Audio Input Noise Levels

[Don Palomaki]

Completed a quick test of the XH-A1 XLR input preamps for noise floor.
Battery powered, phantom power off, SD mode, Capture via firewire.
XLR input not terminated (open circuit while may result in higher noise level))
Analysis performed using Adobe Audition
Reading given is the Average RMS level relative to max sine wave record level = 0 dB.

With Audio Level = Automatic gain control
LINE: -86.1 dB
MIC ATT: -71.5 dB
MIC: -67.3 dB
MIC with XLR 12 dB gain-up: -55.7 dB

With Audio Level = Manual, Level control set to ~MID position
LINE: better than -90 dB (near the 16-bit limit)
MIC ATT: -77.1 dB
MIC: -72.2 dB
MIC ATT with XLR 12 dB gain-up: -72.2 dB
MIC with XLR 12 dB gain-up: -61.8 dB
This gain setting was sufficient to provide -10 dB peaks in speech recorded using an AT825 mic placed about 10 inches from a person speaking in a normal conversational voice. The AT835 has a nominal output rating of -47 dB/Pa, and is substantially less sensitive than many commonly used shotgun such as the ME66, AT897 and NGT-2.

With Audio Level = Manual, Level control set to MAX position
LINE: -86.1 dB
MIC ATT: -60.1 dB
MIC: -55.2 dB
MIC ATT with XLR 12 dB gain-up: -55.0 dB
MIC with XLR 12 dB gain-up: -44.4 dB

No dominant frequency apparent in the noise floor.

Per the manual the nominal XLR MIC input sensitivity is 59.5 dB in AGC mode
And is a max of 71.5 dB in manual mode.
To this you can add the XLR 12 db if you really need.

Als reported a quick test using the 3.5mm mic jack (below the mic holder), inserting a stereo mini phone plug to disconnect the build-in mic and using AGC mode:
Noise was -66.4 dB relative to max record level.
With MIC ATT enabled the noise was -71.5 dB.
At the MIC jack sensitivity is -64 dB in auto mode, (-76dB in manual mode)

In summary, in typical video applications mic preamp noise should not be an issue for most users. However, it is best to avoid using XLR gain-up.

[Martin Saxer]

Thank you very much for this! I spent hours trying to find anything about the audio quality of the XH-Series in all the forums, without much success. So I appreaciate your testing a lot.
The line level perfomance of the XLR inputs in manual mode seams quite good and this is very good news for me.

Just one thing: would you mind posting your actual audio samples here? I can read the figures, but hearing it with my own ears would be even better.

Martin

[Don Palomaki]

OK. It is with some fear that folks will apply unrealistic standards or technical interpreation to these file that I upload them.

Attached are two ~5-second "worst case" .WAV files from my tests. They represent the noise floor out the camcorder HDV/DV output terminal (Firewire). I have added a one-second 440 Hz tone burst in the middle of each file that is recorded at -6 dB below max record level as a hopefully meaningful audible reference for your ears. I've uploaded .WAV files because converting to MP3 results in significant artifacts for the noise component changing the sound appreciably, even to my tin ears.

Again, this is a "worst case" noise measurement made as follows:
XLR input open (not teminated)
Manual audio gain.
MIC setting.
Level control set to maximum.

The MicMax440.wav has a RMS noise level of about -55.2 dB below max record level (-49 dB below the tone burst). However, the noise is random, so individual noise spike may reach as high as -45 dB.

The MicMax12.wav file has the XLR Gain-up of 12 dB enabled, so the noise is about 12 dB louder with an RMS value of -44.4 dB and individual noise spikes as high as -34 dB.

Again these are "worst case," such as using a poorly matched mic (one with rather low output) or placing the mic too far from the talent so you have to crank up the gain.

In the case of the MicMax12 file, the MIC gain is probably sufficient to proved full record level from a 0.067 millivolt signal. My comparison, the standard output rating of the ME66 is roughly 500 time higher.

[Don Palomaki]

PS, I didn't post a line-level sample because there is virtually is nothing to hear.

[Don Palomaki]

Martin: I notice in another post that you were interested in line-level noise as well. The attached 5-second file is recorded:

Open (not terminated) XLR input
Line Input setting
Manual gain
Level control at the mid position
One second 440 Hx tone burst superimposed at -6 dB
Adobe Audition puts the peak noise at -90 dB below max record level.
The average RMS value is substantially lower.
The actual noise in the wave form is an occasional non-zero bit.

