Measuring preamp noise

[JAY X]

Hi!

I want to know if my procedure to measure noise output from a 500 series preamp is correct, as i don't have much experience, nor much equipment...

1). I attach a 200R load between pins 2&3 at the XLR input.
2) I record the output noise from the preamp to the DAW at several gain settings: Fully CCW, 9 o'clock, 12 o'clock, 3 o'clock and fully CW.
3). After recording the noise i run "statistics" on the audio DAW, and i write down the MAX RMS value.

3.1) In my case, at 5 gain settings, i have the following numbers:

Fully CCW : -116,75 db
9 o'clock : -114,20 db
12 0'clock: -91,35 db
3 o'clock : -76,93 db
Fully CW : -59,57 db

The difference between the MIN/MAX gain settings is 57,18db, nearly 60db.

4). I also tested the gain noise with a dynamic/condenser microphone attached. An AKG C5, whose self noise is about 25db. So, let's see if i understand correctly...If i test the mic noise with a condenser mic and 48v phantom, the noise figures i got with a 200R load will worsen by at least the amount of self noise of the microphone (25db). So -116,75db +25db = 91,75db...and so on ¿correct?.

5). ¿Is it better to use a dynamic mic for testing?

6). It is often said that the last quarter turn is the noisiest, but as i don't have other preamps to compare to, i don't know if these figures are normal, average or bad...

7). Another problem is Rail noise....the more gain more rail noise... RC filters? ferrites on rails?


¡¡Thank you for your advise!!

Jay x

 

[user 37518]

First, do not connect a real mic because you would be adding ambient noise and the mic's noise itself, otherwise, just short the input to ground if you want to measure the pre noise alone or with the resistor if you want to consider the effect of the mic's output impedance. Secondly, you really need a very low noise amp at the output of your pre when measuring noise at low gain settings, otherwise, the line input of your interface will be contributing extra noise, this is not really a problem when you crank the gain up to 60dB since the noise of the pre will overpower the contribution of the interface amp, because noise adds in a RMS fashion or the root of the sum of squares (considering there is zero correlation).

Wayne has a good explanation on how he measures noise with a Focusrite interface at the proaudiodesign forum, I believe he also mentions a schematic for the low noise amp. Usually, the preferred candidate for this job is an opamp coupled with a low noise discrete transistor.

To really test the self noise of the mic or to make tests with a real mic you need an anechoic chamber, just having a quiet studio room won't suffice.

Where is the line noise coming from? is it from the pre or the PSU feeding the pre? if the latter, you could use RC filters if the current demand is low. But in the end, its just a matter of how things work, the higher the gain, the higher the junk will be raised, there is no escaping from it. If you want to get rid as most as possible of line noise, try feeding it from batteries, which is pointless IMO, because you will never use the pre in this scenario.

Finally, you mention the MAX RMS value, what exactly is that and how the program computes it? I guess it means that the program uses a prescribed integration time and gives you the maximum of all readings, but, what exactly is the integration time it uses? this is not the best way to measure the RMS value of a signal IMO, taking the MAX RMS might be useful to measure the maximum level in something like a song, but its not the best to accurately measure the RMS value of a random signal such as noise, can the software give you simply an RMS value of the entire signal? no MAX nor MIN? If so, that is the value you want. In fact, in terms of random noise, there is really no reason why you should select the MAX RMS rather than the MIN RMS value, what you really need is a statistically significant RMS value. To do this, you should add the sum of the squares of each sample and divide them by the total number of samples, then take the square root, this usually involves using something like MATLAB/Octave, the longer the signal, the more statistically representative the value will be... Or take 100 captures 1.4861 seconds long each (or, to be precise 1/44100*65536 seconds long, each), then take the RMS value of each and average the total. THAT is much more significant than just taking the MAX or MIN RMS of a single capture within a prescribed integration time.

A guerrilla way to do this in your DAW would be to record 100 clips of short duration (if you knew the integration time of you DAW used to measure RMS, that is the preferred value you should use for the duration of the clip), then, write all the MIN or MAX RMS values (if the clip length is exactly the integration time, both should be the same value, considering this is the way the DAW computes the MAX/MIN RMS) for every clip, if your DAW gives you the value in dB, you might want to convert them to a ratio first by using 10^(db/20), then, sum all of them and divide by 100 and convert back to dB by using 20*log(value_you_got). Or you can just add them as dB and divide by 100. The latter should yield different results.

