Measuring Microphone Preamp Perf in Audio Interfaces

[Blumlein 88]

In french, we say "Un dessin vaut mieux qu'un long discours"

So here we go

I measured my Millennia HV-3C first
This is an analog preamp.
I measured it through the ADC of my RME ADI-2/4 Pro SE.

First for Max gain (60.6dB), with the RME on minimum ADC range (+1dBu)
(Changing the ADC range doesn't change the result here: Preamp's noise is well above the RME noise. What matters is only the Preamp gain.)


View attachment 324244


Then for a gain that gives me just below 0dBFS with a 10mVrms 1kHz signal

I kept the same ADC range first

View attachment 324246

Then I switched to 19dBu range.
To get 0dBFS on the ADC with the same signal, we need to increase the gain by 18dB.
(if I selected the maximum range, which is 24dBu, I couldn't get enough gain to reach 0dBFS with 10mV)

View attachment 324249

As one can see, we get pretty different results.
And we see that for best EIN results, it's better to have more gain and a higher analog output level on the preamp.
(By the way, it's the same if you want to cascade 2 preamps: You'd better maximize the gain on the first stage to get best results)


OK, here it is an analog preamp, and I can play with the ADC range, which allows this flexibility.

Now what happens with digital interface, where the ADC range is fixed ?

Let's measure with the RME 12Mic-D

We don't know the gain, but we may measure the value for 0dBFS in dBu
(Actually, RME is quite clear with the gain: 0dB gain = 18dBu for 0dBFS. Here, the gain is set to 56dB marking, since we target a level of -38dBu for 0dBFS)

View attachment 324250

We can then compare this figure with the previous one:
-127.97dBu un-weigthed vs -127.02 dBu for those 2 high end interfaces/preamps.
(Note that if we had selected a higher level for 0dBFS, like 20mV, as an example, the Millennia would have been able to take benefits of its higher output level - up to 28dBu before saturation - to close the gap with the RME)

Now, at max gain, the RME 12Mic-D gives us this

View attachment 324251

Not dramatically different.

For both examples, the EIN difference between max gain and 10mV range gain is around 0.1dBu.

So, so far, that gives some sense to the idea of measuring all interfaces at max "gain" for noise, if you don't want to go through the pain of explaining this "gain" topic to the readers.
Even if I still think that aligning sensitivities or "gain" is a more scientific way of doing it.

Similarly, measuring dynamic range at lowest "gain" (on mic input, without any pad inserted) probably makes sense.

But we'd still need our "Dashboard" and a set of all usual measurements, to be done at a standardized level AND sensitivity or "gain".
So, say, 4V at -6dBFS (or -3dBFS, if 8V is too big of a range for cheaper interfaces) as an example.



Next, I tried something seriously less good.

Here is the BlackmagicDesign ATEM Extreme ISO video switcher's Microphone input.
"Gain" is set to -1.5dB, for 10mV at -0.2dBFS

View attachment 324484

Interesting to see that the CCIR-2k value here is almost identical to the un-weighted value.
The good thing is that we may now really compare this one with the other ones.
Not really a contest, as expected, of course...
We are virtually at max gain here, so I didn't measure it any further.

So do I understand that last measurement correctly. That 10 mV will clip the input unless gain is less than 0 db?

That is what I thought would be so useful about using 10 mV. We can see what sensitivity is or looking at it backwards what level is max input for the device along with using the same setup to do a useful EIN comparison. Some devices have low max input levels and average gain, others have higher max input levels and average gain. One thing you want to know is will the gain to 0 dbFS work for any given microphone you are wishing to use. I forget, but one interface I saw was optimized for dynamics and ribbons. It had only 50 db of gain, but the ADC sensitivity was such that max input was only 0 dbU for 0 dbFS with 0 db gain. You wouldn't even be able to use a large condenser with it unless you had a PAD added to it.

 

[Rja4000]

do I understand that last measurement correctly. That 10 mV will clip the input unless gain is less than 0 db?

"-1.5 dB" is just the value I had to set for the top (virtual) pot.
It has no specific meaning. It's just graduated in dB.
This is what makes using a "gain" value useless.
Every interface has its own logic.
On the Yamaha, AD8HR you'd see a value of -57dB for the same sensitivity range, as another example.

