• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Microphone preamp measurements: How to get the signal level right

OP
R

Rja4000

Addicted to Fun and Learning
Joined
May 31, 2019
Messages
961
Likes
818
Location
Liège, Belgium
Thread Starter #21
Nobody doing real mixing on a digital mixer will go anywhere near 0dBFS, at least not intentionally. The usual conventions dictate that there should be at least 10dB between the very highest levels likely to be encountered and 0dBFS. 20dB would be a better safety margin to aim for so I would be inclined to take whatever margin you decide on, into consideration when setting up your operating conditions for measurement.
By the way, measuring this on the Focusrite Liquid 4Pre, Yamaha DM1000 mixer and on the Yamaha AD8HR Preamp, it seems Yamaha is labelling the "gain" with a number being the level in dBu to reach exactly -20dBFS,
while Focusrite use positive numbers, representing the negative of the level in dBV to reach exactly -20dBFS :)
Or, in other words, the negative of the level in dBu to reach -22dBFS

Example:
"-60" position Gain for Yamaha => if the input is -60dBU, the level reads -20dBFS after ADC conversion
"60" position Gain for Focusrite Liquid 4Pre => if the input is -60dBV (-58dBu), the level reads -20dBFS after ADC conversion

The bottom line is that as the Focusrite Liquid 4Pre has gain up to "80" and the Yamaha AD8HR up to "-62",
the 4Pre has an additional 16dB of gain (!)
 
Last edited:

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
9,649
Likes
12,894
#22
Sure.
With 24 bits digital and current ADC resolving 18 to 20 bits without issue, -20dBFS should be used as a target.
(Or anywhere between -10dBFS and -20dBFS, depending on the recorded signal expected level variations.)

But that doesn't change the question.

We need to compare the capability of the preamp+ADC as a whole.
For that, we need to know what's the "gain" they allow.
And the only way I could think about to achieve such a comparison is to check what's the dBFS level at maximum gain for, say, a -60dBV (1mV rms) signal.

Of course, that doesn't give us the analog stage gain, because we don't know the dBV/dBFS correspondance after the analog stage, at the input of the ADC.

But that's a figure we can compare.

And if we want to compare to an analog preamp, we have to take the analog preamp headroom into account (which is quite high for the Mackie, as an example).
Because if we add an ADC like the RME behind the Mackie, that's what we will do: align 0dBFS with the max level before unacceptable distortion of the preamp analog stage, as far as possible.

So I'm measuring that.
On top of noise measurement.
(Which may give us a clue of what the actual analog gain is)
Seems to me max input at max gain or max input at zero gain answers your questions comparing one unit to another.

You could use output from a set input like you suggest, but you won't find that quoted in specs. One or both of the other specs are typically quoted for mic preamps.
 

AnalogSteph

Addicted to Fun and Learning
Joined
Nov 6, 2018
Messages
591
Likes
409
Location
.de
#23
I have done some poking around in 402 VLZ4 mixer performance, and my impression was that the tone stage following the mic pre is holding its distortion performance back even at max gain - I got -95 dB H2 / -100 dB H3 @ 1 kHz. (Estimated EIN shorted was about -131.5 dBu, which should translate to about -128 dBu with 150 ohms.) A Behringer Q1002USB preamp (presumably the usual composite with CFP inputs) is up to -75 dB H3 at full tilt, and clearly due to running out of loop gain.

Quite honestly, I would not expect distortion performance or noise to get a great deal better than what the trusty Mackie pre already delivers. Some testing at moderate gain (10-20-30 dB) should be carried out to establish its maximum dynamic range, where I imagine some improvements may still be to be had. The current Sound Devices MixPre II models are claiming 142 dB making it into the ADC at +10 dB input gain, which is close to the limit of what's physically possible and presumably requires a different topology than the usual single-gain-knob instrument amplifier. Expect something like 118 - 128 dB (usually low 120s) for those.
 

dfuller

Active Member
Joined
Apr 26, 2020
Messages
175
Likes
128
#25
In case you're wondering what makes the Millennia different, it's got a very good step response (far better than most mic pres out there). They're "quick" sounding, lots of transient.

