Fixed gain field recorder?

TNT

Member
Joined 2003
Paid Member
Do you think we have come to a state where we could build a field recorder without a gain knob. I am pondering if we have the noise levels and dynamic headroom in both analog and digital domains that we could just set a gain and live with it for any recording really.

Prerequisite is perhaps mic sensitivity and noise figures... I'm thinking really serious state of the art recording like organ in church, closed mic guitar amplified small jazz club etc. So noise level in the chain should just beat any normal recording space by a safe margin (3-6 dB?) but i.e. need not meet the requirement to record mosquito farts in outer space from 100 meters.

Is it feasible? What would it take do you believe? Do the gear exist today. DIY of course ;-)

It would be so nice not have to consider setting the level knobs position. Are they equal? Will it clip? Get the potentiometer out of the signal path... etc... just set up the mics, power on the gear and hit Record.

My hope is that now we have such low noise mic amps and high dynamic/resolution ADC that we can just create a recoding unit and it will take on anything really. Maybe a high and low switch somewhere 🙂 that shifts the whole thing 18 dB in the digital domain?

Could someone elaborate on the gain and noise structure?

Mic in -> 24/96. (44,1)

What say thee?

//
 
My AKG C900 microphones have 122 dB(A) dynamic range (having 17 dB(A) noise and 139 dB maximum sound pressure level), so using an ADC that has a couple of dB more than that, say 128 dB(A), you would never need a gain knob anymore when you only use C900 microphones. I think audio ADCs with such a dynamic range exist nowadays.

If your field memory recorder also has to work with dynamic microphones as well as large-diaphragm studio microphones, then you have the 20 dB or so difference in sensitivity between those to consider. Besides, dynamic microphones can handle much higher levels than 139 dB SPL, which is why they are often used for closed-miked kick drums (or so I read somewhere, no idea how reliable this information is). Even if 139 dB SPL is enough, the sensitivity difference would increase the dynamic range requirement to 148 dB(A). I don't think audio ADCs with such a dynamic range exist yet.
 
These seem to be the numbers....

0 dBu at 133 dB -> 2,2 V p-p ?

//
 

Attachments

  • mic dyn volt.jpg
    mic dyn volt.jpg
    193.2 KB · Views: 377
Last edited:
Yes, at its output, your microphone has a noise floor of about 1.26 uV RMS A-weighted and a maximum level of about 0.72 V RMS assuming a single sine wave, so indeed just over 2 V peak-peak. If you don't want to lose more than 1 dB(A) of signal to noise ratio, the noise of the recorder has to be about half the microphone's noise, so the recorder specs will be:

Input referred noise at < 100 ohm source impedance: 630 nV A-weighted
Maximum input signal level: 0.72 V RMS for a sine wave, so about 2.1 V peak-peak
Dynamic range: 121 dB(A)
 
Interesting concept and seems almost feasible. Since whatever mic you choose isn't going to change (it's fixed gain) it really comes down to the mic preamp built into your recorder.
You don't want to clip the ADC, but you'd want to get pretty close to clipping for good S/N. Are the preamp and ADC clean enough to use when your average recorded signal is -60dBFS? That should not be too hard to test.
 
Looking at the Mouser website, the TI PCM4222 (124 dB(A)) and PCM4220 (123 dB(A)) meet the dynamic range target and the Cirrus CS5381 (120 dB(A)) comes quite close. All of them have a maximum input level around 5.6 V peak differential, so in all cases you would need a low-noise differential amplifier with a fixed gain of about 2.5 between the microphone and the ADC if you want to align the maximum levels.

In fact, if you are willing to accept 3 dB(A) loss of signal to noise ratio instead of 1 dB(A), the microphone could directly drive the PCM4222 without any gain stage. (That is, you would still need some coupling capacitors, resistors and clamping diodes to get to the correct common-mode voltage, but no amplifier.) I doubt if that is a good idea, though, as the output impedance of the microphone could affect the settling of the switched-capacitor input of the ADC.
 
Last edited:
Thanks!!

Yes, maybe a buffer should be there - even an 1:1 just for ADC driving aspects. Maybe it should be 1:2.... But I like the concept of a straight line into the ADC. This buffer stage should have phantom feed in it. I would like to build this with battery feed; phantom - buffer - ADC - USB - laptop (an SD card even better). This aim to be a puristic, possible to carry, field recording solution.