Also, for the ATEM, what is "Max gain" ?
0dB marking or +6dB marking ?

 

[Rja4000]

In french, we say "Un dessin vaut mieux qu'un long discours"

So here we go

I measured my Millennia HV-3C first
This is an analog preamp.
I measured it through the ADC of my RME ADI-2/4 Pro SE.

First for Max gain (60.6dB), with the RME on minimum ADC range (+1dBu)
(Changing the ADC range doesn't change the result here: Preamp's noise is well above the RME noise. What matters is only the Preamp gain.)


View attachment 324244


Then for a gain that gives me just below 0dBFS with a 10mVrms 1kHz signal

I kept the same ADC range first

View attachment 324246

Then I switched to 19dBu range.
To get 0dBFS on the ADC with the same signal, we need to increase the gain by 18dB.
(if I selected the maximum range, which is 24dBu, I couldn't get enough gain to reach 0dBFS with 10mV)

View attachment 324249

As one can see, we get pretty different results.
And we see that for best EIN results, it's better to have more gain and a higher analog output level on the preamp.
(By the way, it's the same if you want to cascade 2 preamps: You'd better maximize the gain on the first stage to get best results)


OK, here it is an analog preamp, and I can play with the ADC range, which allows this flexibility.

Now what happens with digital interface, where the ADC range is fixed ?

Let's measure with the RME 12Mic-D

We don't know the gain, but we may measure the value for 0dBFS in dBu
(Actually, RME is quite clear with the gain: 0dB gain = 18dBu for 0dBFS. Here, the gain is set to 56dB marking, since we target a level of -38dBu for 0dBFS)

View attachment 324250

We can then compare this figure with the previous one:
-127.97dBu un-weigthed vs -127.02 dBu for those 2 high end interfaces/preamps.
(Note that if we had selected a higher level for 0dBFS, like 20mV, as an example, the Millennia would have been able to take benefits of its higher output level - up to 28dBu before saturation - to close the gap with the RME)

Now, at max gain, the RME 12Mic-D gives us this

View attachment 324251

Not dramatically different.

For both examples, the EIN difference between max gain and 10mV range gain is around 0.1dBu.

So, so far, that gives some sense to the idea of measuring all interfaces at max "gain" for noise, if you don't want to go through the pain of explaining this "gain" topic to the readers.
Even if I still think that aligning sensitivities or "gain" is a more scientific way of doing it.

Similarly, measuring dynamic range at lowest "gain" (on mic input, without any pad inserted) probably makes sense.

But we'd still need our "Dashboard" and a set of all usual measurements, to be done at a standardized level AND sensitivity or "gain".
So, say, 4V at -6dBFS (or -3dBFS, if 8V is too big of a range for cheaper interfaces) as an example.



Next, I tried something seriously less good.

Here is the BlackmagicDesign ATEM Extreme ISO video switcher's Microphone input.
"Gain" is set to -1.5dB, for 10mV at -0.2dBFS

View attachment 324484

Interesting to see that the CCIR-2k value here is almost identical to the un-weighted value.
The good thing is that we may now really compare this one with the other ones.
Not really a contest, as expected, of course...
We are virtually at max gain here, so I didn't measure it any further.

Yamaha DM1000

Gain as close as possible to 10mV for 0dBFS

Max Gain (0dBFS for 7.53mV)

Here, we see a drawback of trying to align gains:
The DM1000 is a digital mixer, but it has analog gain pots, and gain setting is a bit random (it's quite old)
The best I could do was to set 0dBFS for 12.30mV, but that's a 1.8dB lower gain than if I could set it exactly for 10mV.

 

[Rja4000]

This is clearly not with EIN that we'll see huge differences between competent interfaces.

We can still use it to highlight very incompetent devices (like the ATEM here)
This will help readers understanding why (and how much) you should use a proper mic preamp interface for critical quality recording.