I may check this out with some preamps I own...
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
9,649
Likes
12,894
#26
I have done some poking around in 402 VLZ4 mixer performance, and my impression was that the tone stage following the mic pre is holding its distortion performance back even at max gain - I got -95 dB H2 / -100 dB H3 @ 1 kHz. (Estimated EIN shorted was about -131.5 dBu, which should translate to about -128 dBu with 150 ohms.) A Behringer Q1002USB preamp (presumably the usual composite with CFP inputs) is up to -75 dB H3 at full tilt, and clearly due to running out of loop gain.

Quite honestly, I would not expect distortion performance or noise to get a great deal better than what the trusty Mackie pre already delivers. Some testing at moderate gain (10-20-30 dB) should be carried out to establish its maximum dynamic range, where I imagine some improvements may still be to be had. The current Sound Devices MixPre II models are claiming 142 dB making it into the ADC at +10 dB input gain, which is close to the limit of what's physically possible and presumably requires a different topology than the usual single-gain-knob instrument amplifier. Expect something like 118 - 128 dB (usually low 120s) for those.
In crude terms that don't correctly explain the how, Sound Devices are running two parallel ADCs and combining the result for extended dynamic range. It allows you to record only the loud parts with say a 100 db wide window, and then record only the low level parts with a 100 db window which is much lower in level, and combine them for more than a 100 db wide window to work within. Then record the result as 32 bit float.
 
Last edited:
OP
R

Rja4000

Addicted to Fun and Learning
Joined
May 31, 2019
Messages
961
Likes
818
Location
Liège, Belgium
Thread Starter #27
In case you're wondering what makes the Millennia different, it's got a very good step response (far better than most mic pres out there). They're "quick" sounding, lots of transient.

I may check this out with some preamps I own...
So, that's a bandwidth difference?
 
OP
R

Rja4000

Addicted to Fun and Learning
Joined
May 31, 2019
Messages
961
Likes
818
Location
Liège, Belgium
Thread Starter #28
Impressively good for a unit costing about €350. The ‘rule of thumb’ for mic. amps. was, ISTR, that -130 was “as good as it gets”.
I was unsure of the method, to be honnest.
Now, I'm more confident.
Results are coming soon :)
 
OP
R

Rja4000

Addicted to Fun and Learning
Joined
May 31, 2019
Messages
961
Likes
818
Location
Liège, Belgium
Thread Starter #29
In crude terms that don't correctly explain the how, Sound Devices are running two parallel ADCs and combining the result for extended dynamic range. It allows you to record only the loud parts with say a 100 db wide window, and then record only the low level parts with a 100 db window which is much lower in level, and combine them for more than a 100 db wide window to work within.
Yes, that's interesting indeed.
But the purpose is more "in the field recording", where you control nothing about levrl or dynamic range.
Switching gain (/ADC) could lead to noise pumping and other artifacts...
Not sure how this would help a normal preamp performance.
 

AnalogSteph

Addicted to Fun and Learning
Joined
Nov 6, 2018
Messages
591
Likes
409
Location
.de
#30
I've been wondering how they may be combining the streams from both ADCs; my guess would be that they're keeping track of relative gain so that both waveforms can be combined seamlessly at a predefined level. (Or maybe take the first bunch of bits per sample from one converter and the rest from the other? Both approaches potentially have their quirks.) Cameras are doing so much HDR stuff these days, and I imagine that the odd idea from that camp would also apply to audio. Guess I should look up the patents given in the user manual, huh?
*patents #US9654134B2 (USA), #CA2973142C (Canada), and #EP3259845 (Europe)
If this concept works well, it could drastically reduce ADC power consumption. High dynamic range converters are rather power-hungry beasts by nature (dynamic range is, after all, still the ratio of signal power to noise power, and the latter is thermally limited), and if you could use two delivering 20 dB less each you'd probably still be looking at substantial power savings.

