1" capsules are the microphone of choice for vocals - especially for those sultry female vocals that are so popular in jazz. As far as recording engineers are concerned, a microphone is a musical instrument, not a measuring instrument. Flat response is neither wanted, nor necessary.
By the way, big resistors have been around for a while. I opened the sealed packaging of a 2G resistor a few months ago that had been packed in 1954. The manufacturer's data sheet (unsurprisingly) says you shouldn't touch the glass body with your fingers. 5G is available in SM, but I'm not sure I'd trust it after a PCB has been washed, so I use wire ended.
To tie the two previous comments together, a typical 1" capsule is of the order of 30pF, so to be 1dB down at 20Hz, you need 500M - which is a typical value in Neumann valve microphones. As the capsule capacitance goes down, you need even more resistance, so 5G wouldn't be unusual.
By the way, big resistors have been around for a while. I opened the sealed packaging of a 2G resistor a few months ago that had been packed in 1954. The manufacturer's data sheet (unsurprisingly) says you shouldn't touch the glass body with your fingers. 5G is available in SM, but I'm not sure I'd trust it after a PCB has been washed, so I use wire ended.
To tie the two previous comments together, a typical 1" capsule is of the order of 30pF, so to be 1dB down at 20Hz, you need 500M - which is a typical value in Neumann valve microphones. As the capsule capacitance goes down, you need even more resistance, so 5G wouldn't be unusual.
Good input EC, but the other reason is that the microphone system becomes MORE quiet, if you use a higher value resistor. The is because the inherent resistor noise is more completely shorted out by the mike capacitance, the larger the resistor value. The noise coupling tracks at 6dB/oct for reduction of noise, and 3dB/oct for the increase of the resistor value. Therefore, a larger value resistor has less noise than a lower value resistor in this situation with a cap across it.
This was first pointed out to me by Mead Killion. 'A Low-Noise Two-Wire Condenser Microphone Preamplifier' JAES, April 1967 pp. 163-168.
This was first pointed out to me by Mead Killion. 'A Low-Noise Two-Wire Condenser Microphone Preamplifier' JAES, April 1967 pp. 163-168.
scott wurcer said:
Sony had a really good idea in one of their first FET mics from the '70s. It had a complementary input pair. The gate leakage currents canceled out nearly completely, and they didn't even need a biasing resistor.
Of course the FETs were special ones that Sony developed for that application. I doubt you can get them anymore, except maybe onesie-twosies from Sony spare parts. Still, a good idea.
I'll try and track down the paper. It might have been an AES paper, I can't remember.
OK, I found the AES paper. It was September of 1975. Unfortunately, it doesn't give the part numbers for the transistors nor the model number of the microphone that used the scheme.
It was a big thing, it looked a lot like a C-38, but I don't think it was that because the C-38 was introduced many years earlier. I remember getting the service manual for the microphone and the transistors were about $20 each as replacement parts from Sony -- and this was back then -- probably closer to $50 each now.
I've done a couple of quick searches and can't find the model of the microphone. Maybe a C-500? I can't remember. Maybe someone else can help here. I'm not so interested in the mic as I am the FETs.
It was a big thing, it looked a lot like a C-38, but I don't think it was that because the C-38 was introduced many years earlier. I remember getting the service manual for the microphone and the transistors were about $20 each as replacement parts from Sony -- and this was back then -- probably closer to $50 each now.
I've done a couple of quick searches and can't find the model of the microphone. Maybe a C-500? I can't remember. Maybe someone else can help here. I'm not so interested in the mic as I am the FETs.
john curl said:
Not impractical if you have the right transistors. You have the AES CD-ROM collection -- look it up. The leakage currents are a function of both temperature and voltage. But with these FETs they got it dialed in where it worked great. Read the article and see for yourself. The only question is whether or not you can track down some of those FETs, and if not, whether you could get similar performance from another pair of FETs.
john curl said:
Well, the article conveniently omits any discussion about the capacitance of the FETs. And the capsule must be a relatively large one, as the capacitance is specified as 56 pF. But they are still able to achieve an equivalent noise of 20 dB SPL (A weighted), which if I recall correctly, is about as good as it gets for a studio grade condensor mic.
One possible pair that would be worth trying would be the Toshiba K246/J103 pair. They give the gate leakage current for the P-channel part on the data sheet but unfortunately not for the N-channel. However, the capacitances are quite low.
At 20 volts (a good starting point) Crss for the N-channel part is 2 pF and only 2.5 pF for the P-channel part. These kind of numbers are not going to cause a lot of problems, especially with a large diaphragm capsule. Plus you get rid of the big resistor. I'd try it if I were building a condensor mic.
Charles Hansen said:
To the contrary this 2pF has a small amount of non linear, voltage
dependant modulation. When driven by an ultra high source Z
it is worth consideration.
I believe tube mics don't sound great just because they have a
contoured freq response and added harmonic spectra, all that can
be done too easily in SS.
T
It is true that a tube will have a more linear input capacitance than will the FET. And it is also true that with a capacitive source (ie, the capsule), a capacitive voltage divider is formed. So any non-linearity in the input capacitance will show up as distortion (easily measured, by the way).