For that purpose, measuring at Max gain is good enough, probably.
And at the same time, you will have to measure the minimum level for which you may still reach 0 DBFS for each interface (at max gain), which is a very valuable information



(Table of values for reference)

Device	Gain	UnW	A-W	CCIR2k-W
HV-3C	10mV	- 127,02	- 129,16	- 124,92
HV-3C	Max	- 127,16	- 129,61	- 125,37
DM1000	10mV	- 126,80	- 128,80	- 125,15
DM1000	Max	- 127,37	- 129,37	- 125,70
12Mic	10mV	- 127,97	- 129,97	- 126,29
12Mic	Max	- 128,07	- 130,03	- 126,34
ATEM	10mV	- 107,29	- 110,25	- 107,31

 

[Rewind]

Other than in one occasion, I have not attempted to quantify the performance of microphone preamp in the context of testing audio interfaces (ADC). A good discussion erupted in the Topping E2x2 review thread on this. I like to have a dedicated thread for it so that the topic is better explored.

Note: please consider explaining things more than arguing with each other. You can disagree but let's make sure the task at hand is properly explained with the eventual goal of me adding a test to the reviews.

Thanks,

I just searched for this. It would be great with some reviews on this.

 

[weme]

A few notes on measuring a microphone preamp (DUT)

(1) Frequency response measurements
For example, the typical frequency response of an IMG Stageline MPA-102.



Cause: Use of a special input OP (SSM2019). Electrolytic capacitors were inserted in the gain adjustment circuit to save components (elimination of the offset voltage).

(2) Measurements of the amplifier stages
The basis for accurate measurements is a calibration of the DAC and ADC (ADC: not required for the AP) with the measured values at the input / output of the device under test: Measure with the AP or with a multimeter at ~ 70 Hz, i.e. not at mains frequency, via a cable adapter.

(3) The optimum usable output voltage (Addition to the standard measurements - Could be a new ASR standard)
Requires a multi-tone signal that best matches an average complex music signal. I use the REW signal 1/4 decade at an SR of 48 kHz.
The output voltage with the lowest value of TD+N is determined for this multi-tone signal at a fixed standard gain (microphone preamp: e.g. 40 dB). The input voltage set for the optimum output voltage is now used for further measurements: THD+N, ... (1 kHz), AES17 (41Hz 7.993kHz), CCIF (19+20kHz), ...
It should be possible to automate this process with an AP.

(4) Phantom power supply
Short-circuit current and pairing of the series resistors:
The two short-circuit currents can be measured very easily with a multimeter at the XLR input. The ratio of the two short-circuit currents [%] shows how well the series resistors are paired. A measurement of up to 0.01% should be possible.

(5) Max. Input voltage (1 kHz, dBu) at which an overload is indicated by the DUT.

(6) Measurement units
mV, dBu (dBV)

 

[Rja4000]

Hi everyone

So, after an attempt with the RME UFX III review, here is my proposal:

01. Sensitivity ranges (dBu, V)
A key topic to understand what we speak about, as "gain" has no standard meaning. Let's go back to sensitivity for each gain marking.

02. SINAD at 4V and best level
Without it, readers would be lost on ASR.
More seriously, a good indicator.
4V is hardly a standard for interfaces, but makes sense for comparison, still.

03. THD+N vs level in V, 0dB gain, compared with line
(High levels)
A very important metric: what about minimum gain for Various input types (if any).

04. THD+N Mic vs level (V) Various gains (Max, 10mV, 100mV, Min)
(Low and intermediate levels,up to high)
Gives an idea of the interface behaviour for 4 gains: Max and Min + 2 normalized sensitivities/gain: where full scale is reached for 10mV and 100mV.

05. Dynamic range 0dB and Line
What's the noise at minimum gain, for line level ?
Could be at 4V or full scale (former being more comparable)

06. EIN Various gains (Max, 10mV, 100mV, Min 48kHz, 20Hz-20kHz, 150 ohm 20°C, dBu
AES-17 Dynamic range is valid at high level but impossible to measure accurately for lowest level/high gain.
EIN is an accepted standard measurement.
Conditions have to be comparable, and temperature or resistance difference to standard compensated for (which is easy).

07. Frequency response 0dB gain, line, various gains (Max, 10mV, 100mV, Min)
important to check if that is not way off, specifically at highest gain and special inputs.

