Wide Band Diff Amp

[Charles Schuler]

Hi:

SwitcherCad III file posted on alt.binaries.schematics.electronics:

This is a SwitcherCadIII file for Fig. 13.11 from The Art of Electronics,
Second Edition, page 870. The SPICE model used for the 2N5179 is:

..model 2N5179 NPN(Is=69.28E-18 VAF=100
+ Xti=3 Eg=1.11 Bf=282.1 Ne=1.177 Ise=69.28E-18 Ikf=22.03m Xtb=1.5 Br=1.176
+ Nc=2 Isc=0 Ikr=0 Rc=4 Cjc=1.042p Mjc=.2468 Vjc=.75 Fc=.5 Cje=1.52p
Mje=.3223 Vje=.75 Tr=1.588n
+ Tf=135.6p Itf=.27 Vtf=10 Xtf=30 Rb=10)

There is no output loading (the 5 pF capacitor has been omitted) and the
amplifier is driven by an ideal voltage source. The bandwidth is a bit less
than expected.

 

[Winfield Hill]

Charles wrote...

SwitcherCad III file posted on alt.binaries.schematics.electronics:
This is a SwitcherCadIII file for Fig. 13.11 from The Art of Electronics,
Second Edition, page 870. The SPICE model used for the 2N5179 is:

.model 2N5179 NPN(Is=69.28E-18 VAF=100
+ Xti=3 Eg=1.11 Bf=282.1 Ne=1.177 Ise=69.28E-18 Ikf=22.03m Xtb=1.5 Br=1.176
+ Nc=2 Isc=0 Ikr=0 Rc=4 Cjc=1.042p Mjc=.2468 Vjc=.75 Fc=.5 Cje=1.52p
Mje=.3223 Vje=.75 Tr=1.588n
+ Tf=135.6p Itf=.27 Vtf=10 Xtf=30 Rb=10)

There is no output loading (the 5 pF capacitor has been omitted) and the
amplifier is driven by an ideal voltage source. The bandwidth is a bit
less than expected.

Without taking time to analyze your circuit, or the one in AoE,
I note that we use Ccb = 0.5pF at 2V, see fig 13.11, whereas the
Motorola pdf file datasheet I have (most of my extensive original
oem datasheet collection was lost due to an unfortunate mistake six
years ago) has no typical curve and simply gives 1.0pF max at 10V.
This is a much higher value than we used, especially considering
that BJT capacitance decreases with voltage.

Spice models frequently use worst-case values, forcing one to go
in and edit them to see how a circuit is likely to really work
(creating loud arguments as to the legitimacy of such editing).

The 1982 NSC transistor databook says they made the 2n5179 with
process 42, and give curves for Ccb showing 0.56pF at 2V; they
also have specs of 0.5pF typ and 0.6pF max at 10V. I have some
other old datasheets at work and can try to explore further, but
I doubt I just pulled the 0.5pF value from a hat 25 years ago. I
believe the 2n5179 parts I used in the late 60s were made by RCA,
and they had much more detailed data. Maybe we can find a copy.

Thanks,
- Win

 

[Jim Thompson]

Charles wrote...

Without taking time to analyze your circuit, or the one in AoE,
I note that we use Ccb = 0.5pF at 2V, see fig 13.11, whereas the
Motorola pdf file datasheet I have (most of my extensive original
oem datasheet collection was lost due to an unfortunate mistake six
years ago) has no typical curve and simply gives 1.0pF max at 10V.
This is a much higher value than we used, especially considering
that BJT capacitance decreases with voltage.

Spice models frequently use worst-case values, forcing one to go
in and edit them to see how a circuit is likely to really work
(creating loud arguments as to the legitimacy of such editing).

The 1982 NSC transistor databook says they made the 2n5179 with
process 42, and give curves for Ccb showing 0.56pF at 2V; they
also have specs of 0.5pF typ and 0.6pF max at 10V. I have some
other old datasheets at work and can try to explore further, but
I doubt I just pulled the 0.5pF value from a hat 25 years ago. I
believe the 2n5179 parts I used in the late 60s were made by RCA,
and they had much more detailed data. Maybe we can find a copy.

