EEVblog #611 - Electret Microphone Design

[DrGeoff]

In my experience, all my quality microphones need the 48v phantom voltage from my mixer.

That has not really covered by Doug in this series so far.

48V phantom supply is a real misnomer. It's not really a voltage source.
The 48VDC is the open circuit voltage, however this is in series with a 3K4 resistance in the console, with a short circuit current of about 14mA.

Many high-end mics (Schoeps, DPA, Neumann etc) have their own phantom power supplies, that supply whatever is required by the mic to operate (essential for tube mics).

[Richard Crowley]

P48 phantom power is well defined in DIN 45596, followed by IEC 61938.
It is indeed a 48V voltage source with 6.8K matched (within 0.4%). Provisions are made for 24V and 12V options with lower-value resistors.

Pro, externally-biased condenser mics use their own custom, mains-supplied, external power supply boxes, they don't operate on P48 phantom power.
For example, the Rode Classic II that was mentioned in the video...

[townkat]

hi everyone,
i have some unclear things about this subject which i'll try to expose here,

i have a Samson c01u usb electret wich i am analyzing for, maybe, improving the noise floor
i noticed that it does have the (50)megohm resistor connected between fet gate and a voltage divider mid point (unlike in the video where it is connected to ground?..), so my understanding is that the megohm resistor purpose is to limit the current from the voltage divider matching it to the current from the electret capsule. is my understanding corect? (is this called impedance matching?)

and the 2nd,
i have seen on internet that some electret pre-amp schematics also have an ac coupling capacitor between the electret capsule and fet gate, what can you say about this? (does it have any effect, is it useless, can you also explain why etc..)

these are probably beginner questions but i tough they are still more suitable asked here than in a new topic in the beginner forum

thank you

[Richard Crowley]

It is almost impossible to have this discussion without referring to a circuit diagram somewhere. Do you have a schematic diagram you are looking at?

50M ohms seems quite low for an electret condenser capsule circuit. If it is connected over to a voltage divider it seems likely it is biasing the FET at midway between ground and the supply voltage. There isn't any current here to "limit".  A very high value resistor is used to minimize draining away the audio signal.

The electret capsule output is very high "output impedance" which is why the FET is used because the FET has very high "input impedance". The FET is doing the impedance conversion with a very high input impedance and a relative low output impedance.

There are so many different circuits for this, it is impossible to make generalizations about all of them together.  In the circuits I've seen, the series capacitor between the mic capsule and the FET input was used for high-pass (low-cut) filtering and/or as an attenuator (to allow the mic to be used in high loudness situations).

Improving the SNR of that mic is an interesting project and it will be particularly challenging where you are starting from.  Because that mic has both a conventional electret mic circuit plus the USB ADC and interface, etc. it will be somewhat different than most of the condenser mic circuits you have seen. Conventional mics are designed to differentially (balanced) drive a long, low-impedance mic cable. But since the ADC is right there in the mic body, they have a rather different design parameter.  We don't really know how much preamp they used to get a signal level (also unknown) appropriate for their ADC. Again, a circuit diagram would practically be required to work on that mic.  Most condenser mics are simple enough that you can draw out the diagram yourself. But that would be almost impossible for a USB condenser mic like that.

[townkat]

someone made some good pictures of the boards of c01u, (and some explanation)
http://jumperone.com/2011/10/samson-c01u-teardown/
is this enough or a diagram would help further? i can also take more pictures if needed

the preamp is built into the adc chip (24+24db programmable gain amplifier)
all the noise is coming from the small board as i bypassed the megohm resistor with a small cap and the noise was gone (but also the signal)
what i am not sure yet is if the noise source is the 50megohm resistor or the capsule itself

[Richard Crowley]

someone made some good pictures of the boards of c01u, (and some explanation)
http://jumperone.com/2011/10/samson-c01u-teardown/
is this enough or a diagram would help further? i can also take more pictures if needed

We can't really have any useful discussions about the circuit design or operation without looking at the circuit diagram.
Randomly poking around blink is not only a waste of time, but very likely to be fatal to the circuit if you poke the wrong place.

the preamp is built into the adc chip (24+24db programmable gain amplifier)

That severely limits what you can do to quiet the audio path.  This kind of "all in one" microphone was not designed for high performance. It was designed for convenient operation.

all the noise is coming from the small board as i bypassed the megohm resistor with a small cap and the noise was gone (but also the signal)

Which is exactly why it is such a large value resistor, to avoid draining off the very miniscule audio signal.

Putting almost ANY value capacitor in that spot is guaranteed to short out ALL the audio signal and leave you with nothing.

what i am not sure yet is if the noise source is the 50megohm resistor or the capsule itself,

It is probably a safe differential test to simply disconnect the capsule and see how the circuit behaves by itself.
Then you will hear only the circuit noise, and nothing from the mic capsule.

VERY IMPORTANT NOTE!!!:  That "solder blob" where the  capsule connects is the VERY MOST SENSITIVE part of the whole circuit.
The reason it is floating like that is because the most minuscule, invisible substance could cause major problems and quite possibly render the circuit non-functional.
I can't stress how important it is to treat that node with extraordinary care and make sure it is scrupulously clean.
Even left over solder rosin residue that is completely invisible under a microscope can severely damage the audio signal.
We use special no-residue solvents and methods to clean that extremely high impedance node to eliminate any trace of leakage.
Messing with that node at all is a very bad idea and to be avoided if at all possible.