08. EiN vs Gain
Interesting to compare interfaces.
X scale has to use sensitivity in dBu or V

09. Linearity (Mic 0dB)
10. IMD (Mic 0dB)
11. Multitone (Mic 0dB)
12. THD vs frequency (Mic 0dB)
Standard ASR measurements

About noise weighting:
I'd keep CCIR-2k for ranking, but specs are usually specified with A weighting. So worth mentioning both values.

About voltage scale:
dBu is widely accepted as a UoM in the Pro world. Standards are about 4dBu, 18dBu, 24dBu,...
So for me it makes sense to use dBu.

Volt may also work.

Here is my 2 €cents.

What do you think ?

 

[signalpath]

Hi everyone

So, after an attempt with the RME UFX III review, here is my proposal:

01. Sensitivity ranges (dBu, V)
A key topic to understand what we speak about, as "gain" has no standard meaning. Let's go back to sensitivity for each gain marking.

Nice work.

Yes, keep dBu. We're no longer working with 600 ohm transmission lines, so dBu is appropriate, and widely embraced.

I prefer unweighted measurements. Been doing this for many decades and have never published a "weighted" spec. Too many games can be played with weighting. I'm on the AES-17 specifications committee and we had this very conversation just a few weeks ago. It was voted to leave weighting as an option. Fine.

Personally, I dislike SINAD in pure audio work (radio, maybe). But that's another conversation. Much better to simply give individual distortion and noise numbers, at varying levels, freqs, sweeps, etc..

 

[Rja4000]

Nice work.

Yes, keep dBu. We're no longer working with 600 ohm transmission lines, so dBu is appropriate, and widely embraced.

I prefer unweighted measurements. Been doing this for many decades and have never published a "weighted" spec. Too many games can be played with weighting. I'm on the AES-17 specifications committee and we had this very conversation just a few weeks ago. It was voted to leave weighting as an option. Fine.

Personally, I dislike SINAD in pure audio work (radio, maybe). But that's another conversation. Much better to simply give individual distortion and noise numbers, at varying levels, freqs, sweeps, etc..

Thanks
I really appreciate the feedback.

Did I miss something important in my measurements, in you opinion ?

 

[AnalogSteph]

Whatever you do, the limitations of basic equipment (think UMC2xxHD, Minifuse, Scarlett 3rd gen or older or 4th gen Solo, maybe even cheaper) better stick out like the proverbial sore thumb. If the treble response takes a nosedive at high gain and/or treble distortion skyrockets, EIN is poor or the ADC has limited dynamic range, that's what we should be able to determine easily.

BTW, having sensitivity graphs go from high levels on the left to low levels on the right feels odd to me. It does mean going from low gain to high gain, so I guess that's where that comes from. Probably a matter of convention / taste and debatable either way.

 

[Blumlein 88]

BTW, having sensitivity graphs go from high levels on the left to low levels on the right feels odd to me. It does mean going from low gain to high gain, so I guess that's where that comes from. Probably a matter of convention / taste and debatable either way.

I agree about the graph feeling odd, but I'll take the info in any form we can get it as long as we get it.

 

[Blumlein 88]

Hi everyone

So, after an attempt with the RME UFX III review, here is my proposal:

01. Sensitivity ranges (dBu, V)
A key topic to understand what we speak about, as "gain" has no standard meaning. Let's go back to sensitivity for each gain marking.

02. SINAD at 4V and best level
Without it, readers would be lost on ASR.
More seriously, a good indicator.
4V is hardly a standard for interfaces, but makes sense for comparison, still.

03. THD+N vs level in V, 0dB gain, compared with line
(High levels)
A very important metric: what about minimum gain for Various input types (if any).

04. THD+N Mic vs level (V) Various gains (Max, 10mV, 100mV, Min)
(Low and intermediate levels,up to high)
Gives an idea of the interface behaviour for 4 gains: Max and Min + 2 normalized sensitivities/gain: where full scale is reached for 10mV and 100mV.

05. Dynamic range 0dB and Line
What's the noise at minimum gain, for line level ?
Could be at 4V or full scale (former being more comparable)

06. EIN Various gains (Max, 10mV, 100mV, Min 48kHz, 20Hz-20kHz, 150 ohm 20°C, dBu
AES-17 Dynamic range is valid at high level but impossible to measure accurately for lowest level/high gain.
EIN is an accepted standard measurement.
Conditions have to be comparable, and temperature or resistance difference to standard compensated for (which is easy).

