On Mon, 01 May 2006 10:36:06 +0100, Pooh Bear
wrote:


Yes - I have a box of old MAT-01s from PMI. They are strictly reserved
for such projects. I don't know if they are still available.

I recall the beast.

Somewhere I think I have some of those similar Nat Semi parts that featured multiple
devices on-die. Forget the part number now. Oh no - I *was* right - the LM394 - just
checked in case. Though that would be an IC but the M just means monolithic. They're not
even insanely expensive now !


Yes that is very similar. The great thing about using these discretes
as front ends is that the knee frequency for 1/f noise is way lower
than the transistors in the average op-amp. So not only don't they
hiss - they don't rumble either. Three of them in parallel is about
right for a mic at about 150 to 200 ohms.

Also took a look at some esoteric fet data a while back. Noise somewhere down in the
500pV/sqrt Hz region. Interfet is the company.


But what about the current noise? You need to multiply that by the
source impedance to add in its effect. I've looked at a few FET input
op amps with amazingly low voltage noise, and this always more than
makes up the difference.


I really wish noise was expressed as a noise figure, rather than a
level. That way it wouldn't matter what impedance you were using, you
would simply have a figure of merit that told you how much worse the
pre was than theoretically perfect.

Hmmm, I wonder how that would go down with those who 'cheat' by using a 150 ohm source
instead of 200. I note that Mackie ( I think ) is now quoting noise with the input
*shorted* too.


Yup, you really do have to read specs with a cynical eye these days.

d

--
Pearce Consulting
http://www.pearce.uk.com