Jellybean hobbyist general purpose transistors..

[orolo]

But there is a catch, that I first heard of from Bob Pease: the higher the beta, the lower the output impedance. It can be seen in this datasheet: the 547A has 55.5k typical output impedance, the 547B 33.3k, and the 547C 16.6k, and it could go as low as 9k!. This can be very important when building current sources/sinks, active loads, etc.

Only relevant to pure current mirrors without degeneration.

Tim

Sorry for going a bit off topic, but honestly, I'm confused beacuse this beta thing has been bothering me for a time. Do you mean that there is a workaround for low rout, or that it's not relevant for current sinks and active loads? I've gone back to basics to see if I got this wrong, redoing example 2 here: http://leachlegacy.ece.gatech.edu/ece3050/notes/ISources/isources.pdf [There is an errata in the exercise: they changed m for n]. I attach the python code I used:

Code: [Select]

beta =  100.0     # Current gain.
Va   =   75.0     # Early voltage.
Vbe  =    0.65    # Base-emitter voltage.
Vce  =    8.00    # Collector-emitter voltage.
rx   =   40.0     # Base spreading resistance.
Ic   =    1.5e-3  # Collector current.
Vcc  =   15.0     # High rail.
Vdd  =  -15.0     # Low rail.
n    =    2.0     # Emitter resistor Vbe multiplier.
m    =   10.0     # Stiff current multiplier at the base.

Vt   =  300.0*1.3806e-23/1.6022e-19   # Thermal voltage.
alpha = beta/(1+beta)

Re  =  (beta*n*Vbe)/((1+beta)*Ic)
R1  =  beta*(Vcc-Vdd-(n+1)*Vbe)/((m+1)*Ic)
R2  =  (beta*(n+1)*Vbe)/(m*Ic)
r0  =  (Va + Vce)/Ic   
rep = (R1*R2/(R1+R2) + rx)/(1+beta) + alpha*Vt/Ic



rout = (r0 + rep*Re/(rep+Re))/(1.0 - (alpha*Re)/((rep+Re)))

print "beta    ", beta
print "va      ", Va
print "Vt      ", Vt

print "Re      ", Re
print "R1      ", R1
print "R2      ", R2
print "r0      ", r0
print "rep     ", rep

print "Output resistance: ", rout

For the same circuit, if beta=100, Va=75, the resistance of the sink is 379k, in agreement with the exercise. Now take beta=200, Va=75/2=37.5. The resistance drops to 215k. Even if beta=400, Va=37.5, the resistance is still 218k. Beta does not compensate for Va drop. If we go to the extreme, beta=600, Va=12.5, resistance drops to 100k. At least we can say that high beta parts underperform in this role. Isn't it just the same with a common emitter with active load: will a piece with 15k-9k output impedance give the same gain than a piece with 50k?

Found Pease's article warning about high beta: http://www.electronicdesign.com/power/whats-all-vsubbesub-stuff-anyhow-part-2 . 

Edit: Aha, I found it! What was spoiling the resistance of the current source was not the Va, but the choice of resistors R1 and R2. For high beta you need a much stiffer current at the base divider, m=100 or so. I think I'll need to think this over, much deeper.    Thanks!

[Vtile]

How disturbing it is that VHF means very high frequenzy in a 30 MHz to 300 MHz range while every dog collar today transmit in Gigahertz range Wifi. Anyone else annoyed. 

[Vtile]

The 4117 were my list just because it is what it is and it still is available in the discrete form. Unfortunately the A-subtype is not hobbyist friendly priced in any form, while MMBF4117 is. (unfortunately in SMD, but luckily SOT23 is still somewhat usable, without PCB and professional equipment.)

There is a US seller who has PN4117As at a decent price:

http://www.ebay.com/itm/QTY-100-PN4117A-N-Ch-JFET-General-Purpose-Amplifier-/371603063222

Maybe he would ship you some, if you ask nicely.

If the seller do refuge to sell outside the US. why you would suggest he would change his mind if someone in Europe would asks. Will no happen, especially since the products are in "high-tech" category, which have had traditionally paranoid export limits / licencing.

In general ordering something from US ebay to Europe is far from hobby priced if the miracle happens and you find a seller who will ship to Europe.

PS. I'm not angry to you, your intention is/was good and someone in US propably will benefit on that information.

[bd139]

Indeed. China to Europe is cheaper than US to Europe!

[David Hess]

Indeed. China to Europe is cheaper than US to Europe!

China to the US is cheaper than the US to the US.

But buying semiconductors and other electronic parts of questionable heritage does not strike me as being particularly economical.

[bd139]

Depends on what you're buying really and what for. If it's for hobby purposes, it's difficult to justify a 10-15x price multiplier on some parts for an assured supply chain. If you're building industrial or commercial electronics, then there's no question however.

