LTspice - bc547 vs bc847

[matseng]

According to what I can tell from the NXP datasheet the BC547 and BC587 should be virtually identical except for the max power dissipation and thermal resistance caused by the different types of packages.

Yet the spice parameters (which I can't decipher yet) are different and simulations with the two different parts gives marked differences.

Why?    Which should I trust - the datasheet or spice-847 or spice-547? Or do I have to build and measure myself?

A simple test schematics:


The result where the 847 is merely a blip close to gnd:

[matseng]

Upping the current thru the transistors a bit by reducing the resistors connected to the collectors down to 100 ohm gives the following result:



Not as bad but still a huge difference for two transistors that should be identical.

[Kalvin]

Try to adjust the V2's first parameter from 0V down to -0.5V in small increments and see what happens.

[matseng]

With V2 initial voltage at -0.25 it definitely looks better.  But why? And is that fix something specific to the simulation or must it be applied in a real circuit as well?

[Kalvin]

I am only guessing here: The Vbe might have been modeled differently.

Also, your design assumes that the forward voltage drop of the 1N4148 is equal to the transistors' Vbe. You should not assume that. The Vbe and the diode's forward voltage drop changes with the temperature and may vary slightly from one manufacturing lot to another.

You might want to try to substitute the D1 as BC847 connected as a diode. Likewise, you may want to try substitute the D2 as  BC547 connected as diode. You will get better match for the Vbe. However, I would not consider this as a robust design.

If your design tolerates that the collector voltages OUT847 and OUT547 don't go to 0V, you could add small resistors or 1N4148 to the transistors' emitters. Better yet, change your design so that the bases are driven either 0V or 5V without any diodes.

[matseng]

Well, this was just a test circuit where I had simplified my original design as much as possible to view the differences between the 547 and 847.  The original design is a DTL nand gate so I really can't remove the diode unless I want to go over to RTL instead :-)  And adding the two dropper diodes in series with the base makes no real difference, nor adding a pulldown resistor between the droppers and the base.



But I simplified it all one step further and changed the timebase by three orders of magnitude, so now it's uS instead of nS.  Then it's apparent that the 847 is modelled quite different than the 547. The 847 is slow as a dead donkey compared to the 547... 1.2us compared to 40ns. Wtf?!?  They really should be identical according to the datasheet.

[Kleinstein]

Not all of the BC547 / BC847 are exactly the same. They are not that well defined - just basic general purpose NPN. So a well designed circuit should work with both modells. It would help to really show the models  they should not be that large. Likely someone here can expain the differenced.

For some parts there are models with different degree of accuracy / detail. Though today there is not that much need for faster simplified models. Back in the days (e.g. 68000 CPU) when it took about minute to get the DC operating point of a 741 type OP, this was a different thing. 

[Kalvin]

Either the model of the BC847 is bad or the E and C are swapped in the model. What happens if you swap the C <-> E of the BC847?

[matseng]

I'm actually trying to simulate a larger system with dozens of discrete logical gates connected in series to try to determine the propagation time of the entire chain.  But if the parts in LTspice is so badly defined/simplified that two transistors that should perform equally have a propagation delay varying from 40ns to 1200 ns it seems like a fools errand to even try to get some rough figures of real life performance figures.

The models have too many parameters for me to even start to comprehend, but maybe someone here do have a better understanding of what and why they are different.

Code: [Select]

.model BC547B NPN(IS=2.39E-14 NF=1.008 ISE=3.545E-15 NE=1.541 BF=294.3 IKF=0.1357 VAF=63.2 NR=1.004
ISC=6.272E-14 NC=1.243 BR=7.946 IKR=0.1144 VAR=25.9 RB=1 IRB=1.00E-06 RBM=1 RE=0.4683 RC=0.85 XTB=0
EG=1.11 XTI=3 CJE=1.358E-11 VJE=0.65 MJE=0.3279 TF=4.391E-10 XTF=120 VTF=2.643 ITF=0.7495 PTF=0
CJC=3.728E-12 VJC=0.3997 MJC=0.2955 XCJC=0.6193 TR=1.00E-32 CJS=0 VJS=0.75 MJS=0.333 FC=0.9579
Vceo=45 Icrating=100m mfg=NXP)

.model BC847B NPN(IS=1.822E-14 NF=0.9932 ISE=2.894E-16 NE=1.4 BF=324.4 IKF=0.109 VAF=82 NR=0.9931
ISC=9.982E-12 NC=1.763 BR=8.29 IKR=0.09 VAR=17.9 RB=10 IRB=5.00E-06 RBM=5 RE=0.649 RC=0.7014 XTB=0
EG=1.11 XTI=3 CJE=1.244E-11 VJE=0.7579 MJE=0.3656 TF=4.908E-10 XTF=9.51 VTF=2.927 ITF=0.3131 PTF=0
CJC=3.347E-12 VJC=0.5463 MJC=0.391 XCJC=0.6193 TR=9.00E-08 CJS=0 VJS=0.75 MJS=0.333 FC=0.979
Vceo=45 Icrating=100m mfg=NXP)


If I flipped the E/C of the 847 around I got this

[matseng]

For fun I flipped the 547 as well - but that really didn't affect anything.  The 547 curve still is fast and beautiful which is strange since the 847 changed significantly when flipped.

[Kalvin]

I thought that you had imported the models from the vendor and somehow the pins were swapped in the process. As you are using the .model statement, that is not the case. Anyway, nice observations

Unfortunately I cannot decipher the model parameters any better. Although, the page http://www.ece.uci.edu/docs/hspice/hspice_2001_2-102.html might provide some insight.

[Kleinstein]

The 547 model looks strange: TR=1.00E-32 is way to fast for any real silicon.

Though essentially all the parameters are different they may still behave rather similar.

I found even different models for the BC547 (though only A and C). So It is more than 2 to choose from.

[Kalvin]

Out of curiosity, how well the 2N3904 and 2N2222 perform?

[matseng]

The 547B/C have virtually identical performance.  The 847A/B/C have increasingly worse performance 1.05us/1.28us/2.10us, the higher the hFE the slower it is.

Graph with 2n2222 and the el-cheapo 2n3904 (which I incidentally have a bag of 1000 of here ^_^ )

[Kalvin]

If you put a small capacitor in parallel to the 1K base resistor does it improve the switching time?

[matseng]

Ok, here's the four transistors again, 847B, 547B, 2n2222 and 2n3904 - all having a 1 k base resistor and a 220 ohm to +5 at the collector.



I added a BAT54 schottky between base and collector as some kind of a single-diode Baker clamp and got the following results where all of a sudden they all perform just about equally.



Same , but zoomed in a bit to see the tiny differences better:

[Kalvin]

Nice! Miller capacitance in action. Here is a nice explanation for the anti-saturation Schottky clamp diode and the speed-up capacitor:
http://electronics.stackexchange.com/questions/23349/what-is-wrong-here-a-simple-npn-switch