An (almost) Automatic Rbb Extractor: the Black Knight

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi Elvee,
Sorry for my mistake, I meant to ask BC327 and BC337.
OK, I have tested a number of examples of BC327/37, plus a few other similar transistors (and some not so similar too).

Conditions are always the same, 2mA - 2V
Code:
Type        Mfr    Prod. year    Ree      Rbb

BC328       Phil    1979         0.09     258
BC327       Phil    1996         0.06     >650
BC327/16    Phil    2003         0.02     49
BC328/40    MultiC  ~2000        0.31     304
BC327       TFK     ~1976       <0.01     208
BC327/16    Sie     ~1988        0.09     467
BC328       Moto    1991         0.06     178
BC327/16    ST      ~1991        0.02     33
BC636G      TFK     ~1985       <0.01     48
BC640       Phil    1985        <0.01     60
2N5322      ST      1989        <0.01     74

BC337       Phil    1980         0.31     269
BC337/40    Phil    2000         0.10     118
BC338/40    MultiC  ~2000        0.34     444
BC537       Fair    ~1982        0.09     64
BC639       Phil    1993         0.07     43
2N4401      Moto    1992         0.09     55
BSS38       Phil    1983         1.10     124
KSP42       Fair    2001         0.08     46
MPSA42      Phil    2006         0.13     118
ZTX651      Zet      ??         <0.01     80
ZTX851      Zet      ??         <0.01     84
ZTX690      Zet      ??         <0.01     114
ZTX869      Zet      ??         <0.01     200
2N5320      ST      1999         0.08     71

What can we draw from these figures? At first sight, values seem to be all over the place.
A first thing to understand is that we do not have to reason in terms of type N°s, or even manufacturer, but in terms of processes: it is perfectly obvious for instance that a BC337 made by Philips in the years 2000 is a completely different beast from the one manufactured in the early eighties: it has vastly improved, with parasitic resistors divided by ~3.
That is indeed a very broad and strong trend for the whole of the industry: more modern processes are almost always an improvement over the previous ones.
There are some rather rare exceptions, like the BC327 made by Philips in 1996 which is a real stinker: it even exceeded the capacity of the instrument.
Probably an isolated production glitch, because in general, the production of Philips is quite consistent.
The semi's from Multicomponent are consistently below the lot, even the modern ones, but since they are cheap as dirt, one might say that you get what you pay for...

Typically, a process cycle seems to last a few years, perhaps 2 to 5 years.
While a process lasts, the production is very consistent, with typical deviations of a few percents, perhaps 10 or 20 maximum.
Then, there is a "switch", and values change completely.
The main driver for these changes is probably economical, streamlining of the production, shrinking the silicon area, but better, more accurate fabrication techniques also improve the characteristics.

That is a sad thing to say, but the semi's in beloved vintage equipment's of the 70's 80's are mostly crap by today's standards....

The list above contains stinkers, but also a few gems: the BC327 from ST is an example of such a gem.

The famed 2N4401 is also among the "good ones", but not the absolute best.
Completely unexpected is the KSP42: seeing a high-voltage transistor as one of the two best is a complete surprise.
Note that I have tested other vintages/makers of 2N4401, and they weren't that good.

The ZTX series is disappointing, but of course they are designed for ultra-low emitter and collector resistances, and not particularly for low Rbb.
Sorry about the missing production dates, but they have no date code, and I don't feel confident enough to guess even an approximate year

If you PM me your shipping address, I would be glad to send you a few samples of the lowest Rbb transistors that Horowitz and Hill measured, in their 2015 survey of 54 different BJTs (Art of Electronics, 3rd edition, p.501). I'll also toss in a couple of MPS6571s; these are the NPNs that Nelson Pass used in the Threshold M1 moving coil pre-preamp, and that he also used in the Threshold NS10 preamplifier phonostage and line amp stage. Datasheet attached.