[townkat]

some more pictures and articles on this subject here
http://www.noisyaudio.com/tag/samson-c01u/
the guy there suspect the a/d chip for the noise but the experiment i made with the megohm resistor ac shorted seem to indicate the resistor or the capsule..

i attached a schematic i made of the small board

the text on fet was not fully readable, what i could read seem to be k.. j? (it may be K596), then on second line j83m

i could not find a microphone symbol so i used a led one

[Richard Crowley]

The transistor is a 2SK596, a special FET designed for electret condenser microphones.   http://www.onsemi.com/PowerSolutions/product.do?id=2SK596S

As we suspected, the 50M resistor is biasing the FET from the voltage divider formed by the 39K and 40K resistors.
The resistors are not equal because note that the transistor actually has a resistor inside.

The node where the 50M resistor, the business-end of the capsule, and the FET gate is the super-high-impedance node.
You won't be able to measure any voltage there because it is extraordinarily high impedance and your meter loads it down too much.

There should be a wire going from the back of the mic capsule to the high-impedance node. 
What happens if you disconnect that wire at the back of the capsule?  (Avoid touching anything to that high-impedance node!!!!)
Or if you connect your capacitor across where the capsule would be?
That would show you how much noise is coming from the circuit (FET + preamp/ADC/USB).

[townkat]

test:
computer gain was set to maximum (teoreticaly adc preamp is at +48dB)

original,
noise floor -49dB

wire from capsule disconnected
huge buzz -11dB
(if i was touching the mic case the signal was lowering to -42dB)

wire from capsule disconnected and 0.1uF (104) capacitor connected between the disconnected wire and capsule ground (the other wire still conected to capsule)
noise floor -56.5dB

all readings where made with microphone assembled to benefit from the case shielding

[Richard Crowley]

How much of the "original noise floor -49dB" was ACOUSTIC noise from the environment? 
It isn't fair to ascribe that to the microphone when it is only doing its duty by picking up ambient noise.

If you don't have an anechoic chamber handy, you could do the next best thing (for home experiments).
You could take all the bedding in the house and pile it on one bed to block as much ambient noise as possible.
Then you could wait until 3am on the most quiet night of the week in your neighborhood.
Then bury the microphone inside the pile of pillows, blankets, etc. to do the experiment.

So the difference between  -49dB and  -56.5dB is (roughly) the contribution of the mic capsule.

Some people claim the capsule is a TransSound TSB-160A  and it looks very similar from the photos.
You could buy a pile of them and select one with the lowest noise.  This appears to be one source....
http://www.ebay.co.uk/itm/Trans-Sound-Unidirectional-TSB-160A-Unidirectional-mic-capsule-Quantity-of-1-/221541095942
Of course for the same $$$ you could have simply bought a quieter microphone and let the manufacturer sort the capsules for you.

But I would agree with some of the comments that the most likely source of the mediocre performance is that AK5371 chip.
It is rather pushing it to even use such a low-end chip in that kind of microphone product.
This is clearly a low-end circuit design and practically impossible to "improve" without simply replacing the entire circuit with a better design.

[townkat]

the mic was wrapped in a blanket and between pillows in the first test were the capsule could pick up ambient noise, also the room was quiet at the moment of measurement with the mic conected to a laptop as the desktop pc was raising the noise with 3dB, the blanket and pillows reduced the noise with about 13-15dB

the normal gain needed is about 30dB
the noise could also be reduced by 1-2dB if the official app is used which commands the amplifiers separately (a little bit of gain staging)

i find it quite usable anyway just that i was wondering if there are any improvements possible
(i wouldn't change the capsule with one of the same model)

thank you very much for your help 

[JustSquareEnough]

I posted this in the youtube comments but figured I'd have a better chance here:

for the life of me I dont understand why gain goes down as the external resistor goes down which has the effect of increasing Vds, correct?

I've simulated as much and can see that the amplitude of the output wave does in fact go down as the external resistor gets smaller (1k vs 5k for example) but I don't understand why.

anyone?

[JustSquareEnough]

I posted this in the youtube comments but figured I'd have a better chance here:

for the life of me I dont understand why gain goes down as the external resistor goes down which has the effect of increasing Vds, correct?

I've simulated as much and can see that the amplitude of the output wave does in fact go down as the external resistor gets smaller (1k vs 5k for example) but I don't understand why.

anyone?

researching my own question, common source amplifier configuration (https://en.wikipedia.org/wiki/Common_source) we see from the voltage gain equations, Rd being in the numerator now makes sense mathematically as to why as Rd gets small gain goes down.  However, it still doesn't make sense intuitively, to me yet, why when Ids is fixed and Vds is increasing due to Rds decreasing gain goes down?

[Richard Crowley]

No transistor (inclucing the one inside the electret capsule) is perfect.  They all have some absolute limit of how much current they will pass, AND they all have an intrinsic series resistance which limits how much current we can see externally.

But then the capsule isn't supposed to be a power or current device.  It is supposed to be a signal or voltage device.  If you need more current use a buffer stage. That is how most electret condenser mics are designed.  For example here is a circuit using a typical electret capsule, and then a pair of resistors transistors as an impedance buffer to drive the low-impedance balanced typical microphone circuit: