Ugh. Feeling quite humiliated/humbled/stupid at this point. Spent 12 hours trying to run measurements using the Cosmos. For the life of me, I can't get REW, Phonalyser, Sound Recorder or even the basic Windows Input sound test (when connected to the Cosmos) to recognize a signal.
Cosmos first powered on via powerbank, then plugged in via USB. Windows recognizes the Cosmos and my processor.
Front R XLR connected to the Cosmos (via Y-XLR cable), have tried both left and right. Left and right on Cosmos. Java. FlexASIO.
If I measure using a VMM or FINSRI, voltage is measured when I send a signal via REW (or just a bluray). Below is a chatgpt summary of what I've tried and results, if helpful.
Honestly at a loss as to what to try next.
Current status: We have confirmed that REW generates a signal that reaches the AT-600's analog XLR output. However, REW is not receiving a usable signal from the Cosmos ADC. We have also observed an unusually large voltage reduction when the Cosmos is connected and powered.
1. Equipment and configuration
| Component | Details |
|---|
| Processor | Tonewinner AT-600 |
| Computer | Windows laptop, HDMI to AT-600 |
| Measurement ADC | E1DA Cosmos ADC |
| ADC sensitivity | DIP switch row 3, nominal 1.66 Vrms at −0.5 dBFS |
| Software | REW V5.40 Beta 139 |
| Audio interface | FlexASIO initially; Java subsequently |
| Oscilloscope | FNIRSI DSO-TC3 |
| Multimeter | Klein CL120 |
| XLR connections | Hosa YXM121 splitter and XLR screw-terminal breakout |
| Test signal | 100 Hz sine, −6 dBFS |
| AT-600 volume | −10 dB |
| Sampling rate | 48 kHz |
| Processor EQ | Off |
| Power amplifiers | Muted |
We previously verified the XLR breakout pin assignments:
| Breakout terminal | XLR pin |
|---|
| GND | Pin 1 |
| L | Pin 2 (hot) |
| R | Pin 3 (cold) |
The FNIRSI oscilloscope was previously calibrated, and the breakout wiring was checked for continuity and shorts.
2. Analog output and Cosmos loading measurements
These tests investigated whether connecting the Cosmos changes the AT-600's output voltage.
At 100 Hz — Klein CL120
| Cosmos connection | Measured voltage | Change vs. disconnected |
|---|
| Cosmos XLR disconnected | 0.288 Vrms | Baseline |
| Cosmos XLR connected, USB power disconnected | 0.222 Vrms | −22.9% |
| Cosmos XLR connected, powered by USB power bank | 0.126 Vrms | −56.3% |
At 1 kHz — Klein CL120
| Cosmos connection | Measured voltage | Change vs. disconnected |
|---|
| Cosmos disconnected | 0.344 Vrms | Baseline |
| Cosmos connected, unpowered | 0.265 Vrms | −23.0% |
| Cosmos connected, powered | 0.142 Vrms | −58.7% |
Important: The Klein CL120 is specified for AC voltage measurements at 45–400 Hz. The 100 Hz measurements are therefore more reliable than the 1 kHz measurements.
What we established
Connecting the Cosmos produces a substantial measured voltage reduction, especially when powered.
The Cosmos's nominal balanced input impedance is approximately 8.8 kΩ. Such a large reduction is unexpected for a conventional low-impedance preamplifier output.
We have not established why this happens. Possible causes include loading, wiring interactions, ADC input configuration, or an output-stage issue. We should not conclude that either device is defective yet.
3. Earlier oscilloscope measurements
Using the FNIRSI DSO-TC3 at 1 kHz, measuring individual balanced XLR legs relative to pin 1:
| AT-600 volume | Cosmos | Approx. voltage |
|---|
| −20 dB | Disconnected | 0.05 Vrms |
| −20 dB | Connected | 0.04 Vrms |
| −10 dB | Connected | 0.15 Vrms |
| −10 dB | Disconnected | 0.17 Vrms |
These measurements also suggested a voltage reduction when the Cosmos was connected, although the reduction was smaller than in the later Klein tests.
Because the oscilloscope measured individual XLR legs while the Klein measured the differential output, these results are not directly interchangeable.