I would probably just set the ADC to 24/88,2 and leave it there - I want an analog pre filter to set in at say 24 kHz and the combo of analog input and digital filters should yield better than -150 dB at 44,0 Khz. Die alias, die 🙂

If anyone has an idea of such a pre-stage (phantom+buffer) - please describe... 🙂 If without capacitors in signal path - great if possible...

//
 
....You don't want to clip the ADC, but you'd want to get pretty close to clipping for good S/N. Are the preamp and ADC clean enough to use when your average recorded signal is -60dBFS? That should not be too hard to test.

Yes! I have nited my mic gain setting on my Edirol and it actually comes up quite close in quite different situation. Organ i church, trad jazz band with mic just 1 or 2 meter away - partly amplified song/band...

So I would guess that maybe one would dare to up your -60 dBFS to maybe -40?

I think one should include an estimation of what is the absolute quietest background one might experience... add a few dB (3-6) margin downwards and then look upwards.. 🙂

//
 
Thanks!!

Yes, maybe a buffer should be there - even an 1:1 just for ADC driving aspects. Maybe it should be 1:2.... But I like the concept of a straight line into the ADC. This buffer stage should have phantom feed in it. I would like to build this with battery feed; phantom - buffer - ADC - USB - laptop (an SD card even better). This aim to be a puristic, possible to carry, field recording solution.

I would probably just set the ADC to 24/88,2 and leave it there - I want an analog pre filter to set in at say 24 kHz and the combo of analog input and digital filters should yield better than -150 dB at 44,0 Khz. Die alias, die 🙂

If anyone has an idea of such a pre-stage (phantom+buffer) - please describe... 🙂 If without capacitors in signal path - great if possible...

//

The digital decimation filter inside the PCM4222 only has a bit over 100 dB stopband rejection, and actually only 6 dB at the Nyquist frequency - the stopband starts at 0.5465 fs instead of 0.5 fs. You would need something like a ninth-order Butterworth filter to get the rejection up to 150 dB from 0.5465 fs onwards, or 25th order from 0.5 fs onwards. With elliptical filters, the order could be much reduced, but you would still need a pretty high-order filter with high dynamic range requirements.

An alternative would be to take the multibit sigma-delta modulate from the PCM4222 and pass it through your own decimation filter, for example made with an FPGA. The analogue chain can then be kept nice and simple.

https://www.ti.com/general/docs/sup...=26&gotoUrl=http://www.ti.com/lit/gpn/pcm4222
 
Hi!

I have a hunch that "digitalis" is due to poor filtering. In both ends really. To be what it promises, anything above Fs/2 need to be kept OUT - both at ADC and DAC. Your last statement is what I think I would like to do.

//
 
So I would guess that maybe one would dare to up your -60 dBFS to maybe -40?
If so that would be great and you shouldn't have much trouble. I like to record hot because that's what we learned to do in the days of tape, but it may not lbe needed these days. In your live recordings, if RMS is at about -40dB, where are your peaks?

I looked at those Swedish mics, had not seen them before. For an 8mm capsule that omni is very flat. Do they build in EQ? Back in the '80s Jean Hiraga said that he was going to have a pair of high end mics made in Sweden for his personal use. I wonder if it was very early Line?
 
SAR ADC for high performance audio ADC project [LTC2380-24]

Glancing this for a long time - I had hoped this had materialised sooner but cant of course blame the good designer.

//

Edir : Or here atually:OSVA - Open source Versatile Analyzer

So additional options for the ADC are AKM's AK5394A (123 dB(A)) and AK5397 (127 dB(A)) and the LTC2380-24 (unknown but apparently close to the AK5394A).

The AKM chips also use halfband filters that don't completely prevent aliasing, and the LTC is a SAR ADC with built-in sample-and-hold that doesn't have a decimation chain at all, only a digital averaging filter.
 
Last edited:
Hi!

I have a hunch that "digitalis" is due to poor filtering. In both ends really. To be what it promises, anything above Fs/2 need to be kept OUT - both at ADC and DAC. Your last statement is what I think I would like to do.

//

In case you use a sigma-delta ADC, it may be useful to know that they are sensitive to interference on their reference or clock at all frequencies except the sigma-delta modulator sample rate (typically 128 times 44.1 kHz or 128 times 48 kHz, but check the datasheet of the chip you want to use) and its exact integer multiples. It is therefore handy to let the clock of the FPGA's DSP blocks run at an integer multiple of the sigma-delta modulator sample rate.

