measuring mic pre noise

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

briomusic

Well-known member
Joined
Mar 1, 2009
Messages
411
Location
London/Berlin
I am trying to find a way of evaluating all my mic pres (most of them DIY) in terms of EIN (equivalent input noise), without shelling out for any additional equipment or software. I want to stress that this is not a 'proper' test setup, but I think my 'errors' are the same for all of the mic pres under test, giving me a relative value rather than an absolute one. I just wanted to know which mic pre I should go to for very quiet accoustic signals/ribbon mics etc.
This was my test setup/procedure:
1. Connect Behringer  :D test tone generator/cable tester to mic pre input with a -50dB signal.
2. Adjust gain to a specified output level as measured by Electroacoustics Toolbox (turned out to be -8.6 dB in my case)
3. Disconnect test tone generator and replace with dummy source (120 Ohm resistor between XLR pin 2-3)
4. Measure mic pre output using A-weighted curve on Electroacoustics Toolbox' sound level meter.
5. Repeat steps 1-4 with each mic pre.
6. EIN is measured noise level minus gain
Device under TestInput (dB)Output (dB)Gain (dB)Noise (dB)EIN (dB)
EZQN* (DIY)-52,6-8,644-79-123
API 512c-52,6-8,644-78,3-122,3
SSL9k (DIY)-52,6-8,544,1-103,8-147,9
Chandler LTD-1  -52,6-8,644-90,7-134,7
DRIP 47 (DIY)-52,6-12,839,8-90,9-130,7
*"EZQN" is my custom 'channelstrip' with a partial EQN (minus line input transformer and output stage) inserted into the 'trimmer insert' of an EZ1290.

My biggest surprise was the fact that my EZQN performed quite a bit worse than the Chandler LTD-1, both of them being Neve 1073/84 clones of sorts. I am happy to hear opinions and suggestions to improve my test rig/procedure. Loving the quiet gain of the SSL9K  ;D

 
The A weighting should help ignore bandwidth related measurement errors (for better or worse).

What is the output impedance of the signal generator? Mic pre's inputs should be terminated with 150-200 ohms for measurement. I would roll my own resistive pad to drop level down and control source impedance.

JR

PS: That 20 dB difference seems large to me.
 
Hi John,
thanks for your input. I have no idea what the output impedance of the behringer is, the test tones are kinda thrown in as an extra, it's primarily a cable tester. (damn handy, though, to put a mic level noise down a line from the other side of the window).

I guess the more predictable/reliable alternative would be to use a known balanced line output and pad that down to mic level. Rick Chinn's paper recommends a U-pad with 2 times 7k5 resistors inline, and a 150R between hot and cold to achieve a 40dB drop. I'll also make a new dummy source with a higher resistor between pins 2 and 3 as you suggested.

And for my next trick.....
planning to measure microphone noise. am planning to point the mic at a small practice amp putting out a sine wave test signal. then turn the amp off and measure how much 'quieter it just got'. repeat with different mic in same position and with same mic-pre gain setting. viable method?

thanks
briomusic
 
The pad can effectively balance the feed... I would use slightly higher than 150 in the short leg to get 150-200 overall.

Keep in mind your mic noise test will also pick up room noise which may be significant.

Consider perhaps a mic instead wrapped in rags and inside a box, vs. a fixed resistance termination. (again 150-200 ohms).

JR
 
-147 dB(V?) from 120 ohms is not creditable.

A well-made SSL9k should be within a couple dB of quietest-possible.

 
Yep, the SSL 9K result looks goofy. Note that dB is not a meaningful unit for an EIN measurement (actually not a unit at all)--we want dBu.

Also measuring EIN with different source resistances (50 to 1k Ohm) and at lower gains will tell you much more about real world performance of mic preamps. If a particular design measures well at high gain it does not necessariliy do so at low gains; same applies for source resistance variations.

