see a version adapted for use in a music studio, to have an alternative to Motu, Lynx and RME.
I would love to see a version adapted for use in a music studio, to have an alternative to Motu, Lynx and RME.
Last edited:
Yes, one need to make sure the settings in Audio Midi Setup for the device is what you want.... easy to miss...
//
I don't have an SMSL DAC. Nor do I have a Windows laptop. But, I do have a MacBook Air M1.
This is a plot I just made of a decade old Victor Oscillator into a Cosmos Scaler into the Cosmos ADCiso. Computer is a MacBook Air M1. All powered by batteries. I think the settings for the spectral plot are similar to SylphAudio's. With one exception - I maxed out the Input Device Buffer Size to 512k.
Note: No notch was used here, so the distortion measured of the oscillator is somewhat limited by the ADC. Plus, this is a 10 year old oscillator. I know Victor's newer units are lower distortion.
This is a plot I just made of a decade old Victor Oscillator into a Cosmos Scaler into the Cosmos ADCiso. Computer is a MacBook Air M1. All powered by batteries. I think the settings for the spectral plot are similar to SylphAudio's. With one exception - I maxed out the Input Device Buffer Size to 512k.
Note: No notch was used here, so the distortion measured of the oscillator is somewhat limited by the ADC. Plus, this is a 10 year old oscillator. I know Victor's newer units are lower distortion.
I just noticed something else.
It looks like you are sampling the right channel of your ADCiso, at least with the Mac. This is what the instructions say about that:
"MONO/STEREO modes are switchable by the UAC2 volume commands, for example, if your Windows volume slider position is less than 50%(0 to 49%) the device works as 2 channel ADC, if more than 50% it is a MONO ADC with summed Left+Right inputs mapped on the Left channel(Right input is always mapped to the Right channel). In the MONO mode, both channels need to be tied together L+ to R+, L- to R-."
In my case, I usually use the Cosmos Scaler because it automatically does all that combining. It also gives a good way to add gain to the input signal, if needed, as well as presenting a much higher impedance to the device being tested. With power amps, that may not matter. DACs and preamps - maybe. Naturally, it adds a little noise and maybe some distortion, but life is full of compromises.
It looks like you are sampling the right channel of your ADCiso, at least with the Mac. This is what the instructions say about that:
"MONO/STEREO modes are switchable by the UAC2 volume commands, for example, if your Windows volume slider position is less than 50%(0 to 49%) the device works as 2 channel ADC, if more than 50% it is a MONO ADC with summed Left+Right inputs mapped on the Left channel(Right input is always mapped to the Right channel). In the MONO mode, both channels need to be tied together L+ to R+, L- to R-."
In my case, I usually use the Cosmos Scaler because it automatically does all that combining. It also gives a good way to add gain to the input signal, if needed, as well as presenting a much higher impedance to the device being tested. With power amps, that may not matter. DACs and preamps - maybe. Naturally, it adds a little noise and maybe some distortion, but life is full of compromises.
Well, as you say, the right output channel is always the right input channel regardless of mode, and has no bearing in noise floor.
To check the noise floor or the ADC, it's neither needed nor favorable to actually measure a device.
The Cosmos work well with inputs open due to it's low impedance design, and in mono summing mode you can get an FFT noise floor below -170dBFS and a 22Hz..22kHz RMS noise floor of -124dBFS, peak mode! Equivalent to ~30nV/rtHz noise spectral density.
That (4M FFT @ 441.1kHz) will take soooome tiiiime though ;-) but it means we can measure harmonics at -120dBc and below easily even when the input signal is 40...50dB down from fullscale. This is quite spectacular.
No, averaging will make the curve smoother. but it will not lower the noise floor.
To reduce the noise floor you need to increase the FFT window size, as done by KSTR above with a 4M-point FFT. Doubling the number of points will reduce the noise floor by 3dB.
Unfortunately the RMS noise stays the same of course.
To reduce the noise floor you need to increase the FFT window size, as done by KSTR above with a 4M-point FFT. Doubling the number of points will reduce the noise floor by 3dB.
Unfortunately the RMS noise stays the same of course.
Okay I will burn my data acquisition books than and ask my money back from my university.....
I am sorry, but noise averaging has been a default method to reduce noise for many many decades.
It's in fact a fundamental principal.
That has nothing to do with smoothing, but actually averaging the noise out.
For this reason this also works in any kind of analog circuit as well.
Go back to signal fundamentals, first year of every science or engineering student.
I am not even gonna explain this since you can find thousands of great articles online about this within 10 seconds.
Although I would recommend just buying and reading a good book about it.
I am sorry, but noise averaging has been a default method to reduce noise for many many decades.
It's in fact a fundamental principal.
That has nothing to do with smoothing, but actually averaging the noise out.
For this reason this also works in any kind of analog circuit as well.
Go back to signal fundamentals, first year of every science or engineering student.
I am not even gonna explain this since you can find thousands of great articles online about this within 10 seconds.
Although I would recommend just buying and reading a good book about it.
Averaging is a "boxcar" LP filter. LP filtering the noise has the effect of smoothing the noise floor so you get more of an averaged value. However averaging is not the same as reducing the noise per bin (then averaging/smoothing or not).
https://en.wikipedia.org/wiki/Boxcar_averager
Last edited:
Have a look at the last sentence in page 7 before figure 6:
https://www.analog.com/media/en/training-seminars/tutorials/MT-001.pdf
https://www.analog.com/media/en/training-seminars/tutorials/MT-001.pdf