Thanks,
- Win

Caution: CJC, in the model, is the value at *0V*. So if you wanted
to do it right you should do a back-out simulation on CCB at 2V to get
the proper CJC value at zero.

It's a function of MJC and VJC and can be calculated by hand. Just
see a Spice reference for the equation. (I'm not going to try doing
it in ASCII here ;-)

...Jim Thompson

 

[Winfield Hill]

Jim wrote...

Caution: CJC, in the model, is the value at *0V*. So if you wanted
to do it right you should do a back-out simulation on CCB at 2V to
get the proper CJC value at zero.

It's a function of MJC and VJC and can be calculated by hand.
Just see a Spice reference for the equation. (I'm not going
to try doing it in ASCII here ;-)

That's right, I hadn't forgotten that, Jim.

IICC, the back-biased portion of the Spice formula simplifies to
Ccb = CJC / (1 + Vcb / VJC)^MJC which gives us Ccb = 0.756pF at
2V and 0.54pF at 10V. This is lower than the 2n5179 worst-case
value, but it's 51% higher than the presumably-typical value we
used. (Simply using the junction-capacitance formulas without
adding/subtracting the TO-18 lead capacitances included in the
data-sheet values will certainly cause us trouble, but we'll
ignore that.)

If Charles edits his CJC from 1.042p to 0.689p, he'll get 0.5pF
at 2V, IICC. If he further evaluates and corrects the f_T values,
etc., to more realistic values, he can better compare our simple
back-of-the envelope calculations to the mathematically-intensive
possibly-more accurate Spice answers.

As for me, SWMBO reminds me (again) that the lawn mower has been
calling for several hours now, while the day gets ever hotter.


Thanks,
- Win

 

[Charles Schuler]

If Charles edits his CJC from 1.042p to 0.689p, he'll get 0.5pF

at 2V, IICC. If he further evaluates and corrects the f_T values,
etc., to more realistic values, he can better compare our simple
back-of-the envelope calculations to the mathematically-intensive
possibly-more accurate Spice answers.

As for me, SWMBO reminds me (again) that the lawn mower has been
calling for several hours now, while the day gets ever hotter.


Thanks,
- Win

Hi Win:

Thanks for your input. The above increased the bandwidth a little. I have
been playing with simulations of wide band circuits, and others, mostly to
get comfortable with SWCAD. Your book has long been a major source of ideas
and down to earth analysis methods.

Thanks to Jim as well.

Chuck Schuler

 

[Winfield Hill]

Charles Schuler wrote...

To satisfy my curiosity, I reduced TF in the SPICE model to 10p and
the bandwidth increased to 400 MHz.

Oops, unless you change the transistor to a GaAs part, or some such,
I'm not sure if you're allowed to play with the forward transit time.
Perhaps Jim can give us more guidance.

Another trick was to place a 0.22 uH peaking coil in series with the
300 ohm resistor. That also did a nice job of increasing BW but, of
course, the pulse response (overshoot) got a little nasty. I guess
I miss the smell of solder these days!

That's a nice trick, not shown in our circuit but a good improvement.
Of course, you can control the peaking level with the choke's series
resistance (add R if needed), Q = sqrt(L/C) / Rs = 1.5 etc is good.

Thanks,
- Win

 

[Tony Williams]

The 1982 NSC transistor databook says they made the 2n5179 with
process 42, and give curves for Ccb showing 0.56pF at 2V; they
also have specs of 0.5pF typ and 0.6pF max at 10V. I have some
other old datasheets at work and can try to explore further, but
I doubt I just pulled the 0.5pF value from a hat 25 years ago. I
believe the 2n5179 parts I used in the late 60s were made by RCA,
and they had much more detailed data. Maybe we can find a copy.

My 1977 RCA cat has only a single page of data on the 2N5179.

Ccb = 0.7pF typ, 1pF max, at Vcb = 10v and IE = 0mA.

Cib = 2pF max, at VE-B = 0.5v.

Collector-Base RC (rb'*Cc) = 14pS max.