Nothing blatantly obvious, unfortunately (though I'd look for a better window function like Kaisel-Bessel and can't vouch for Virtins in terms of accuracy and would probably double-check that M/S mode works as expected). Still, if the manufacturer of a high-dollar pro device like the HV-3C with high input and low output impedance bothers to spec the gain to 0.5 dB and you find a constant ~1 dB offset, that sounds fishy to me. Got any other DUT to throw on there?
Thanks for taking the time to answer
I've read in different places that the noise is supposed to be measured with Rectangle window, which is said to be the most accurate for level measurements.
Virtins and REW measurements do match.
I use Virtins because it offers more flexibility and I'm able to automate it.
M/S is basically doing it, which means lower ADC noise by a few dB by averaging it.
But the noise we measure here is way above the RME ADC noise (since it's amplified).
I measured in non-M/S mode with same results.
So I keep the M/S method since that's how I perform my measurements.
The 1dB gain difference is strange, indeed, so I measured it several times.
I see that the source impedance makes a difference.
When using the passive attenuator with high attenuation, I get even lower gain, around 57.9dB with PAD and 59.6dB without PAD.
If I use the RME output directly, I get 60.4dB.
When any combination with the transformer-based JDI at the end of the chain (with or without passive attenuator first), I get 60.5dB.
I repeated that several times for both channels, which repeated exactly the same figures.
In any case, I compare to direct RME loopback, so that should compensate for RME mismatches.
I can't measure the 10mV-level input signal directly with anything like 1dB accuracy, since I don't own a high accuracy multimeter. And to get that kind of accuracy in true RMS, you need an expensive one.
By the way, that gives us a clear idea how impressive the RME performance is.
So, even if I don't give too much trust to the EIN raw figure, its still interesting to look at the trend and to compare with other devices, using the exact same setting.
Also, looking at Millennia specs, they spec it for
Minimum Gain 8.5 dB
Maximum Gain (1.5 dB per step, 36 steps) 60 dB (up to 70 dB on request)
Which is pretty much matching my measurements.
For EIN above, I've used gain figures I got by feeding the Preamp directly from RME, so respect. 8.67dB and 60.43dB.