Samuel
 
Okay, Electroacoustic Toolbox, offers measurements in LP(V) or Leq(V). dbV = dbu if I am not mistaken? The help file says:
The status of this radio button determines whether a time-weighted sound level (Lp) or an equivalent sound level (Leq) is measured by the SoundLevelMeter.
Which one should I select?

Other things to improve next time:
  • use padded line signal instead of behringer test tone
  • measure at two gains (20dB/40dB)
  • use higher resistance for dummy source (200 Ohm)

The -147dB were never really measured as such, but are the result of substracting the applied gain from the measured noise level. I guess (additionally) measuring at another gain setting would maybe show that some of the noise is independent of gain, thereby rendering that value wrong.

I am very thankful for all the advice, but please keep in mind that my primary objective is to find the quietest mic pre in my studio in relative terms, so the next time I have to record some avantgarde musician scratching his viola with a feather, it doesn't sound like I have a tap running in the background  :-[
 
0 dBV is a relative voltage ratio referenced to 0dBV=1V.

0 dBu is referenced to 0dBu=.775 (note: this is same voltage as 0 dBm terminated by 600 ohm load).

JR
 
briomusic said:
Okay, Electroacoustic Toolbox, offers measurements in LP(V) or Leq(V). dbV = dbu if I am not mistaken? The help file says:
The status of this radio button determines whether a time-weighted sound level (Lp) or an equivalent sound level (Leq) is measured by the SoundLevelMeter.
Which one should I select?
Leq is a calculation of the average sound level over a 1 second period of time. Since the noise of your preamp is almost continuous, that should not make much difference. But are you sure that there is no weighting applied to the measurement?
Acoustic measurement are by default A-weighted. For electrical noise measurement (white-ish noise), it indicates about 10dB less than unweighted.
Other things to improve next time:
  • use padded line signal instead of behringer test tone
This is not very important.
  • measure at two gains (20dB/40dB)
  • use higher resistance for dummy source (200 Ohm)
You should also measure the input impedance of your preamps. The worst preamp I know in this respect is GT's The Brick, with ca. 500r input Z. A standard 200r microphone loses more than  2dB of useful signal compared to a mic pre with the recommended 2kohms input Z.
The -147dB were never really measured as such, but are the result of substracting the applied gain from the measured noise level.
That's always how it should be measured, and I presume that's how you've proceeded for the others, so obviously, something must be wrong. Noise from a 120r resistor is -131.8 dBu.
please keep in mind that my primary objective is to find the quietest mic pre in my studio in relative terms
That's understood, but even then, your basis of comparison for the SSl 9k is erroneous.
 
Thanks to everyone that's chipping in!
0 dBV is a relative voltage ratio referenced to 0dBV=1V.

0 dBu is referenced to 0dBu=.775 (note: this is same voltage as 0 dBm terminated by 600 ohm load).
my bad, you're completely right of course! well the program deals in dbV, so it's still voltage based.

@[silent:arts] - I'll definitely take you up on that offer. is your measuring equipment portable? Otherwise I'd have to see which pre amps I could bring around (definitely my new snazzy portable SSL9k  8) )

@PRR - what is 'quietest possible'? combined self-noise of all the resistors and opamps?

@abbeyroad - how do I measure the input impedance of a mic pre?
 
briomusic said:
@abbeyroad - how do I measure the input impedance of a mic pre?
You feed it with the lowest impedance you can, set the gain as to have a well readable output; this is V1. Then you insert a resistor in series with the source in order to produce 6dB attenuation; this is V2. The value of the resistor is about equal to the input impedance.
The exact calculation is the following: with Z= input Z of mic pre, G=gain of mic pre (in Vout/Vin, not in dB), U=generator voltage (all voltages in volts, not in dB)
V1=U.G.Z/(Z+R1)
V2=U.G.Z/(Z+R2)
that resolve to V2/V1= (Z+R1)/(Z+R2)
and finally Z=(R2.V2-R1.V1)/(V1-V2)
As you can see, the result is independant of the source voltage and the preamp gain, so you can set them at your convenience, but you have to make sure there is no clipping.
I also recommend that you monitor the signal acoustically for all these measurements, because hum can offset the measurements significantly.
 