[Martin Saxer]

Don, thank you very much for this!

Martin

[Ryan Postel]

Don, Your time and attention to detail in the audio tests are amazing.

From my experience recently recording Voice Overs, the amount of background noise is closer to your audio sample of "MicMax12" than anything else. Possibly even louder.

I am using the A1 and a Shure Beta 58A Voice Mic.

I have done some experimenting between the Mic setting and the Mic ATT setting, and it seems, either way I do it, it ends up the same. The noise at the Mic ATT setting only gets amplified when I edit the audio in Audition in order to bring the audio to an audible level.

[Don Palomaki]

Some of this may relate to the Beta 58A and how it is being used. What is the format of the voice over, and what is the mic placement vs. the person speaking. Are several people speaking to the same microphone?

The 58A (I have one) is a dynamic mic with a corresponding output level that is substantially lower than typical (powered) condenser microphones. The 58A output is rated for -51.5 dB/Pa, while shotguns like the Sennheiser ME66 with an output of -35 dB/Pa, have a output that is about 16 dB higher for the same sound level reaching the microphone. One Pa is fairly loud sound. The 58A is best used as a very close mike, or with fairly loud sources.

Lets do a simplified analysis of using the two microphones, and to keep things simple, lets forget about ambient noise and directional effects. Let us also assume for the moment that the baseline noise level is set by the A1's preamplifiers and circuits, not the microphone's internal circuits (self-noise). (This is OK for high quality microphones, but many modest cost condenser microphones have a rather high self-noise level.)

To achieve the same output voice level, audio recorded with the Beta58A will require 16 dB more amplification than the ME66. Thus the background noise from the preamplifier circuits will also be about 16 dB higher.

Solutions may include combinations of moving the microphone closer to the speaker, using a louder voice, using a high quality external preamplifier, or using a different microphone more suited to the application.

[Ryan Postel]

Don,

I appreciate your thoughts.

I have suspicions that the problem lies in the camera, or the settings, rather than the mic. Aside from that, there is one person speaking in an isolated room less than 6 inches from the microphone.

As far as a higher quality preamplifier, this has been something I have tried to avoid for budget reasons. Nonetheless, I am open to explore all possibilities.

I have adjusted listening to the project through tv speakers rather than the headphones, since obviously this is the end medium. I still can very audibly hear the hiss begin and end each time a voice over clip begins or ends.

[Don Palomaki]

Can you e-mail me a short piece (a few seconds) of the voice-over audio as captured from the camcorder (as a .wav file, no audition filters, etc. applied).

And what camcorder audio setting are you using?

Because the beta 58A it is a dynamic mic (no active components inside it) it should be very low noise.

What is the voice-over going over? If it is going over silence, almost any noise floor short of silence will be audible, and one possible solution is to mask it with a steady but not-intrusive noise floor. My recollection is that the old Cool Edit Pro (from before Adobe bought it and packaged it as Audition) documentation had some information on it in this.

[Petri Kaipiainen]

I made similar tests some months ago and posted them here. Only compared line in to mic at normal levels. Copy from my previous post:

Test setup: Loud pink noise from a good stereo system, quiet room, Rode NT-1A super-quiet studio condenser microphone. First pink noise via SD302 to line in, camera pots open, level on SD302 just starting to hit the limiter set at +17 dBVU, then the mixer pot turned down to kill the input (to get the system noise floor). On the second sample the output levels pulled down on the SD302 so that when using mic in the camera pots are at midrange (optimum?). Same pink noise at same levels, also "quiet" recorded after the noise.

Result: The line level gave max peaks at -4.60 dBFS and minimum average RMS at -93.93 dBFS. An amazing dynamic range for a camera, approaching the theoretical limits of a 16 bit system. Almost 90 dB!

With mic levels the peaks were at -3,17 dBFS and with no signal average RMS at -79.21 dBFS. Thus the mic in was 10 dB noisier than the line level even with the pots all the way open with the later.
---------

Quite in line with Don's findings. Line in on XH-A1 is superb.

[Ryan Postel]

Don,
masking the noise is a good idea. I think i may also use a very light soundtrack as well.