Topic: Is system noise a thing of the past, which can usually just be ignored?

[Billy Mumphrey]

...I just don't think a noise floor that low is particularly important for "what we do".  The mic's self-noise really only needs to be somewhat lower than the noise-floor of our quietest recording environment.

I agree. Outside of classical music and nature recordings, you're much better off putting your money and energy into other aspects of the recording process IMO. For any newbie's reading this, don't hesitate to grab that $60 Tascam Dr60d on marketplace. It will be just fine for anything moderately loud (and still acceptable for quieter stuff *shrugs*).

With that being said, I still wanna hear a rock (or loud music) audience recording with CMD 42's (hopefully into a PR-4?) and crank it on my home stereo and see what my ears tell me. One day a 4-mic CMD 42 (or equivalent) setup > 4-channel AES3 recorder will be my go-to rig. Especially if they make a CMC1 version like grawk mentions.

[capnhook]

...I just don't think a noise floor that low is particularly important for "what we do".  The mic's self-noise really only needs to be somewhat lower than the noise-floor of our quietest recording environment.

I agree. Outside of classical music and nature recordings, you're much better off putting your money and energy into other aspects of the recording process

Billy Mumphrey, to your point of putting your energy into other aspects of the recording process, for me it often involves getting set up early to get a baseline of the room noise, and maybe hear the HVAC run

No solution for the chompers yet, Lee 

[Ozpeter]

That you heard noise with only 30db of gain says noise is still an issue. To know if it’s from the recorder or environment, using an analog preamp in front of the recorder and see if it gets louder. Recording silence in the quietest area you can find with varying levels of gain from -10 to 60db would be an interesting test.

I suppose what I am testing is the standalone system - the demonstrable system noise produced from one end (built in mics) to the other (digits).  To do that in such a way as there would be zero chance of the mics actually detecting any sound in the 'silent' environment is not simple (without perhaps a high tech soundproof room).  And that's what is needed to be sure that any noise heard is system noise, not environment noise.  As the gain and thus the recorded noise is fixed in the 32 bit float system, there is nothing to adjust.  The recorder just assumes it's confronted by a rock band and it's been configured appropriately by the makers.  What surprised me when running my test was that despite that configuration, the reverberation of the originally inaudible bird songs (inaudible by me on location) was clearly audible when 30dB of normalisation was applied.  Not just the bird calls, but the reverb thereof.  That sound must be close to the quietest sound one could normally encounter in life.

That scenario would, or should, be improved by connecting external low noise mics and preamps etc, indeed.  But it would be interesting to demonstrate that even using inexpensive built in mics, noise is not a practical problem in most circumstances.  Not all, of course.

I will see if I can devise another test... put the recorder between two mattresses in a countryside bedroom at 3am...

As for using noise reduction, I did actually also have the recorder set up with its built in AI NR being applied to another pair of tracks.  That did drop the background noise (system noise or distant environmental noise, whichever) by 12dB, but it did subtly mess with the birdsong when normalised.  How well it might work when recording music in an air-conditioning compromised performance space (for instance) I don't know.

[Ozpeter]

Actually, the test should be fairly simple.  Record 'silence' in a room.  Then do whatever is possible to reduce the silence further by burying the recorder under something suitable.  Then put the two recordings into one file, normalise as one, and see if the part with the recorder 'buried' is quieter and in what way.  If the recorded noise reduces, it must be environmental.  If it doesn't, the noise must be self noise.  In fact the result might not be that definitive but it might reveal in the quality of the noise what is environmental and what is system.  I could have done that during my original test by simply getting into my car and recording the external 'silence' in there, to compare with what I had recorded outside.  No, not with the engine running...

[Ozpeter]

OK, I've tested along those lines. 

I made an H2e recording with the device lying on the bed in a quiet room.  Then I put a duvet, rolled up, over it, and recorded some more.  Then I took the recorder into the toilet adjacent to that bedroom, which has no windows, and recorded some more silence.  Finally I put the duvet over the H2e in the toilet - yes, really - and made a final recording of silence.

Playing those back, normalised equally, I could hear no difference with the duvet over the recorder, compared to it not being over, and actually to my surprise there was a very slight increase in noise in the toilet.  All in all, I would say that what I recorded was very close to 100% system noise.

Next, I put together a plot of the noise in the park and the noise in the bedroom.  With luck it is attached.  The noise in the park is in red.  In the bedroom, probably just system noise, it's green, and it is a smooth curve indicating that no significant variable environmental noise was present.  Note that the two lines converge at 9kHz.  The noise from the park, which included environmental noise on top of system noise, is a maximum of about 9dB greater than in the bedroom.  The level of system noise goes down from about -72dB at 400kHz to about -101dB at 20kHz.  Remember that the recordings were made at the same analog level, as it's not adjustable.

So having now got a recording which I'm confident is system noise from this device, I went back to considering under what circumstances it would matter.  I spent some time in Audition multitrack, overlaying commercial classical recordings with the system noise.  I chose works which have significant dynamic range, and finally settled on Taverner's choral work "God is With Us", where the final chord in the organ is full scale loud, and where other parts are pianissimo.  I adjusted playback to threshold of pain when playing the end of the work! 

Interestingly, after the final chord, there are at least ten seconds of what seems to be 'live silence' before the original track is faded.  With the H2e system noise added, I couldn't hear that noise at all.  If I raised the H2e noise playback level by about 6dB I was just about able to hear it - with headphones, and remember the music peaks were maximal.  With the H2e noise muted, the original recording's noise in the end silence was about 12dB louder than the H2e noise on its own, though it may well have included some environmental noise from the cathedral where it was recorded.

So, I go back to the suggestion that even this inexpensive modern recorder's system noise (and that from other such devices) is unlikely to be a significant problem, so long as when using it you make sure that the signal to noise ratio is optimal.  You can't do that with digital gain setting, but you can do it with positioning.  Make sure using appropriate proximity that the level reaching the mic, the signal, is as high as possible, with the proviso that when using the device for acoustic recording and the location's reverberation is important, you are not too close.  The end result, in terms of system noise audibility, should be fine in the context of the likely uses of such a device.  And if it's amplified music, I would be really surprised if system noise is any kind of problem at all.

[Gutbucket]

Nice real world test, Peter!

Kevin, yeah on the chompers.  What I'm wondering is if you've found the samples you use to train the nose reduction still useful if they contain ANY amount of background conversation or other transient noise, or if they need to be completely clean of all that to be useful, containing nothing else but the steady state noise being targeted.

Perhaps the details are best discussed in a thread on noise-reduction. I'm just trying to get a rough idea of what works for you and what doesn't based on your experience with it as a regular part of your process.

[Ozpeter]

Before Kevin gets a word in (and before I head to bed), the "AI NR" on this little Zoom device samples the ongoing noise just for a couple of seconds before it's ready to record, and I imagine it is aimed at room environment noise - and its own self noise, which is also lowers by 12dB.  But as I said before, I wouldn't use it for music.

Lastly, I just did a quick test on the digital level settings they recently introduced.  All as expected.  The default "70" level, the max "100" level, and the unintuitive "1" level, which produces what looks like a flat line recording, all amount to the same thing when normalised - and the setting chosen has no bearing on system noise in relation to signal level.

[VibrationOfLife]

Doesn't Curtis have an xlr plug that sends in a frequency he can record and then just pull it up in software to reveal self-noise?