07. Frequency response 0dB gain, line, various gains (Max, 10mV, 100mV, Min)
important to check if that is not way off, specifically at highest gain and special inputs.

08. EiN vs Gain
Interesting to compare interfaces.
X scale has to use sensitivity in dBu or V

09. Linearity (Mic 0dB)
10. IMD (Mic 0dB)
11. Multitone (Mic 0dB)
12. THD vs frequency (Mic 0dB)
Standard ASR measurements

About noise weighting:
I'd keep CCIR-2k for ranking, but specs are usually specified with A weighting. So worth mentioning both values.

About voltage scale:
dBu is widely accepted as a UoM in the Pro world. Standards are about 4dBu, 18dBu, 24dBu,...
So for me it makes sense to use dBu.

Volt may also work.

Here is my 2 €cents.

What do you think ?

I think this would be enough for most people wanting to compare interfaces. Much more info than you normally find for sure. All makes sense. It will mean quite a bit more testing by Amir. So we need Amir's input on this too. @amirm

 

[Blumlein 88]

Was noticing in your RME review, that at max gain it reached 0 dbFS with only 1.1 mV input. 1.1 mV is the stated sensitivity of the Shure SM7b at 94 db SPL. It is about the least sensitive microphone one runs across. I do hope Amir incorporates at least mic sensitivity and EIN at a couple different levels. These are two of the things one needs to know when choosing interfaces.

I have a Focusrite 18i20 1st gen. 60 db of gain, decent EIN, and it works with ribbons only about 6 db more sensitive than the SM7b. Some of the 3rd and 4th gen devices have lower gain, and not quite as good EIN. But they may have shifted things down so you get similar sensitivity as they still seem okay when Julian Krause lets you hear them connected to the SM7b.

 

[Rja4000]

having sensitivity graphs go from high levels on the left to low levels on the right feels odd to me

Fair point !
I didn't realize that.

Indeed, my idea was to have the graph following the "gain" order.

But I understand we usually have high levels to the right, indeed.

That woul mean all my plots with gain as X coordinate would have to be reversed, though, to keep things coherent.
As an example, EIN vs gain.

I'll give it a try.

 

[signalpath]

Did I miss something important in my measurements, in you opinion ?

Professionals deal with EM/RF interference, so a "common mode rejection" measurement is helpful. In multi-channel micamps, a channel-matching spec couldn't hurt (e.g., +/- 0.08dB worst case channel variance). We usually spec phase response and slew rate, as well as input and output impedance numbers. Nobody in pro audio specs "SINAD" to my knowledge, just individual distortion and noise measurements.

 

[Rja4000]

Thanks

 

[Blumlein 88]

Professionals deal with EM/RF interference, so a "common mode rejection" measurement is helpful. In multi-channel micamps, a channel-matching spec couldn't hurt (e.g., +/- 0.08dB worst case channel variance). We usually spec phase response and slew rate, as well as input and output impedance numbers. Nobody in pro audio specs "SINAD" to my knowledge, just individual distortion and noise measurements.

At least in the consumer and prosumer specs I usually see THD+N either in db or percent which of course is SINAD. I wish they would split out individual noise and distortion numbers. They often do the noise, but not always distortion. Dynamic range is one of the specs lots of manufacturers play games with as well.

 

[Rja4000]

having sensitivity graphs go from high levels on the left to low levels on the right feels odd to me. It does mean going from low gain to high gain, so I guess that's where that comes from. Probably a matter of convention / taste and debatable either way.

Hi

I changed this in the review for outputs.

 

[Rja4000]

Yamaha AD8HR - 8 channel preamp - interface

(NB: This seems to be the exact same preamp/ADC section than in the RIO1608-D and RIO3224-D. This probably makes it the most widely used Pro Micro preamp+ADC in the world - at least for live events. This probably also explains, with the fact its only outputs are AES, why it's also one of the cheapest on second hand market.)

EIN at max gain

NB: I found a trick to compensate for the temperature drift.

 

[Rja4000]

Hi
I explained my method to measure EIN noise in a new thread.

Comments welcome, of course (in that thread)