Not really surprising. Zener breakdown is a quantum tunneling of carriers through the bandgap. It is the dominant breakdown mechanism for highly doped p-n junctions, therefore low voltage zener diodes are going always have the lowest noise.
Higher voltage diodes, like LM329 with 7V breakdown, have a combination of zener and avalanche noise, therefore much higher overall noise. What the LM329 does, it buries the junction, thus avoiding the edge effects, usually the worse offenders when it comes to controlling the avalanche breakdown and the associated noise, but even the planar breakdown still has a significant component of avalanche.
Comparing a 2.7V zener noise with the LM329 is not fair; find a BZX6V8 zener with lower noise that the LM329 (hint: you won't).
TL431 are indeed bandgap references + a shunt regulator, not intended for low noise. Maxim makes a few very good low noise references, see the MAX6126, 35nV/rthz at > 1KHz. Linear does the LT6655 with 18nV/rthz, but is specified so only for the 1.25V output (likely little to none gain stage noise contribution).
Last edited:
They are usually not stocked in quantity so they had to build them. I also wanted a better price but that didn't help much. I think they cost about $12 each even at 1K quantity. Not bad really especially compared to magic audiophile cables. I think they are interesting but probably not really a practical audio part especially with a good Toshiba inventory.
I'd heard but maybe just dreamed that the singles are actually with two chips inside, but without the leads brought out. I guess they make them that way and probe for matching in case it is good and they can make a dual.
Mouser has the duals for about 50 bucks in single quantity, I see.
The BZX84C6V8 is the purple trace; that's where the LM329 should be and the BZX84
1/f corner is at 10 Hz and not at 100. The BZX84 is probably ion-implanted, too,
nowadays that's no big thing for even mature production lines.
But hard for backyard start ups it is.
In my plot, you can easily see the progressively higher noise levels from 2V7, 3V3, 4V7 and 6V8.
The _Z12 etc in the file name of the traces is the inscription on the SOT-23, so I can re-find them easier.
A bag with LT6655 lies on my lab table; I had no more time last sunday evening.
Even more interesting is the LT3042, a complete regulator.
Last edited:
I should mention this up front. The LM329 results are right in middle of the datasheet spec (7...100uV RMS in 10KHz BW, that makes 70...1000nV/rtHz, or 36...60dB in the graph scale). What's really stunning is the 6V8 zener results, I would double check this for experimental errors. I just checked a no name 6V8 zener and has 65dB in the same graph scale.
Last edited:
Prob. a single, o/w there would be a symmetrical arrangement of stripes for D-S --- presuming that things are not interdigitated. Scott W. do you concur?
I don't know where I got that notion. I may have been thinking about the THAT Corp. bipolar arrays, which run out of package pins to bring out all of the transistor connections.
http://ams.aeroflex.com/metelicsHRC/pdfiles/1N4099-1N4135_1N4614-4627.pdf
Aeroflex, 6V2 has a max noise of 5uV/rthz (73dB) and 6V8 40uV/rtz (92dB).
http://www.vishay.com/docs/85770/mmbz4617.pdf
Vishay, same 5uV/rthz for 6V2
http://www.centralsemi.com/get_document.php?cmp=1&mergetype=pd&mergepath=pd&pdf_id=CMPZ4614-27.PDF
http://www.centralsemi.com/get_document.php?cmp=1&mergetype=pd&mergepath=pd&pdf_id=1n4099-4135.PDF
Centralsemi, claimed as "low noise zeners", same values.
What Gerhard measured is likely a truly exceptional 6V8 sample (all of the above are specifying the max values). I would not count of such low noise being the norm.
All low value zeners are spec'd above at 1uV/rthz (60dB). Even those seem to be outstanding as measured on Gerhard's graph.
Yes that looks like a multi-bond option on a single, the gates are shorted anyway if they are the substrate. Remember the standard dual pinout is mirrored, so these could possibly be plonked down in either rotation and bond up.
Could at least get four amps on that die.
- Status
- Not open for further replies.