_
OK, you do not need to send more than one sample for each different production batch: as I said, the production is quite consistent, at least for post-70's semi's

A reminder of the warning I gave in the introduction: this tester is not a lab instrument, and values have to be taken for what they are: indicative.
That is particularly true for PNP's, for technological reasons, and for "good" transistors, especially the ones having a largish Ree: in this case, the Ree*β term is heavily dominant, and its subtraction from the raw Rbb value maximizes the errors.
 
Member
Joined 2011
Paid Member
If you PM me your shipping address, I would be glad to send you a few samples of the lowest Rbb transistors that Horowitz and Hill measured, in their 2015 survey of 54 different BJTs (Art of Electronics, 3rd edition, p.501). I'll also toss in a couple of MPS6571s ...

Sent today: 16 samples.

_
attachment.php

_
 

Attachments

  • Transistors_To_Elvee.jpg
    Transistors_To_Elvee.jpg
    435.8 KB · Views: 1,246
Horowitz and Hill report rbb = 1.67 ohms for the ZTX851 while you report 84 ohms, that's rather a large difference.
That's worrying: as I don't suppose H & H used heavily simplified measurement methods and gear like I did, my figure is most probably incorrect.
I probably went one or two simplifications too far.

I'll have to review the possible causes for such an error.
The direction of the error is unexpected: most simplifications and errors should in principle result in a measured value lower than the actual one.
Among the few plausible culprits are the self-heating and the emission factor.

The thread will return on the back-burner for the time I investigate....


Elvee,
What do you think of this method of measuring Rbb based on the transistor voltage noise? I saw this circuit years ago but had questions because some transistors advertised a low Rbb had high voltage noise.
It is a known and recognized method for measuring Rbb, but as shown, it is simplified and only gives approximate results, but even the full accuracy method cannot be 100% accurate, because there are excess noise sources unaccounted for.
Of course, all external methods of measurement for this parameter use some kind of approximation.
 
No, I had hoped for an illumination, but so far I have seen no light.
The ideality and self-heating fixes I added work as intended, but some higher order effects still pollute the measurement, and I have no clue as to which might be dominant: the number of minor deviations mechanisms in semiconductors is overwhelming, and it's one that not taken into account by spice, because in sim, it works perfectly.
This also means that the sim is useless at providing clues.
 
I use a variant of this method as a sanity check, because although not very accurate, it is reliable and predictable (it provides an upper bound for the resistance value).
For relatively ordinary transistors, it is in good agreement with Horowitz & Hill results, but for the more exotic types, the discrepancies become enormous.
Maybe those transistors have extremely low Rbb and extremely high excess noise, or else.....
 
Yes, more or less.
This is the principle:
https://www.diyaudio.com/forums/att...lack-knight-base-spreading-resistance-rbb-pdf

The real thing needs to take in to account some practicalities, like the ability to test both polarities or the usability.

This is the complete tester, with an umbilical-connected supply:

attachment.php


This configuration eliminates hum issues.

4 switches control the configuration. The "safe insert" mode mutes the supplies, discharges all the coupling caps and lights the socket zone:

attachment.php


The input resistor is a 20 turn, 220 ohm trimpot:

attachment.php


Two outputs are available, but further filtering is normally necessary:

attachment.php


A connector gives access to the resistor value:

attachment.php


The switch shorts the resistor:

attachment.php


Everything is carefully shielded, of course
 

Attachments

  • NoiseJig0.jpg
    NoiseJig0.jpg
    240.8 KB · Views: 437
  • NoiseJig1.jpg
    NoiseJig1.jpg
    256.3 KB · Views: 413
  • NoiseJigB.jpg
    NoiseJigB.jpg
    231 KB · Views: 426
  • NoiseJigC.jpg
    NoiseJigC.jpg
    198.1 KB · Views: 421
  • NoiseJigA.jpg
    NoiseJigA.jpg
    204 KB · Views: 414
  • NoiseJig5.jpg
    NoiseJig5.jpg
    207.2 KB · Views: 397
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.