I used a 1 mA Ic and a Vce of about 10V in a differential test jig. (See: http://dragonflyalley.com/synth/images/ ... 11_144.pdf)
My circuit was a little different with the collectors to +10V Vcc, bases grounded, the emitters pulled to -10.6V through precisely-matched 10KΩ resistors. (About 0.01% ratio-matched using a Kelvin bridge measurement).
-10.6V (approximately) sets the combined collector current at 2 mA which I monitored with a mA meter.
I had 13 ZTX851 samples from one lot and matched them first.
I then pulled another 10 from a second lot of 100 that had been shipped to me months later and matched them.
18 of 23 ZTX85 matched within 1 mV.
Of the 18 matched pairs, 12 ZTX851 were matched better than 500 µV.
The remaining 5 were outliers trypically 4 mV, one was about 15 mV off from the rest.
The yield for matching within 1mV was about 78%; for 500 µV or less it was 52%.
Of the ones within 500 µV many were well below 250 µV.
I found that common "bubble pack" packaging laid over the test jig provided a nice thermal blanket to insulate from the air conditioner vent blowing towards the workbench.
I used needle nose pliers to handle the transistors to reduce test time from the heating of handling.
After about a minute under the blanket the pair would stabilize without being thermally-coupled.
The one minute settling time was with the jig continuously-powered.
There is a fair amount of self-heating even at 1 mA.
If one device is powered for a significantly long time as the other device is switched-out, the settling time, for the devices to reach thermal equilibrium, is longer.
Switching off the supply when changing devices yields much faster settling times.
If you're doing a lot of devices or double-checking your work powering down could be a significant time saver.
Differential measurement makes pairing easier.
The key is to pick a good reference transistor and then match to it.
If you grab one and start seeing a whole run of them with 4 mV offset, set that transistor aside.
Grab two of the ones that measured similarly and then pair them.
I'm satisfied that someone wanting a couple of matched pairs could buy 10 and likely get two if not three pairs within 500 µV.
Hfe Measurements of the Paired Devices
Once I had paired up devices, I measured the Hfe of each pair and recorded the lowest Hfe of the two.
If the Hfe measured 147 I marked it ">140" and set bins 10 units apart.
The Hfe average of the 18 pairs, based on bin value, was 143.
The lowest Hfe was about 130; the highest 180.
Since devices were binned at values lower than they actually were the real Hfe average is a little higher.
The ZTX851 is not the "Ultimate" Low Noise Transistor
I'm sure this thread could degrade (in other fora) into a pissing match about what the lowest noise transistor is and that the Hfe is low, the capacitance is high, its not spec'd for low noise, it must have high 1/f yadda, yadda, yadda. Negativity on parade.
Show me something that can be bought.
The reality is that the most-desireable low noise transistors are simply no longer available.
The ZTX851 is available and is dirt cheap.
You don't have to parallel some ridiculous number of them to get low noise.
The ZTX851 has a high Cob. (45 pF typ.)
The ZTX851 was designed for auto horns and fluorescent emergency lamp electronic ballasts.
Low noise performance of the ZTX851 in a mic preamp seems to be the polar opposite of its use in auto horns.
It's big geometry, indestructable and is through hole.
But: You can buy them.
The rbb of the ZTX851 was measured by Horowitz and Hill to be about 1.6 Ohms. (Art of Electronics, 3rd Ed. pg. 535.)
"Although its noise voltage is unspecified, we found excellent noise performance good consistency with some Zetex bipolar transistors, notably their ZTX851 (NPN) and ZTX951 (PNP)."
I'm going to epoxy two devices back-to-back and call it a "ZTX2851" in a "TWO-92" package.
My hunch is that this degree of thermal coupling, without added mass, will likely not have short-term thermal drift requiring excessive servo correction once in a shielded (from air currents) enclosure. If it does then we'll just epoxy it into a nut like before.