The simplest way would be if your DAW is able to give you simply the RMS value of the entire signal (no MIN nor MAX, nor integrated within a prescribed time, but rather integrated through the entire length of the signal), record 60 seconds of noise, take the RMS value and that is your real RMS value.

Another (tiresome) method is if you have a scope which can write values to a USB flash memory or if it can be connected to a computer, measure the noise with it (preferably with a low pass filter at the output of the pre or if your scope has digital filters use one of them), take many shots, how many depends on you scope's samples per second, in my 2GS/sec scope, if I select a time base of 20ms/div, it says it is working at 50kS/s, which allows me to visualize up to 25KHz (considering no alias because of the LPF), and since the screen has a total of 10 divisions, the screen shot time length will be 200ms, so lets say I take 100 screen shots equivalent to 20s of signal and save them to a file or send them to the PC, then, if you don't have Octave/Matlab, you can use Excel: for each shot square each sample value and sum the total, then, sum the 100 values you computed from each shot, divide the total by 100*(number of samples in each shot), take the square root of the resulting figue and convert to dBu or dBV.

 

[JohnRoberts]

It might be useful to make a resistive pad so you can feed in an attenuated input signal. With a -40dB pad you can estimate the preamp gain from output level. Design the pad to have 150-200 ohm source impedance and it can serve as the dummy mic input termination.

JR

 

[JAY X]

WOW!...it was a simple question...
Thanks! Great advise to take!!

My daw also gives me the average rms value: at all gain settings, with the 200R load at input, the RMS average value is: - infinite.db.

But this is with short noise samples. Still have to try longer 100 seconds samples.

Anyway, every time i record a mike and turn the gain over 12 o'clock noise starts becoming nasty...despite the nice figures from a resistor load.(due to microphone self noise).

The point is that i start feeling insecure, as the load test says something and the audio recording from a microphone another different...

Anyway i have to test the preamp at a friend's studio, more ears, better judjements, working alone as me, i think i loose perspective of things..

Jay x

 

[user 37518]

WOW!...it was a simple question...
Thanks! Great advise to take!!

My daw also gives me the average rms value: at all gain settings, with the 200R load at input, the RMS average value is: - infinite.db.

But this is with short noise samples. Still have to try longer 100 seconds samples.

Anyway, every time i record a mike and turn the gain over 12 o'clock noise starts becoming nasty...despite the nice figures from a resistor load.(due to microphone self noise).

The point is that i start feeling insecure, as the load test says something and the audio recording from a microphone another different...

Anyway i have to test the preamp at a friend's studio, more ears, better judjements, working alone as me, i think i loose perspective of things..

Jay x

Hmmm, no, that average RMS value doesn't sound to me like the real RMS value, perhaps it is just takes the average within a specified time window. However, that is much better than just taking the MAX RMS, you should definitely record a longer signal and use that instead.

What do you mean the audio recording from a microphone? You can't measure the noise of the pre with a microphone connected at the input, the microphone will contribute its own noise plus the ambient noise, it doesn't matter how quiet is your studio, you would need an anechoic chamber to do reduce the ambient noise to a minimum, and even still, I don't know how useful would be this, you should be interested on measuring the pre noise by itself, not the whole specific mic + pre arrangement. If you want to model the effect of the mic more realistically, you might want to create a mic electrically equivalent model, which would necessarily involve inductors, capacitors, etc... this sounds a bit extreme to me if you just want to test the noise of your pre.

 

[JAY X]

Ok!. But given these figures, is it possible to arrive at a conclusion about how noisy is the preamp?

A few weeks ago, I went to a studio, to test a preamp i built. The first test they always do is to atach an SM58 dynamic mic and crank up the gain to the max...to test the preamp noise and compare it to others they have. That makes me nervous...

 

[user 37518]

Ok!. But given these figures, is it possible to arrive at a conclusion about how noisy is the preamp?

A few weeks ago, I went to a studio, to test a preamp i built. The first test they always do is to atach an SM58 dynamic mic and crank up the gain to the max...to test the preamp noise and compare it to others they have. That makes me nervous...

You need to test the electrical noise of the amp without a mic, just attaching a mic and cranking the gain means nothing, the ambient noise plus the induced electrical noise captured by the mic will completely overpower any noise produced by the preamp. That is not the way to test things. You should test with a source resistance (you can use different values to emulate different mic impedances) if you want to account for the mic impedance, but connecting a mic, cranking the gain and having a listen is barbaric... Considering the level of sophistication of your friends, just make your pre with less gain and they will love it for its low noise...

