Excessive high frequency noise in RME Fireface UFX+ in 192 kHz operating mode (ADC)

[kartoffelmann]

Is this a normal noise floor for an RME Fireface UFX+? This looks insane to me. Almost 40 dB difference between the typical 20-20k range and the limit near 96 kHz. At the high frequency range this is actually a higher noise floor than my old Fireface 800. I'm asking because I'll have to return it if it is damaged. I've seen devices at less than a quarter of the price of this unit that have a lower noise floor at high frequencies.

Note: This is with channel gain set to 9 dB to achieve a -10 dBV reference level.

Edit: This is of the ADC, not DAC. Sorry for the confusion.

 

[RayDunzl]

Is this a normal noise floor

Maybe.

Not calibrated for level

Focusrite with (assumed) USB noise spikes.

---

Ooh....

 

[Blumlein 88]

Perhaps @MC_RME could answer your question. He is with RME so should be able to find out for you.

In case he doesn't see this notice you could start a conversation with him using the link in this post.

 

[kartoffelmann]

@RayDunzl

Pardon, what do you mean by calibrated for level? I can choose between +4 dB and "Lo Gain". The older ones used to have -10 dBV too but they dropped it so now you have to use digital gain and lose 9 dB of dynamic range. In other words, without gain the curve would simply be 9dB lower (I tried this). But for instance switching between +4 and "Lo Gain" does not seem to affect the noise floor, presumably because it's an actual electrical switch between different sensitivities and independent of the ADC.

@Blumlein 88

That would be awesome, thanks!

 

[RayDunzl]

Pardon, what do you mean by calibrated for level?

There are several display choices in REW's RTA display - the SPL display can be (and has been here) calibrated to arbitrary 0dB levels - so don't pay attenition to the absolute level in the graph above.

 

[kartoffelmann]

There are several display choices in REW's RTA display - the SPL display can be (and has been here) calibrated to arbitrary 0dB levels - so don't pay attenition to the absolute level in the graph above.

Ah, gotcha. Thanks for explaining.

 

[dinglehoser]

I run into this same issue with USB packet noise - 8khz fundamental, with harmonics that keep going forever - when there's a ground loop somewhere AND the USB source is noisy. I actually just noticed this with my own test setup with two otherwise well-engineered devices, and am presently trying to see if the issue is more about process than the hardware itself. Will keep you posted on what, if anything, works best.

 

[JG2020]

RME Fireface UCX 192K noise floor.

 

[dinglehoser]

As I suspected, it had to do with my measurement setup. Using an Apogee Element 24 as ADC, my Focusrite Scarlett 18i8 v2 was initially showing an 8khz spike at -90dBFS, with harmonics not far below that. They were plugged into the same laptop. Turns out that was causing a ground loop that not even a balanced connection could solve completely. Using a different machine to drive the 18i8 solved the problem.

 

[kartoffelmann]

Oooh, I'm sorry, I should have mentioned that the picture of the noise floor I've posted is of the UFX+'s ADC.

 

[mansr]

Is this a normal noise floor for an RME Fireface UFX+? This looks insane to me. Almost 40 dB difference between the typical 20-20k range and the limit near 96 kHz. At the high frequency range this is actually a higher noise floor than my old Fireface 800. I'm asking because I'll have to return it if it is damaged. I've seen devices at less than a quarter of the price of this unit that have a lower noise floor at high frequencies.

View attachment 60731
Note: This is with channel gain set to 9 dB to achieve a -10 dBV reference level.

Edit: This is of the ADC, not DAC. Sorry for the confusion.

I can't say for sure that there's nothing wrong with your unit, but noise like that is not uncommon in sigma-delta ADCs.

 

[kartoffelmann]

@mansr I see. Well, on the RME forum the admin told me that it's normal too. So now I'm conflicted on whether I'm being paranoid.

With the 9dB of gain needed to reach -10 dBV input level, the high frequency noise on the UFX+ is almost 15 dB worse than on the FF800 and even appears to get outperformed by a now cheap Steinberg MR816, though sadly I don't have a 192 kHz test recording made of that one. Even a Korg Audio Kontrol 1 has a bit less noise in some areas of the spectrum, and that unit is as cheap as it gets. Of course noise floor isn't the only metric, I understand that, but still.... that's thing costs 1% of the UFX+. And the (unweighted, I presume) level meter in Audition shows -71 dB for the FF800 @ 192 kHz and -58 dB for the UFX+ @ 192 kHz. I feel like that's a serious downgrade, but I'm being told by the RME guy that this is not a big deal and made up for by all the other features ... maybe I'm suffering from some Dunning-Kruger here, but I completely disagree.

