From: Mike Monett
X-Mailer: Mozilla 2.02 (Win16; I)
Subject: Re: Prof Searle's 100MHz 30dB 1968-era 3-trannie amplifier
References: <3E2358BE.email@example.com> <3E2A0C89.7FAB@sneakemail.com> <3E2A4E8D.78BE@sneakemail.com> <3E2A868D.1D0E@sneakemail.com> <3E2AE81D.73AC@sneakemail.com>
Date: Sun, 19 Jan 2003 17:16:10 -0500
NNTP-Posting-Date: Sun, 19 Jan 2003 17:15:36 EST
Organization: Bell Sympatico
Winfield Hill wrote:
> Mike Monett wrote:
>>Here's the SPICE files:
> OK, that's a shunt-series pair, as I posted, but you call it "Prof
> Searle's 100MHz 30dB 1968-era amplifier." My question is, did you
> simply use my thread title, or do you remember, was Prof Searle's
> design a shunt-series pair, or possibly was it something else? For
> example, Jim Thompson's design isn't a shunt-series pair, yet
> sports a certain simplicity.
You were right - sorry for the confusion. I just used the thread
title, and have changed the html file to show the relation to this
thread in case anyone else joins the thread in the middle.
I don't know for sure if this is the design he used, but I strongly
suspect it. There was an instant flash of recognition as soon as you
posted it, then my mind starts playing tricks, the same as changing
your answer on a multiple-choice quiz.
My gut instinct is this is the one he used. I have a strong
impression the feedback went to the emitter, which limits the number
of configurations in a two-transistor design. I believe the output
went straight into 50 ohms, which accounts for the emitter follower.
But this is just speculation. I don't have a copy of the original
> . Mike Monett's mods, V+ V+ = 15V
> . shunt-series pair | | G = 30.5dB (no load)
> . (Art of Electronics R4 | -3dB BW = 560MHz
> . page 872) V+ | C Q3
> . | +----B all MMBR941
> . R2 Q2 | E
> . | C | 8.9V
> . +------- B +---------- out
> . | E |
> . | | 5V | R1 = 50 (or ext source)
> . ,--- | --- R6 --+--, | R2 = 750
> . | | | | | R3 = 100
> . | C R5 === | R4 = 470
> . -||-- R1 -+- B Q1 | | | R5 = 470
> . | E ground | R6 = 10k
> . R9 | | R7 = 3.9k
> . | +---||--- R7 -----+ R8 = 470
> . gnd | | R9 = 15k
> . R3 R8 bypass R5
> . | |
> . gnd gnd
> WRT your version, I see you've added R9 to bias Q2's emitter up at
> 5V. Is there a reason for that (e.g., higher Q1 Vce), or was it to
> reduce R2, thereby extending the R2-Cob Miller integrator
> bandwidth of the 2nd stage, etc.?
I used the time-honored "tune for maximum smoke" to get a feeling
for the voltage stackup. Maximum smoke means max output swing with
both hands firmly planted on your seat.
I set the collector resistances low to drive the stray capacity (not
shown) and Miller effect. The bias drove me crazy. I recalled that
R1 just pointed left in your original diagram, and it was not
connected to anything, so I added R9 to make a voltage divider. The
values for R6/R9 are way too high, but it was getting late and I was
running out of time to make the simple change and have to redo all
The problem with my procedure is R5 got quite large. This means a
bypass cap is needed across R5. Your circuit didn't need one, so I
really would like to see the more organized approach that you would
> Another comment, with R1 = 50 ohms (i.e. the RF generator's
> impedance) and R6 = 10k, the first R6/R1 feedback loop of a usual
> shunt-series pair is basically wide open, increasing the open-loop
> gain available for the second R7/R3 feedback loop. Hmm, I wonder,
> is this a good idea, or a bad idea?
It's definitely not a good idea to have such high resistance values
for R6/R9. Lowering them would tend to fix Vb1 which stabilizes the
voltage across R3.
I like to have a resistance in series with the source just to get an
idea of the input impedance. I raise it during the analysis to see
how much effect it has on the results. In this case, the feedback to
the emitter boostraps the input impedance, so it is quite good.
> BTW, in a real design I'd scale back R6 and R9, to reduce the
> uncertain effect of Q1's beta. Usually high-bandwidth transistors
> have very poor DC gain. In this case, 1.7V of Q2's 5V emitter
> voltage is from Q1's uncertain base current dropping across the
> 10k resistor. :>)
Yes, you are absolutely correct. It was just getting late.
> - Win
Win, I'm sure there must be an optimum supply voltage where anything
less would simply not give enough to bias the trannies on, and more
would simply waste power as my approach did.
Can you show us your approach? I would like to sit back and be
enlightened as always when you talk about how to design.