BTW, from the screen shot you provided of your DAW it is confusing, it says Min. and Max Energia RMS, energia is not the same as voltage. Perhaps its just a wrong label or translation to spanish.

 

[JAY X]

Energía means energy, so it may be measuring loudness...

 

[JAY X]

Less gain...good idea!..
Really, now i understand why people like behringer audio interfaces...they only offer around 30db of clean gain...

 

[user 37518]

Jay, I am mexican and spanish is my native tongue, I am very well aware what energia means, energy and loudness are also not the same, it is also different from voltage. I guess that they mean RMS voltage but say energy, the formulas to calculate energy and RMS voltage are different, the latter involves division by the total number of samples and a square root, the former does not

 

[ruffrecords]

Ok!. But given these figures, is it possible to arrive at a conclusion about how noisy is the preamp?

Not really because the values you have are relative the the sensitivity of your interface. You really need to make absolute measurements in dBu rather than in dB relative to something unspecified.One way to do this is to connect a real meter to the mic pre output. Then feed in a tone at the input and adjust the gain for 0VU on the meter which is the same as +4dBu. Keep everything the same except disconnect the meter and feed the mic pre into your DAW and see what the DAW meter reads. Whatever dB this reads is equivalent to +4dBu. You can now convert andy DAW meter reading to dBu which is what you want.

Cheers

Ian

 

[Newmarket]

Ok!. But given these figures, is it possible to arrive at a conclusion about how noisy is the preamp?

A few weeks ago, I went to a studio, to test a preamp i built. The first test they always do is to atach an SM58 dynamic mic and crank up the gain to the max...to test the preamp noise and compare it to others they have. That makes me nervous...

As others have implied / stated - you should not be introducing a real transducer into the test setup for fairly obvious reasons of ambient noise. For accuracy you should put a resistance (150 / 200 Ohm depending on what you are using as a comparison reference - value tends to differ between North America and UK for example) across the input. If you simply short circuit this then you are not taking into account the input current noise of the preamp that develops a noise voltage across the transducer impedance.

 

[Matt Syson]

The test measurement needs to have properly designed characteristics for which an audio interface may not be correct. the bandwidth needs to be defined usually -3dB points at 20Hz and 20KHz and using a true RMS rectifier. Then some manufacturers quoted alternative measurements like Quasi Peak and also 'non flat' response curves like CCIR, 12db boost at 7kHz (IIRC) Understanding the different rectifiers and weighting can reveal different aspects to the gear beaing measured. You of course have to know (or measure) the gain that your preamp is giving before you can measure the noise and then compute the 'EIN' or noise figure before you can say whether it is in tolerance or acceptable. Having the input 'terminated' with a resistor (150 or 200 Ohms typically, is also necessary as it will give results that may be misleading if you don't. May be wrong but I think the CCIR curve was supposed to make the 'noise intrusion' more meaningful as peoples hearing is more sensitive in this band. It also conveniently reduced thev effect of mains frequency 'hum' which could make some gear look better on paper. The 'self noise' of a 200 Ohm resistor is -131.5db when measured 20 - 20KHz with a RMS reading meter/rectifier. (or is that 150 Ohms? It is late and I forget now). A 'decent preamp will produce an EIN of about -129.5 dbu (20 - 20K) when terminated correctly.
The EIN being the sum of the noise value you measure PLUS the actual gain. So for an EIN of -129.5dB for an amplifier with 60dN gain you would measure noise as -69.5dBu. Of course you have to remember if you are working with 0VU (+4dBu) as a reference level. An uncalibrated 'soundcard' or interface would simply add more uncertainty to whatever you are doing as you won't know what the band limiting is or the rectifier characteristics.
Matt S

 

[ccaudle]

. I attach a 200R load between pins 2&3 at the XLR input.

That is appropriate. The thermal noise of a 200 Ohm resistor in a 20kHz bandwidth is 0.25 uV, which is -132 dBV or -130 dBu.

I record the output noise from the preamp to the DAW at several gain settings: Fully CCW, 9 o'clock, 12 o'clock, 3 o'clock and fully CW.

To be useful for comparison you need to know (or measure) the gain at those settings. Giving a noise figure for "fully clockwise" makes it meaningless to compare to another preamp with a different gain range.

Also, at lowest gains the noise is likely to be from a combination of the preamp and the line input on the audio interface. At high gains the noise should be enough above the input noise that the preamp will dominate the measurement.