So, with what you said about sigma-delta ADCs, what kind of ADC for a reasonable price point might I look at to get a lower noise floor in high frequencies? I am sadly not very knowledgeable about all these technicalities.

Edit: Pardon, it's of course a NI Audio Kontrol 1, not a Korg.

 

[mansr]

For most purposes, the noise level below 20 kHz is more important than above 80 kHz. I'd pick the UFX+ over the Audio Kontrol 1 (that looks like a Cirrus Logic ADC, btw), no question, if price wasn't a factor. There is rarely, if ever, any good reason to record at a higher rate than 96 kHz anyway.

So, with what you said about sigma-delta ADCs, what kind of ADC for a reasonable price point might I look at to get a lower noise floor in high frequencies? I am sadly not very knowledgeable about all these technicalities.

They're all sigma-delta, so a high-frequency noise hump is unavoidable. That said, some ADCs have a flat noise floor extending higher than others. The TI PCM4220, for example, is flat to beyond 50 kHz, and above that the noise level rises by about 20 dB.

 

[dinglehoser]

I really like the Apogee Element series for the price - great DAC and ADC. Only hitch is that it's Mac only ... I wish they'd write some PC drivers for it. I'll run a quick FFT of the noise floor of my Element 24 and post.

 

[dinglehoser]

Did my best to match your scaling and parameters (of which the FFT length is most important). This is at -10dBV reference, 192khz sample rate.

 

[kartoffelmann]

@mansr That PCM4220 looks really promising. What's a good unit that has this or any other way to get this to work? I would need a way to use it as an input and route it into the UFX+ via ADAT or AES/EBU at 192 kHz 24 bit.

Do I understand correctly that the Tascam UH 7000 has this? If so, is that unit usable for my intents?

@dinglehoser Thanks, much appreciated. Could that unit just be used as a standalone converter?

Edit: Looking at the values, they seem pretty close to the Fireface 800. Would probably buy a new FF800 before buying this. Price would be about the same. Thoughts?


In general, any thoughts about the MOTU 8A? I have seen some measurements here and other places, but no noise floor at 192 kHz.

 

[dinglehoser]

It can't be used completely standalone, unfortunately. That said, as long as it's plugged in to a Thunderbolt Mac, it'll go about acting like a standalone ADC once configured (I'm currently using it like this as an ADC for my phono pre).

I think "better" depends a lot on what you're trying to do with it, and why. FFT noise floors are fake and can be manipulated by increasing or decreasing the FFT length. For my purposes, I primarily look at whether/to what extent there are idle tones, as well as the actual unweighted floor within the audio band. The FF800 appears to have a few spurious tones above the audio band, which the E24 doesn't have. They're inaudible, though, so of more import to me is the in-band unweighted noise floor. Not sure what it is for the FF800, but it looks like it's around -108dBFS for the E24.

EDIT: -108dBFS at -10dBV; -115dBFS at +4dBu with no visible idle tones. Noise floor starts increasing above 35kHz.

 

[mansr]

@mansr That PCM4220 looks really promising. What's a good unit that has this or any other way to get this to work? I would need a way to use it as an input and route it into the UFX+ via ADAT or AES/EBU at 192 kHz 24 bit.

Do I understand correctly that the Tascam UH 7000 has this? If so, is that unit usable for my intents?

The UH-7000 is more or less discontinued. It also has an issue with noise from the built-in power supply.

 

[kartoffelmann]

@dinglehoser Those spurious tones might be from still having another device attached that was just powered off a few seconds earlier. Sadly that's the only test file I still had. Overall I think the E24 might be a slight upgrade, but not really worth it for me. Thanks for the effort tho!

@mansr Mh yeah I read about that. However I did end up finding this thread: https://www.audiosciencereview.com/forum/index.php?threads/cheap-high-quality-adc.101/
I think I'm in love. It's probably gonna be this unless I have missed some important detail.