> Acoustic measurement are by default A-weighted.

Acoustic noise measurement are usually A-weighted.

A-weight is an approximation of hearing curve for sounds much softer than speaking levels.

I've seen A-weight on LOUD sounds. There may be some validity if the question is "hearing loss", but that's an accidental and perhaps incorrect use of A-weighting.

> measuring ... ... at lower gains....

For simple (one gain control) preamps, you should be able to measure twice. Max gain gives input hiss. Min gain gives output hiss. In-between, input hiss falls as gain is reduced but output hiss is constant, sum the two hisses. (The actual source of hiss may not be IN the output stage, it is often NFB resistors, but conceptually it is not affected by gain.)

> next time I have to record some avantgarde musician scratching his viola with a feather, it doesn't sound like I have a tap running in the background

Can this guy scratch VERY consistently? Set up one good ordinary mike and your six preamps. Set them all for EQUAL scratch loudness (why he has to play consistently). Tell him pizza is in the breakroom, and compare the hiss levels.

Highly compressed pop tracks through a speaker which can be turned WAY down, back-background level, may be a more consistent reference tone.

> scratching his viola with a feather

Put a pickup ON the viola.

Sharp-cut at 3KHz or 6KHz.

I'm crying too. There is a modern concerto where the whole third act is played as soft as possible. With expert fiddlers the acoustic level is right in room-noise. It's actually interesting musically. Putting piezos on the bridges was not an option (live public performance by virtuosos with antique Italian instruments). It's 40 minutes of strong and soft melacholy moaning, so catchy that over the years they did it again and then again. I never liked the recordings, but we got good crowds in the room and that was the point.
 
The definitive article on this, (I think so anyway), is by Paul Buff for REP magazine in 1976(77?). It was the beginnings of the discussions about consoles and mic pres and ersatz mic pres for (mainly) MCI 500 series, and begat the TransAmp and a good meter that Valley People manufactured. But it did give a easy to understand and reasoned argument for doing it properly.
 
+1... the Paul Buff article was one of the better discussions of noise measurements including bandwidth and source noise.

Prosounweb has started re-running some of the classic old RE/P articles so maybe this one will come around again.

JR
 
The articles under discussion are:-
Console Noise Specifications  --- Fact or Fiction?
and
Towards Microphone Transparency.

They are from the same issue of RE/P. The first one is the rant about noise specs, BUT it does tell you how to do it; and that some of your figures (Briomusic) are exactly what he is talking about!
The second one has a lot of useful information but is mainly about how revolutionary the TransAmp is and that everyone should use them, especially people with MCI 500A consoles!

There is later article further to this called "Perceiving Audio Noise & Distortion" which shows why the A-Curve is a non-runner for measuring noise, (you should be using the CCIR one), but people don't like the results it gives. You will also learn a great deal about console gain structure in this one.

If someone else chooses to post these, fine, but I'm not.
PC
 
Following on from the post above,

http://www.beis.de/Elektronik/Electronics.html

There is a design for a good set of filters for just the BrioMusic application. So if you do not have a Neutrik A2 or an old Sennheiser "pegelometer" (I can't remember the number) this is worth building.

PC
 
Off-topic: the www.beis.de site has some notes about low voltage noise.

http://www.beis.de/Elektronik/LNPreAmp/LNPreAmp.html

Reports low-low Vn with a 400V 8A part.

 
I saw that and it seems odd that a BF459 would be that low noise.  The parts only have a beta of around 25, most low noise parts I am familiar with are much higher current gain (several hundred).

I guess anything is possible.

JR

 

Latest posts

Back
Top