So yes, in absolute terms a tube mic will sound (and measure) better than a FET mic (assuming all else is held equal). But the question is how much better?
At 100 dB SPL, the mic capsule in the paper will output 20 mV. And if you look at the curves for Crss on the Toshiba FETs running with 20 volt supplies, the input capacitance will vary by a factor of 2x with a change of input voltage of 10x. So in this case we can make a rough calculation of the distortion created by the non-linear capacitance of the capacitors.
At the neutral position, the capacitive voltage divider will give 56/(56 + 4.5) = 0.92561983 voltage gain. And at 100 dB SPL, the input capacitance of one FET will increase by roughly 0.0002 pF while input capacitance of the other FET will *decrease* by roughly 0.00025 pF for a net change of 0.00005 pF. So now the capacitive voltage divider will give a voltage gain of 56/(56 + 4.50005) = 0.92562059, or a change smaller than 1 ppm or 0.0001% distortion.
So either I've made a big mistake somewhere in my calculations, or else we should look elsewhere for the reason that tubes sound better than FETs.
So yes, in absolute terms a tube mic will sound (and measure) better than a FET mic (assuming all else is held equal). But the question is how much better?
At 100 dB SPL, the mic capsule in the paper will output 20 mV. And if you look at the curves for Crss on the Toshiba FETs running with 20 volt supplies, the input capacitance will vary by a factor of 2x with a change of input voltage of 10x. So in this case we can make a rough calculation of the distortion created by the non-linear capacitance of the capacitors.
At the neutral position, the capacitive voltage divider will give 56/(56 + 4.5) = 0.92561983 voltage gain. And at 100 dB SPL, the input capacitance of one FET will increase by roughly 0.0002 pF while input capacitance of the other FET will *decrease* by roughly 0.00025 pF for a net change of 0.00005 pF. So now the capacitive voltage divider will give a voltage gain of 56/(56 + 4.50005) = 0.92562059, or a change smaller than 1 ppm or 0.0001% distortion.
So either I've made a big mistake somewhere in my calculations, or else we should look elsewhere for the reason that tubes sound better than FETs.
john curl said:
The curves for the FETs shown in the Sony paper are very similar to the curves for the Toshiba J74/K170 pair. (I know these are not suitable for this application, but they are the only datasheets I have handy for both polarities of FETs.) Basically, the N-channel parts have a sharp knee in their gate leakage current at around 20 Vds, while the P-channel parts have a very linear increase in Igss versus Vds. So if you run the parts at +/- 20 volts, the bias current will naturally "find" a spot where they are equal and opposite.
This won't be exactly the same with each pair of transistors, nor with temperature, so you would have to AC couple the circuit to keep any DC drift out of the picture. But the operating point should be easily maintainable within 10% or better, which should be fine under the circumstances.
My guess is that it was not used widely because Sony probably patented it. By the time that the patents expired, everyone forgot about it. But I still think it is a brilliant solution that solves many problems at once. And if I were designing the electronics for a condensor mic, this is the path I would start with. Sony was able to get it work well enough that they included it in a commercially available production product, so it couldn't have had any fatal flaws. At the worst case, it would require a lot of hand-matching of FETs. But, hey, you and I know a little bit about that, don't we?
Charles, we never forgot, it was just too impractical. P ch fets are not too good at these impedences. You forget that we can go up to 100G or more, in theory. Sooner or later, enough is enough. There is also reverse diode biasing, AND the real problem is not so much linear or non-linear capacitance, but: delta C / C Think about the capacitance modulation of the mike capsule itself when it is vibrating. This is how it makes signal. Then the FIXED capacitance of the load can be a problem. Even the so called cap. PAD is a real distortion producer. We have measured it.
Charles that's why I posted my circuit. It's a very simple updated version of what Neumann was doing with some of their KM series mikes. I know you hate it but feedback works for other things too. In the charge amplifier mode there is no voltage change on the Vdg of the FET. Also there is no matching involved in the "leakage" biasing. A diode forward biased by picoamps has a very large equivalent resistance (~Is/Vt), the trick is to find a junction with a very small Is, like a FET. The 2N4118a's (smoke detector) are perfect at 100fA or so leakage.Charles Hansen said:
BTW have you ever opened up one of those cheap Panasonic capsules? There's nothing in it but a FET with it's gate tied to the back plate. Even Bob Pease was scratching his head untill I pointed out a FET will equilibrate at slightly over Idss when the drain to gate leakage just barely enhances the gate to source. I think the 2SK123 data sheet shows this if you read between the lines (IIRC 1mA Idss with a 1.4mA max in (mike) circuit Id).
Try it some time, very slow on a curve tracer sweep Vds with the gate floating. If you have a lowish Idss FET you can make a source follower with a resistor to ground and couple the floating gate with a small cap, works fine for those .50$ mikes. Distortion and DNR OK for cell phones etc.
But I have built a version of my charge feedback mike with a higher quality capsule and the 2N4118A (as diode) and there is virtually no distortion even at almost rail to rail out. The 5G version is easier to build and works fine. I was asked not to tell who but there is an (expensive) commercial mike that uses the BAV?? pA leakage diodes in the same way.
The cheap mikes are now all MOSFET with b-b diodes tying down the gate, same idea.
- Status
- Not open for further replies.