To start you should measure the noise of your interface with pin 2 and 3 shorted, and with the 200 Ohm resistor. That will give you the noise floor of your interface. You can not accurately measure noise below that value.

After recording the noise i run "statistics" on the audio DAW

What is the input sensitivity of your interface? Decibel values are a logarithmic ratio, they don't tell you much unless you know what the reference value of the ratio is.
Common reference values are 1V for dBV, and 0.775V for dBu (for historical reasons it isn't worth explaining right now).

Your DAW is operating on digital audio samples, so the only available reference is the maximum value of a digital word, dB FS, decibel referenced to full scale output.
The value quoted for low gain, around -116 dB is about what you would expect for input noise of a decent quality AD converter, but without knowing what voltage level results in full scale you have no way to compare to other analog equipment.

self noise is about 25db
...
So -116,75db +25db = 91,75db...and so on...

Microphone noise is given as the equivalent acoustic sound pressure level which would correspond to that output. So the decibel reference in that context is the ratio of sound pressure level to 20uPa pressure change, then converted to electrical output based on the microphone sensitivity.
The AKG page for the C5 gives a sensitivity of 4mV per pascal (so 4mV output for 94dB acoustic). 25 dB acoustic is 69 dB less that 94 dB acoustic (notice that AKG spec the mic as 69 dB SNR), so the microphone noise output would be around 1.4uV (someone double check my math).
1.4uV would be -117 dBV or -115 dBu (both of those are rounded values).

That is the minimum electrical noise you could get from that microphone, but as others pointed out, when you connect a microphone to a preamp you also measure all the amplified acoustic noise in your room, so not useful for direct comparison.

And no, it also isn't -116dB + -115 dB, the way you have to add noise values is complicated by the randomness and by the fact that decibel values are logarithms. You add decibels if there is analog gain, so if you have a -116 dB noise level and that goes to an amplifier stage with 40dB gain you would get -116 + 40 = -76dB noise at the output.

It is often said that the last quarter turn is the noisiest

Are you familiar with the term "tautology?"
The highest gain is of course going to be the highest noise, that is not really useful or surprising information. The noise of the 200 Ohm resistor sets the noise floor, and the microphone preamp amplifies that noise, and adds additional noise as well. The amount of additional noise at each gain setting lets you evaluate how low noise the preamp design is.

 

[transient]

What Matt said, plus you should read the Equivalent Input Noise section in Bob Metzler's Audio Measurement Handbook. The AMH has been mentioned in the thread on Audio Reference Books, if I recall correctly. Ideally, read the entire volume for a good general background on audio measurement techniques.

 

[user 37518]

As others have implied / stated - you should not be introducing a real transducer into the test setup for fairly obvious reasons of ambient noise. For accuracy you should put a resistance (150 / 200 Ohm depending on what you are using as a comparison reference - value tends to differ between North America and UK for example) across the input. If you simply short circuit this then you are not taking into account the input current noise of the preamp that develops a noise voltage across the transducer impedance.

All true, yet you can short the input to gnd and still get the contribution of the noise current through the pre amp's own input resistance. But I agree with you, is best to include the aource resistance

 

[user 37518]

A bit of thread derail: I own several distortion/noise measuring equipment. I have an AP SYS-1, a QuantAsylum QA402, an HP 8903A and even an old HP 334A. And to be honest, every time I get more and more convinced that there is nothing better than MATLAB/Octave + a Focusrite Scarlett or similar.
I can only think of one good reason to buy an AP and that is due to its ultra low residual distortion and noise, that's it. AP is even selling the software alone for 3K, and I think that anyone with a bit of knowledge of DSP would do infinitely better with MATLAB/Octave. It is a fact that sometimes getting the audio interface up and working with Matlab is not easy, but once you do, there is nothing better.

 

[kags]

Great info on noise measuring! Another consideration based on your in-use experience that after a certain gain the signal get “nasty.” Are you feeding the output of this mic preamp into an interface with a preamp? If so, you need to bypass or at least zero out the gain of the interface preamp or you are injecting a line level signal (from your 500-series preamp) into your interface, which may be expecting a mic-level signal.

Im all for measuring the noise, but make sure your gain structure is optimized

 

[JAY X]

Hi!

Really I appreciate your advise!. Not that "easy" to make good measurements, but at least now I know better the process. I will try matlab/octave software. In any case, it is clear i need to read more, and a scope...i found some 1 or 2gb/sec models.

When recording in my Daw, i tend to have the peak signal reading around -12dbfs. My interface has preamps, but i set them at minimum gain.