REW Beta Release Strange difference between two RTA graphs

[jeannotp]

Hello everybody.

Disclaimers :
I searched a little on the forum but didn't find any clue, sorry if it's pretty obvious.
I'm totally new to REW and in fact pretty far away from audio stuff, unfortunately.

I want to make spectrum graphs from noise in Datacenter (in order to find the best noise protections).
A friend of mine who is very good at this stuff but very busy advised me to buy an Umik-1 (maybe because he's going to buy it back ) and to use REW to make spectrum graphs.

So I use this Umik-1 (with the 90° calibration file) and REW 5.20 on Mac OS 10.15

I have these two RTA graphs.

First graph, first room, the loudest (to the ear it's obvious).



Second graph, second room, less noisy.


I set Forever averaging mode.

If I understand correctly, I have on average:
- for the first room (the loudest) 87.81 dB SPL
- for the second room (less noisy) 78.46 dB SPL

It corresponds perfectly to my feeling. But when I check the black lines, the less noisy room seems much more noisy that the loudest one, the line is always above).

By transparency:


What do I don't understand? (alas, a lot of things)
I have other spectra for these two rooms but it's always like that.

Does the peak sample at 0.00 dBFS mean that the input noise was to high for the Umik and has been clipped?
(it's maybe not appropriate to datacenters...)

Sorry for my clumsy english and thank you for your attention (and by making available free software).

Jeannot

 

[John Mulcahy]

Yes, 0 dBFS peak samples mean the input is clipping. You need to be careful with the input volume setting with the UMIK-1, if the input volume is set to full scale 24 dB of gain is applied to the data, so it will clip easily. Check the volume in the Audio Midi Setup app (in Utilities) and adjust it for 0 dB gain.

Make sure you select the UMIK as the input device and input in REW so the cal file sensitivity figure is properly applied.

Capture images that show the settings you are using (just leave the graph controls panel open when doing the capture).

 

[sam_adams]

As @John Mulcahy said, when you plugin the UMIK, open Audio MIDI Setup.app and look at the settings for the UMIK. The gain should be set at 0 dB:

You may also want to try the build of RC4 that @John Mulcahy posted over here that eliminated dropouts in measurements. If you choose to use that build, before you install it you should run the REW uninstaller then install that build.

If you are trying to measure noise levels and exposure durations you may want to take a look at the logging features of the SPL Meter.

 

[jeannotp]

Hello.

Thank you for your responses.
(I only see now that I wrote 2020 instead of 2021 for the legends of my previous graphs, !)

I made new measures with the gain at 0 dB, like the screenshot Sam Adams made (and with V5.20 RC5).

I also made the measures in different conditions as I was walking around the rooms for the old ones... My friend told me that it was a bad idea (and it's also in the manual of REW, but as I wasn't making measures for a studio... Well. Please don't boo! )
So I cannot compare exactly the before and after.

Anyway, here are (some of) the new RTA traces I made (static and with 0 dB gain), if you're curious about noise in datacenters.

EQX room (less noisy)



IX3 room (the loudest)


It's more interesting because I see that there are especially low frequencies around the air conditioners units but that they disappear a few meters away.
Now that I properly made measures, the RTA graphs seem logical and I don't have any question anymore.

Again, thank you (I hope my english has been about correct)!