[Vtile]

... and PN/MMBF4393 in most packages are easy to find and inexpensive. ...

Obsolete, according to Farnell

[Cerebus]

... and PN/MMBF4393 in most packages are easy to find and inexpensive. ...

Obsolete, according to Farnell

That's Farnell's odd way of saying we aren't stocking this any more.

Digikey have about 45,000 MMBF4393 on hand and about 1000 each of 2N4393 and PN4393. Looks like OnSemi are dropping the TO-92 packages soon, but there are lots of others out there who aren't e.g. PMBF4393 from NXP, marked active.

Never trust Farnell or RS's indications of part lifetime status.

[T3sl4co1l]

Sorry for going a bit off topic, but honestly, I'm confused beacuse this beta thing has been bothering me for a time. Do you mean that there is a workaround for low rout, or that it's not relevant for current sinks and active loads?

Degeneration was the key I mentioned.  Negative current feedback improves the constant-current-y-ness of the output.  It doesn't take much: a current mirror at 1mA (r_e ~= 26 ohms) benefits greatly from R_E = 100 ohms, while losing only 0.1V in saturation.

Edit: Aha, I found it! What was spoiling the resistance of the current source was not the Va, but the choice of resistors R1 and R2. For high beta you need a much stiffer current at the base divider, m=100 or so. I think I'll need to think this over, much deeper.    Thanks!

Hmm, I'll have to think about that.

Does Vbe change much with respect to Vce?  I know that it does -- it's an oft-overlooked mode of operation of the BJT, and can be used for detecting saturation -- but I don't have a feel for how much, and if it's exaggerated for high hFE or not.  (I can imagine it might, as the base-thinning effect should act to increase r_bb' and Vbe.  This would also be consistent with Vbe falling in saturation.)

Tim

[Vtile]

My go to list atm..

BC549/559 C-type, low-noise version of 547/557. Vcb is substantially lower though. I think I get them in 2(npn):1 relation. For those frontend project failures.  What about BC550/560 the noise figure seems to be specked even a bit lower, with same other specs.
PN2222/PN2907 For true general use, rather low beta compared to BC459, 2n2222 is a such a classic, comparable to ua741 and 1n4148 and PN-series is fraction of the price.
BD139/140 hmm...

Small amount of both
J113 as a general purpose JFET
MMBF4117 as for a toying around part

For a igfet, BS170/2N7000, seems to work with 5V logic, while not fully on, but usable. I need to search on the mosfets a bit more.

Anyone seen sot-23 to to-39 canning machine cheaply on chinese evilbays??? 

[Zero999]

How disturbing it is that VHF means very high frequenzy in a 30 MHz to 300 MHz range while every dog collar today transmit in Gigahertz range Wifi. Anyone else annoyed. 

The designations LF, MF, HF, VHF, UHF etc. came about when electronics were much slower, than they are today.

[Vtile]

How disturbing it is that VHF means very high frequenzy in a 30 MHz to 300 MHz range while every dog collar today transmit in Gigahertz range Wifi. Anyone else annoyed. 

The designations LF, MF, HF, VHF, UHF etc. came about when electronics were much slower, than they are today.

Yep, when the silicon were still used for rectifier casings and not rectifying material. 

[David Hess]

Anyone seen sot-23 to to-39 canning machine cheaply on chinese evilbays??? 

That would be nice; there is a much larger variety now in small signal surface mount than leaded components.  There are SOT-23 to leaded adapters but I have not seen any that I like.

[floobydust]

Whatever happened to transistor technology improving?

I found TIP30's in my junkbox date code 1973. hFE measured 5. Tossed them in the garbage.
40+ years later, TIP30 still available and same specs. fT 3MHz, hFE 15-40, 30W.

Planar, perforated emitter, epitaxial, mesa, alloy junction etc. I thought improvements occurred over the past few decades.
I don't see why antique part numbers are still around.

edit - added some silicon crystal configs

[T3sl4co1l]

TIP30 is another one of those ancient parts, tied to ancient specs.  I don't know if TIPxx is JEDEC registered (they're originally Texas Instruments Plastic transistors), but they're widely available and genericized.

You might get one that overperforms, or you might get one that's shite.

I would be surprised to see a modern (since 80s) part perform as poorly as the originals, for the same reason I suggested regarding 2N3055 and such: the old diffusion lines are long since scrapped.  Compatible parts are made on newer processes, but not benefiting from modern designs, so they still suck, like in hFE linearity and switching performance.

You're more than welcome to buy new-designed parts, but, surprise surprise, they're more expensive.....