The same holds for the boost converter for the phantom supply, if any, but unfortunately 128 times 44.1 kHz is a bit high for a boost converter. Very often frequencies close to odd multiples of half the sigma-delta sample rate cause most interference, so you can at least try to keep boost converter harmonics away from that.
 
Yes, maybe -40 is a bit optimistic and I do want the peaks intact... I dont know for how long Line Audio has been running... seems at least from 2005...

I believe that the project by Frex do contain some logic and SW to accommodate custom filtering but must confess that I have lost track of that project a bit. I do live in the state that it shall be possible.

I have been looking at a lot of phantom power circuits and must say that I have not found one that I was really hocked on... if one could get those caps out of the way.... 🙂

A low noise 2x gain line stage should not be to hard to find I suppose...


//
 
I think I will ask Line Audio to build me a pair of mics with discrete wiring for the power feed i.e. use a 5 pin XLR. Then I can get rid of the phantom feed drawbacks in the reception circuit. This is all purpose built anyway... 🙂

Or mod my pair... 😉

//
 
Last edited:
I did a few op-amp unity gain stages for AKM ADC chips, but I can't for the life of me remember what I did for a low pass filter. :xeye: Whatever it was worked and was straightforward. Should not be too hard to come up with something of your own.

Knowing what is the dynamic range of your recordings would be a big help in setting gain. The wave editor Goldwave is very handy for finding out things like that. I suspect that if you normalize the peaks of a raw recording to 0dBFS that your RMS level will be around -40.

The Line Audio website says "Since 1989" which is why I wondered. I am now tempted by a pair of their cardioid mics. 🙂 You like the Omni mic, right?
 
1989! didn't find that 🙂

Yes, I think they seem to sound very good.

Using 200ms window analysis for whole "song":

Church organ recordings showed these figures:

Minimum sample value -0.7198 -0.8257
Maximum sample value 0.7291 0.7713
Peak amplitude -2.7 dB -1.7 dB
DC offset -0.0000 -0.0000
Minimum RMS power -62.2 dB -62.9 dB
Average RMS power -22.9 dB -22.8 dB
Maximum RMS power -12.3 dB -11.7 dB
Clipped samples 0 0

One more from that session:

Minimum sample value -0.1777 -0.1740
Maximum sample value 0.1978 0.1699
Peak amplitude -14.1 dB -15.2 dB
DC offset -0.0000 -0.0000
Minimum RMS power -49.8 dB -50.1 dB
Average RMS power -29.3 dB -29.4 dB
Maximum RMS power -23.9 dB -24.7 dB
Clipped samples 0 0

Here I would shift out 2 bits to the left for 12 dB gain and probably add dither.


Take of a street performer playing mandolin in a tunnel:

Minimum sample value -0.5357 -0.6282
Maximum sample value 0.6140 0.7078
Peak amplitude -4.2 dB -3.0 dB
DC offset -0.0000 -0.0000
Minimum RMS power -28.8 dB -30.6 dB
Average RMS power -22.4 dB -23.4 dB
Maximum RMS power -14.1 dB -11.5 dB
Clipped samples 0 0


Marching band from 10 meters:

Minimum sample value -0.4579 -0.3540
Maximum sample value 0.2883 0.2314
Peak amplitude -6.8 dB -9.0 dB
DC offset -0.0000 -0.0000
Minimum RMS power -54.3 dB -55.9 dB
Average RMS power -31.6 dB -33.0 dB
Maximum RMS power -20.7 dB -21.8 dB
Clipped samples 0 0


Trad jazz band 2 meter:

Minimum sample value -0.2131 -0.2830
Maximum sample value 0.5640 0.4955
Peak amplitude -5.0 dB -6.1 dB
DC offset -0.0000 -0.0000
Minimum RMS power -59.7 dB -59.4 dB
Average RMS power -32.0 dB -31.3 dB
Maximum RMS power -21.0 dB -20.8 dB
Clipped samples 0 0

As I remember it the gain knobs where about in the middle on all these 🙂

Short excerpts below in FLAC (16/44):

//
 

Attachments

Last edited:
Knowing what is the dynamic range of your recordings would be a big help in setting gain.

When the dynamic range of the recorder exceeds the dynamic range of the microphone, you can set the gain so high that the noise floor is dominated by the microphone's noise, yet so low that the recorder's clipping level exceeds the microphone's clipping level. Why would the dynamic range of the recordings matter then? Whatever the microphone can handle, the microphone + recorder can handle without any need to change the gain.