4. Windows and REW troubleshooting
We spent considerable time determining whether the problem was Windows audio routing, REW configuration, or the Cosmos ADC.
| Test or change | Result |
|---|
| REW with FlexASIO | No usable signal in RTA |
| FlexASIO output device changed | Did not resolve problem |
| Cosmos selected as FlexASIO input | No usable signal |
| Windows Sound Recorder using Cosmos | Initially unstable; subsequently remained running but showed no meaningful waveform |
| Windows microphone test | Initially 0% |
| Changed Cosmos Windows format to 24-bit, 48 kHz | Windows microphone test subsequently reported 29% |
| Checked Windows Device Manager | Intel Display Audio and Cosmos ADC both listed |
| Checked Windows Playback devices | AT-600 recognized and set as default |
| Windows audio activity bar | Pulsated, indicating Windows playback activity |
| Changed REW driver from FlexASIO to Java | AT-600 and Cosmos became explicitly selectable |
| Selected EXCL: AT-600 output | REW recognized HDMI device |
| Selected EXCL: Cosmos ADC input | REW recognized ADC |
| REW output set to R | Output meter showed −6 dBFS |
| REW input set to L, then R | Neither showed a usable tone |
The Windows microphone test reaching 29% was encouraging, but it did not establish that REW could capture the correct analog signal.
Current REW settings
| Setting | Value |
|---|
| Driver | Java |
| Output device | EXCL: AT-600 (Intel Display Audio) |
| Output | Master Volume, R |
| Input device | EXCL: Cosmos ADC |
| Input | LINE_IN (Master Volume) |
| Input channel | L or R tested |
| Sample rate | 48 kHz |
| Generator | 100 Hz, −6 dBFS |
5. Definitive REW-to-AT-600 output test
We measured the Front Right balanced XLR output using the Klein CL120 while toggling REW's generator.
| Generator state | Measured XLR voltage |
|---|
| ON — 100 Hz, −6 dBFS | 0.126 Vrms |
| OFF | 0 Vrms |
Conclusion: REW is successfully producing a signal that reaches the AT-600's Front Right XLR output.
This rules out a complete failure of the REW-to-HDMI-to-AT-600 signal path.
It does not establish that the output level is correct or that the signal is distortion-free.
6. Cosmos input channel comparison
With the 100 Hz generator running, we switched REW's input selection between Left and Right.
| Cosmos input channel | REW input meter |
|---|
| Left | −89.52 dBFS |
| Right | −87.68 dBFS |
The RTA also displayed approximately −125 to −237 dBFS in different captures, without a clearly visible 100 Hz tone. These RTA figures are not directly comparable to the input-meter readings.
Conclusion: Neither selected input channel is showing the expected signal. Switching L/R did not fix the problem.
At the Cosmos's nominal sensitivity, a genuine 0.126 Vrms balanced input would ordinarily correspond to approximately −22.9 dBFS, assuming no additional attenuation. Our observed readings are dramatically lower.
7. What is confirmed versus unresolved
| Question | Status | Finding |
|---|
| Does Windows recognize the AT-600? | Confirmed | Yes |
| Does Windows recognize the Cosmos? | Confirmed | Yes |
| Does REW send a signal to the AT-600? | Confirmed | 0.126 Vrms ON, 0 V OFF |
| Are XLR breakout pins identified? | Confirmed | Pins 1, 2, 3 verified |
| Does Cosmos connection affect measured voltage? | Observed | Yes, substantially |
| Is REW capturing the expected signal? | No | Neither input channel shows the tone |
| Is the Cosmos ADC itself defective? | Unknown | Not established |
| Is the AT-600's XLR output defective? | Unknown | Not established |
| Can we calculate reliable SINAD yet? | No | Valid ADC capture required |
8. Recommended next steps
I would stop changing REW's output settings. We have already proved that the generator reaches the processor.
The remaining work should focus on two separate issues:
- ADC capture: Verify the Cosmos's physical input connection, USB data connection, and actual signal capture independently of REW's generator. This should establish whether the ADC can correctly digitize a known analog signal.
- Unexpected loading: Once capture works, investigate why powering the Cosmos appears to reduce the AT-600's measured voltage from 0.288 to 0.126 Vrms.
Only after those issues are resolved should we calibrate REW's full-scale voltage, switch to a 1 kHz distortion test, and calculate SINAD.
Bottom line: We have made meaningful progress identifying the problem. The AT-600 receives the REW test tone and produces an analog output. The unresolved problem is the Cosmos measurement path, including the unexpectedly large voltage reduction when it is connected and powered. We should isolate that path before doing any more processor performance testing.