Tim

[bd139]

2n3055 is an interesting one. Modern ones are a lot better which means in some older kit the damn things oscillate violently. Have to stick a few pF across base/collector to make them shit again

[Vtile]

But back in the day they used to handpick the 3-transistors used to every lunar lander and much details were given to select the right biasing currents for certain tolerance lot of transistors.  ...But I can bet that at least half of the writers on this topic do have hands-on experience on this particular thing and not just "I have read about this historic procedure" like me.

[David Hess]

The 2N3773 is another ones of those parts which morphed from 0.2 MH to 4.0 MHz.

At one point Tektronix was using both the 0.3 and 0.8 MHz versions of the 2N3055 in the 76xx oscilloscopes and grading the 0.8 MHz parts for something (hfe or Ft?) to make 3 versions in the same product.  The 2.0+ MHz 2N3055s came later and I gather that they are the junk bin of TO-3 power transistors where other parts which meet the 2N3055 specifications but not their own get tossed.

I know of a couple times now where people had problems with oscillation in Tektronix high voltage inverters when replacing old and slow 2N3055s with fast modern ones.  Using a 2N3771G or 2N3772G which are 0.2 MHz or a 2N3055AG or MJ15015 which are 0.8 MHz solved it.  Tektronix changed something in their later inverters so they could use the 40 MHz D44H11 series (ring or perforated emitter transistor?) but I am not sure what, maybe the transformer?

But back in the day they used to handpick the 3-transistors used to every lunar lander and much details were given to select the right biasing currents for certain tolerance lot of transistors.  ...But I can bet that at least half of the writers on this topic do have hands-on experience on this particular thing and not just "I have read about this historic procedure" like me.

In designs which rely on matching, trivial grading can be the difference between excellent performance and adequate performance or working and not working.  I try to avoid designs which rely on absolute specifications.

[exe]

I found TIP30's in my junkbox date code 1973. hFE measured 5. Tossed them in the garbage.
40+ years later, TIP30 still available and same specs. fT 3MHz, hFE 15-40, 30W.

That's how it supposed to be. Manufactures should not change specs for the same part number. Besides that, I'm not sure BJTs are evolved much last ~20 years.

Still, there are nice things like 2STA1942, MJE15028, BD139 (I like it because of low Vce(sat), there are smd equivalents). Use them, why using old junk?

EDIT: fixed typo

[bd139]

bd139 ftw

[floobydust]

I don't know the politics of jelly bean transistors. TIP31 is the cheapest TO-220 in town and not very sexy.

I believe all these old parts got moved over to modern lithography, standard semi fab processes.
Any improvements in the technology gave a "die shrink" to get costs down and fT up.

I learned from IC's such as Signetics/Phillips/Raytheon NE5534 which I think was 28 masks.
The TI NE5534 part behaves totally different, so I called them and asked WTF. They told me they purchased the rights and are using modern op-amp lithography/standard process that still "meets the specs" of the old part. So it's a different part with an old number and new fab process. Old parts don't have a lot of tests and datasheet specs so they are easy to rip off.

The TI NE5532 I found terrible, literally kicked out of a recording studio for using it in repairs. Noisy and crappy sounding, compared to the old Signetics version.

[exe]

Old parts don't have a lot of tests and datasheet specs so they are easy to rip off.

Yeah, this what I don't understand at all. There are many parts with the same part number, but from different manufacturers. And they are different! Esp. opamps. How can I make purchase decisions? For example, I measured bandwidth of NE5532 from TI and from ONSemi and there is a noticeable difference. Same for MC33079. Or transistors, I found 2STA1942 from ST to be better in my use-case than corresponding part from Toshiba (although, I tested only one of each).

What's worse, nobody tests this. And even worse, manufacturers keep adjusting their processes... And on top of this a lot of rumors (esp. about audio stuff where audiophiles "reason" about sound quality with no understanding of how opamps work and no measurements).

[Vtile]

Most of the shared parts what I see, are from the low end and/or old designs. OPAs in general aren't "amps" in sense of audio amp, but arithmetic calculators. If I set up the LM358 as a PID-controller I doubt (since I haven't tested) that I would see really any difference between the manufacturers.

[bd139]

This is why people buy stuff from Linear.

Datasheets not datashits and proper specifications that are reliable.

[T3sl4co1l]

What's worse, nobody tests this. And even worse, manufacturers keep adjusting their processes... And on top of this a lot of rumors (esp. about audio stuff where audiophiles "reason" about sound quality with no understanding of how opamps work and no measurements).

If it meets the datasheet, and your application works based on datasheet information, then you're fine.

A perfect reason against designing something to be critical on the parts used.  Like hFE ("suicide") versus emitter degeneration bias, to use a simple and popular case.

Tim