Getting 142 dB(A) through any analog preamp is quite impressive as-is. That's a serious, serious amount of dynamic range. Given that internal levels seem to max out at +20 dBu, that's an input noise of 0.195 µV(A) at 10 dB gain (-132 dBu(A) short-circuited), or in all likelihood less than 1.5 nV/√(Hz) real-life. Good luck finding any suitable balanced line drivers or receivers, too - are there even any with a noise level anywhere near -122 dBu? I'm finding -101 dBu for THAT line drivers, a bit lower for receivers. Guess you'd have to roll your own with discretes...
 

dfuller

Active Member
Joined
Apr 26, 2020
Messages
175
Likes
128
#31
So, that's a bandwidth difference?
Could be; the only paper I found on it referenced slew rate vs transformer coupled mic pres (in this case API 512C and a Neve 1073 clone), with the millennia being considerably faster (~24V/us for the Millennia HV-3D vs ~10V/us for the API and ~8V/us for the Neve clone).
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
9,649
Likes
12,894
#32

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
9,649
Likes
12,894
#33
I've been wondering how they may be combining the streams from both ADCs; my guess would be that they're keeping track of relative gain so that both waveforms can be combined seamlessly at a predefined level. (Or maybe take the first bunch of bits per sample from one converter and the rest from the other? Both approaches potentially have their quirks.) Cameras are doing so much HDR stuff these days, and I imagine that the odd idea from that camp would also apply to audio. Guess I should look up the patents given in the user manual, huh?

If this concept works well, it could drastically reduce ADC power consumption. High dynamic range converters are rather power-hungry beasts by nature (dynamic range is, after all, still the ratio of signal power to noise power, and the latter is thermally limited), and if you could use two delivering 20 dB less each you'd probably still be looking at substantial power savings.

Getting 142 dB(A) through any analog preamp is quite impressive as-is. That's a serious, serious amount of dynamic range. Given that internal levels seem to max out at +20 dBu, that's an input noise of 0.195 µV(A) at 10 dB gain (-132 dBu(A) short-circuited), or in all likelihood less than 1.5 nV/√(Hz) real-life. Good luck finding any suitable balanced line drivers or receivers, too - are there even any with a noise level anywhere near -122 dBu? I'm finding -101 dBu for THAT line drivers, a bit lower for receivers. Guess you'd have to roll your own with discretes...
This is one of those places where dynamic range and signal to noise ratios aren't the same thing and can confuse you. The noise floor is not -142 db from max signal level. But at low signal levels they can get lower noise floors so the dynamic range between max signal and minimum signal can be this wide a range.

This still doesn't let you cheat physics and get noise below around -130 dbv EIN.

https://www.sounddevices.com/how-is-a-32-bit-float-file-recorded/

https://www.sounddevices.com/noise-in-32-bit-float/
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
9,649
Likes
12,894
#34
Yes, that's interesting indeed.
But the purpose is more "in the field recording", where you control nothing about levrl or dynamic range.
Switching gain (/ADC) could lead to noise pumping and other artifacts...
Not sure how this would help a normal preamp performance.
The way they are accomplishing this is multi-stage ADC's working in tandem at all times. So you won't have any artifacts. That by itself doesn't help the preamp, but if you combine it with a good preamp optimized for each of the signal levels you can get wider dynamic range than any single preamp/ADC could manage.

You are correct it is mainly for on site, live or nature recordings where you don't control the levels or can't know what they are ahead of time.

The Zoom F6 also uses dual ADC's and 32 bit float to allow recording without concern for setting levels.
 
Last edited:
OP
R

Rja4000

Addicted to Fun and Learning
Joined
May 31, 2019
Messages
961
Likes
818
Location
Liège, Belgium
Thread Starter #35
You probably know about this, but for anyone wondering, these two short pages explain the noise on preamps. And how to measure them in a way that allows for an apple to apple comparison.

https://www.sounddevices.com/microphone-preamp-noise/

https://benchmarkmedia.com/blogs/application_notes/12139801-measuring-mic-preamp-noise
Thanks. Yes, I do.
I think there are more conditions to be able to compare different preamps noise.

1. Temperature should also be the same. Or compensated for.

2. "Gain" has a different meaning for preamps with integrated ADC, since it also depends on the ADC analog input range (what V for 0dBFS?), and this is not standardised. So can we compare such preamp with fully analog preamps ?

