low-noise sot-23 BJT transistors

Some of the DIY audio folks seem to be using 2N4401/3 which are available in SMT versions (MMBT...). Probably have less capacitance than those matrix base Zetex parts.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

Though the '5089 (and even better the '5087) have nice low noise figures, they aren't best for low impedance sources. I did some 'speriments a few years ago with the smt parts we had kicking around and found that the venerable '2222 (mmbt2222 in this case) did better than I expected. As I recall, it was actually slightly better than the mmbt4401, though that may have been a function of the particular lots I was sampling for my test. I'm remembering measuring input- referred voltage noise in the vicinity of 0.7-0.9nV/rtHz, at 1kHz. I was looking for an amplifier I could use to good advantage at the output of a standard diode DBM, source impedance of about 60 ohms when driven from a 50 ohm source.

Cheers, Tom

Hi Win,

Something strange is going on. One of my favorite low-noise BJTs was the 2N6429. At a few mA and > 100 Hz there was no way it was anywhere near 3.5 nV.

I couldn't find a data sheet for the pmbt6429 (is that a Phillips part?). But I found ones for MMBT6429 from both ON Semi and Motorola.

And their curves of en vs. freq at various collector currents were not only different from the 2N part, but strange, too. It doesn't make sense to me that they show 3.5 to 4 nV at 100 kHz at all currents from 300 uA to 10 mA.

The original 2N6429 had about 1.4 nV at 1 mA -- which jibes well with the ~ 100 ohm rbb' they show in the typical model. To get 3.5 nV, you'ld need 800 ohms of rbb'. Did they change or mess up the chip that much? Ccb and ft look the same. So, from them it seems to be the same chip. But I also note that hFE isn't shown in the SOT parts to hold up as well at low currents. The 2N part amazingly had typical hFE of

900 at 1 mA and was still 800 at 10 uA -- at least that's what the=20 data sheet typical showed.

Noise current In vs. freq. looks strange, too, on the SOT parts. They show it going _up_ at low currents and freq. above 10 kHz.

Perhaps you could measure en on your SOT parts and see if it matches the data sheet.

--=20 Regards, Howard snipped-for-privacy@ix.netcom.com

I'd like to take noise measurements now, but my venerable HP 4470A has just gone on the fritz. Perhaps in a week or two I can find the time to repair it. Meantime, I've ordered a selection of currently-available transistors from several manufacturers to evaluate.

Actually, all the '6429 datasheets I have in my computer show it to be a small-die transistor, best suited to operation below 100uA or so. What 2N6429 datasheet do you have that shows better results for low-Z source impedances (e.g., under 500-ohms)?

Yes, looking at the 24401 datasheet, I hadn't noticed the noise curves in the past. If you examine the noise-figure graph for Ic = 1mA (Motorola datsheet figure 10), the 3dB point occurs near 40 ohms, which corresponds to 0.8nV/rt-Hz at 1kHz. Not too bad. One could hesitate to use a high-current transistor (low r_bb') that didn't have a noise pedigree, but here one has been claimed.

Spef points out the DIY audio folks are also using them, and Graham says they're "fairly quiet for low Z audio sources".

I ordered a selection of smd versions from various manufactures.

The ON Semi data sheets I have for the MMBT- 4401, 4403, 2222 and 2907 all show noise figures below 3dB at 50 ohm source resistance at 1kHz, for 1mA Ic. The data sheets suggest the PNPs as slightly better than the NPNs, and for sure better for higher source impedances. The curves for both NPNs are as far as I can see identical, and same for both PNPs. I suppose the noise figure at 50 ohms would be slightly improved for slightly higher collector current. A bootstrapped cascode can help keep the power dissipation and therefore the die temperature of the input transistor low, and of course also gets rid of most of the capacitive effects.

Cheers, Tom

Hi Win,

The Motorola data sheet for 2N6428 and 2N6429 dated 1977. At 1 mA and 1kHz, this shows En as about 1.4 nV per root Hz and In as 0.8 pA per root Hz as typical. I assume this is for the 2N6429, which had the higher hFE. Cob is spec'ed as 3 pF max at 10 V. fT typically peaks at 400 MHz at about 10 mA and 5 V Vce.

Where it really shown for us was with a 50 ohm source impedance and 10 Hz. I calculate about a 6 dB noise figure at 1 mA bias.

--=20 Regards, Howard snipped-for-privacy@ix.netcom.com

Hmm, the 2N6428 and 2N6429 aren't in my 1974 databook, but they do appear in the 1983 databook, as a two-page spec summary, no graphs, etc. There's a little table for noise, which says Vn = 4.1nV for Rs = 500 ohms. 500-ohms has 2.9nV of Johnson noise, implying 2.9nV is from the transistor. It also says f = 10Hz. No sign of 1.4 nV and 0.8 pA. Weird.

There is also a 2N6428A. Just a thought.

Al