Proper Noise Floor testing procedure

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

mitsos

Well-known member
Joined
May 4, 2007
Messages
2,886
Hi all, I am trying to test some projects here and don't really know if I am doing it properly. I am currently testing mic pres, but hopefully test methods for other dear is similar so this can serve as some sort of guide.

I will start by listing my current method here and hopefully some of you geniuses (genii?) can correct my mistakes.

What I am testing right now is an SSL9K pre. I am trying to do an FFT type plot. First of all, I use PT LE but don't have the necessary plugin so I downloaded Audacity to use their Analyze function. I didn't want to figure out Audacity and I didn't want to read any manuals so I recorded in PT, bounced, imported to audacity and ran the analyzer. Is this ok to do?

I shorted pins 2 and 3 of the input with with a 180R resistor and hit record (did 2 recordings, one at mid gain, one at max gain). I used a pot so I have no real exact idea what the gain is (I suppose I could measure the pot if I need to). 

After analysis, at mid-gain I am getting -88dB.
At max gain I get -71dB.

Does these results sound right? Am I doing anything right/wrong?

What gain level am I supposed to test at?
Is there a standard way to do this? 
Anything else I'm missing?

Thanks in advance!
 
> Anything else I'm missing?

Useful noise measurements are NOT simple.

> What gain level am I supposed to test at?

Measure both Noise and Gain at both maximum gain and minimum gain.

> Is there a standard way to do this?

Yes, measure both max and min gain, but then report only the max-gain number. The EIN is best at max gain. In many preamps it degrades badly as you start to turn-down. But you do not want to admit that to your customers.

(One brand does make a point that their mid-gain noise is less than others.)

You are really going to need a calibrated mike-level signal source. You could measure resistance and calculate preamp gain, but it is too easy to make a mistake. Get a good 0-1V sine generator and a 120dB pad (which is NOT an easy thing). At the most basic level, you put the sine in series with your dummy-mike, and turn-up until the sine is about the same level as the hiss. Since you know the sine, you know the actual input hiss level.

Be careful with the resistor type. Carbon-comp has big excess noise. Carbon-film may be good, and is cheap. This is one of the few places I would use Wire-Wound. In a shielded connector.

A much more practical test is to A-B two preamps. First put matched mikes on them, play the radio, trim for identical gains. Then put dummy-resistors on them, compare hiss. You won't know your exact hiss, but you can quickly know which is quieter (at that particular gain). You can sort all your mike amps for noise. Most should be very close one to another. If one is 10dB hissier than most others, it is sick. If a mike amp is not sick, it is good enough for all ordinary work. 
 
If ya can't hear it at usable levels it is okl!! I personally don't test stuff If I can't hear noise at monitoring levels. But I do test stuff above monitor levels after a new build. To make sure I did ok. I also test phase etc...

In protools LE you should have a Bomb factory noise meter. Leave the filters off.
 
Stagefright, I'm with you. In most cases I wouldn't even test, but I want to know how my projects stand up to what I could buy from the big manufacturers. Thanks for the bomb factory tip, gonna look into that.

PRR, thank you for the answer. It is more complicated than I initially thought.

I started reading about EIN after I read your post and, while it's a bit clearer now, I still have some questions. Can I accurately measure gain by sending a sine wave from my interface through a DI to the preamp and measuring the input and output of the preamp with an oscope? I have done this to test for problems in the pres before, and it seemed accurate. In this case I wouldn't need the dummy load?? Can I use this to do the "most basic level" test you describe?

And about measuring noise at min gain, I went back and did it with my rudimentary "record->analyze" method after I posted but did not edit my post to show this. So in fact I did min, mid, and max gain. At min gain I got nothing on the graph when I ran the analysis, which tells me that my method is not correct, because according to you and everyone else online, noise should be worse at min gain...

thanks again..

 
mitsos said:
Can I accurately measure gain by sending a sine wave from my interface through a DI to the preamp and measuring the input and output of the preamp with an oscope?

Accurately is a big word, and most 'scopes aren't too good at showing mV-level signals, but for low to moderate gains this procedure should work fine. For higher gains you can build a simple U-pad (like Example 1 from this page), connect it between DI and pre, and have the scope measure the difference in level between the signal before the pad and after the pre. Mic pre gain is now the measured difference plus the pad's attenuation. Oh, and when measuring with scopes ground levels and probe grounds may matter (unless you happen to have a few differential probes lying around, or you are sure your scope is one of the few models with true independently floating inputs).

(A simpler setup may be to work the other way around. Build a few pads, say 20dB, 40dB, 60dB, and a Y-cable. Put one of the pads before the pre, connect the Y-cable to your signal source, the pad and a VU meter. Connect the output of the pre to another VU meter. Adjust pre gain until both meters show the same level; at this setting the gain of the pre equals the attenuation of the pad. For VU meters you can substitute metered DAW input channels, of course).

[quote author=mitsos]At min gain I got nothing on the graph when I ran the analysis, which tells me that my method is not correct, because according to you and everyone else online, noise should be worse at min gain...[/quote]

No, that's to be expected. Noise is referred to the input of the device. Take the THAT1510; according to its data sheet it has 17dB more input-referred noise at a gain of 20dB than at a gain of 60dB. However, with no signal present its output will show 23dB more noise at 60dB gain than at 20dB gain. This may sound weird, but consider that (barring clipping) an input signal will appear 40dB stronger at the output at 60dB gain than at 20dB gain, so the signal-to-noise ratio will be better at higher gain, as expected.

JDB.
 
PRR said:
Useful noise measurements are NOT simple.
Do not agree. In the case of good mic preamp (input impedance >>600 Ohm), you could make
simple 600 Ohm resistor shorted by this: http://www.imagesco.com/catalog/sensors/tilt.html
to do simple estimation of input noise resistance {to make two measurements - one with
resistor opened and the other  with resistor shorted between hot and cold termination of mic connector. That two measurements should be done at one preamp gain.

If noise energy doubles {+3 dB} increase while resistor is open  then noise resistance
of your preamp is equal of 600 Ohm. If smaller change of noise is seen while short X open
the resistor, preamp is worse.

For many simple preamps holds, that input noise resistance rise with decreasing gain.
This {paradoxial} situation is result of resistive feedback, the greater is feedback,
the lower is gain and the greater is amount of noise from the feedback resistor.
 

Latest posts

Back
Top