3. Also, is noise level linear with frequency ? Its not. Should we use A-weighting ?

4. How is done (and how efficient) the bandwidth limitation ?

5. Do we measure after FFT ? If yes, does the ADC sampling frequency and the FFT window size make a difference (even small, like 0.1dB) ?
Can we compare figures measured with FFT, in digital domain, to figures measured by analog instrument ?
...
 
OP
R

Rja4000

Addicted to Fun and Learning
Joined
May 31, 2019
Messages
961
Likes
818
Location
Liège, Belgium
Thread Starter #36
Could be; the only paper I found on it referenced slew rate vs transformer coupled mic pres (in this case API 512C and a Neve 1073 clone), with the millennia being considerably faster (~24V/us for the Millennia HV-3D vs ~10V/us for the API and ~8V/us for the Neve clone).
Well. Those can't really be compared to the Millennia to begin with, as they have been designed with some "sound" in mind, while the Millennia is meant for "Wire-with-gain" philosophy.
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
9,649
Likes
12,894
#37
Thanks. Yes, I do.
I think there are more conditions to be able to compare different preamps noise.

1. Temperature should also be the same. Or compensated for.

2. "Gain" has a different meaning for preamps with integrated ADC, since it also depends on the ADC analog input range (what V for 0dBFS?), and this is not standardised. So can we compare such preamp with fully analog preamps ?
Sure, analog designs have max input levels with certain gain settings specified. The extra step of adding an ADC and seeing what input creates 0 dbFS isn't hard to compare at all.
3. Also, is noise level linear with frequency ? Its not. Should we use A-weighting ?
Most specs use A-wtd.
4. How is done (and how efficient) the bandwidth limitation ?
Most of the time the spec is to 20 khz only.
5. Do we measure after FFT ? If yes, does the ADC sampling frequency and the FFT window size make a difference (even small, like 0.1dB) ?
Can we compare figures measured with FFT, in digital domain, to figures measured by analog instrument ?
...
Depends of course on the particulars. Remember there are analog spectrum analyzers or at least were at one time.

There is good info on all this from Rane
https://www.ranecommercial.com/legacy/pdf/ranenotes/Selecting_Mic_Preamps.pdf

The main difference would be once digitzied gain can be applied with no real penalty. If my interface gave me 0 db into an ADC equivalent to 1 volt input, that could be 2 volts or 4 volts or whatever if I play it back on another DAC which has higher output for 0 dbfs. The beauty is this can be done without noise penalty.
 

dfuller

Active Member
Joined
Apr 26, 2020
Messages
175
Likes
128
#38
Well. Those can't really be compared to the Millennia to begin with, as they have been designed with some "sound" in mind, while the Millennia is meant for "Wire-with-gain" philosophy.
Not really, they were designed (at the time) to be as clean as possible. It just so happens that "as clean as possible" in the early 70s is not the same as "as clean as possible" in 2005.
For one, transformers have their own issues with hysteresis reducing transient response - the Millennia is transformerless. API and Neve used transformers at the time because in the early 70s good enough quality solid state differential amps did not exist - at least, not at the noise levels you need for a mic preamp (remember these can add ~60dB or so of gain). They used transformers because they were the best option at the time - plus, transformers have some useful side benefits like blocking DC and acting as impedance matching devices. It helps of course that transistor tech, even discrete transistors, has improved dramatically in that time. Interestingly, the EIN of all 3 is around -130dB.
 

Pluto

Addicted to Fun and Learning
Forum Donor
Joined
Sep 2, 2018
Messages
694
Likes
1,011
Location
Harrow, UK
#39
They used transformers because they were the best option at the time
Old man Rupert claims that transformers are the key to the sound of Neve consoles of that era, and that he would not have used a solid-state approach to a mic. input stage even had a suitably low noise one been available at that time.
 

scott wurcer

Major Contributor
Audio Luminary
Technical Expert
Joined
Apr 24, 2019
Messages
1,072
Likes
1,688
#40
Simply place a low noise* 200Ω resistor from each of pins 2 & 3 of the XLR input down to earth.

*NB this is truly important
If you have resistors that have more than sqrt(4KTR) noise (or less for that matter) with no DC current please sent me some we can share the prize.
 

Similar threads

Top Bottom