Resistor values in instrumentation amplifier

[gutsofgold]

Hello. I am building up an instrumentation amplifier to use as a basic microphone preamp. I am using two NE5532s at +/-15V - one IC handling both input buffers/gain and the other IC as the differential amp. Based on the schematic below my R1 = 5k, Rgain = 100 ohms, and R2 = R3 = 10k. My breadboard works, amplifies the hell out of the input signal. The question I have is in the workings of the amp. Assuming the gain of the differential amp is R3/R2 and I don't need any gain here as the buffer amps are doing all the heavy lifting.. what difference does the value of R2+R3 make?? I chose 10k arbitrarily and it works but what is the correct value ??

[gutsofgold]

edit - not to confuse anyone but by "what difference does the value of R2+R3 make??" I mean the R2=R3 value... not adding them or anything.

[Digital Larry]

I think it's going to be similar to choosing the values in any typical inverting op-amp stage.  The ratio is what's important to the gain.  If the value of the resistors is high (in the megohms range) then you will introduce additional noise.  In addition, parasitic capacitance will have greater effect with high impedance circuits.  If the value of the resistors is too low, then currents in the op-amp might exceed specs leading to distortion or in extreme cases cooking.  To avoid worrying about it too much, use everyone's favorite, 10k.

That's my seat of the pants answer anyway.

[PRR]

10K is an entirely fine answer generally, and specifically here.

Uber-geeks who tweak the LAST dB of hiss would go 5K even 2K. However they certainly would first use a lower OSI part than the 5532. 5532 is optimized (and near-ideal) for 6,000 Ohm sources. Two of these series, 12K. This is a long way from the 100-300 ohms of your microphone. The Noise Figure may be 10dB or more. This *will* be fine with most modern Condenser mikes (if you can feed Phantom without killing the 5532). It will be hissy recording harpsicord with ribbon mikes or even SM58.

Look-up $10 mike preamp. There's a chip which does all this better than you can roll yer own.

[PRR]

The resistors you should be asking about are the 5532's internal noise resistance and the value of Rgain.

R(hiss) should (depending on approach) be similar or less-than source resistance. At 200 Ohms this means FAT transistors not common in ICs.

Rgain should be significantly less than mike resistance. The topology shown has Rgain variable and at some mid-gain point it won't be less than mike resistance. Scaling-down impedances can only go so far. Practical limits: R1 (||R2) should be fairly low but must not strain the amp, Rgain ends up low but too-low gets realy trouble with potentiometer (or switch) contact resistance and the inductance of wires to the front panel.

[mth5044]

Look-up $10 mike preamp. There's a chip which does all this better than you can roll yer own.

+ 1 and 2

[tubegeek]

Look-up $10 mike preamp. There's a chip which does all this better than you can roll yer own.

THAT1510 or THAT1512 are excellent possibilities for this purpose - essentially the same instrumentation amp but all on one chip, with laser-trimmed/precision matched 10K resistors, or at least that what it says on the box.

[R.G.]

Just as a bit of icing on the cake, resistor matching and tracking are critical in instrumentation amps. Many of the fine qualities of IAs are dependent on how well the resistors match and track.

Mouser sells the TI INA126 for $3.33 in ones, and it has all the internals set up so you only change Rgain to change gain. Works from +/-1.5V to +/-18V, and will beat anything you can hand tweak. There are many others; that's just the one I'm familiar with.

[PRR]

> TI INA126 for $3.33

I think Guts is looking for "microphone" preamp.

INA126 is a terrific high-impedance diff-input, and stunning low offset voltage, but has voltage noise at 35 nV/rtHz which is 5uV across a 20KHz band. It hisses like a 125K resistor, and we want lower hiss than a 150 Ohm resistor.

Slew of 0.4V/uS is *worse* than our beloved/despised 741. And if you try for gain of 500 (54dB) the bandwidth is 1.8KHz (9KHz at Gv=100 or 40dB). THD may become non-ignorable 100X lower.... it's pretty dubious for audio at gain=5.


The THAT parts are always good for what they say they good for. I think the THAT151* parts were what I was thining of.

TI offers the PGA2500 which is also an *excellent* mike preamp but is digi-controlled so it makes more sense in large arrays than for one/two channels.

[tubegeek]

I've built with THAT1510 and 4 9V's stacked for +/-18V rails, can't beat the results considering how easy it is to implement. A dozen parts, tops, the only one that's the least bit hard to find is the 10K reverse log pot for the gain set, you should really go switched-attenuator there anyway, I just got lazy.

Straight out of the app note, works like a charm.