toggle quoted messageShow quoted text
I appreciate your time taken for this response .... I have got the “gist” of this, although I will ponder a few of the issues further ... eg. I am assuming that the top part of the rise time response curve was being driven by the ‘troubled’ transistor .... and bottom section driven by the ‘good’ transistor ... Push Pull??
And I don’t yet understand the HF compensation circuitry.
Anyway, another brick in the wall ...
Hello Ian, in response to your message further down...
I've got busy in the past couple of days which prevented me from answering
First of all, please don't expect too much... because I think that, in the
end, it had a 50% of plain luck...
On this 7623A, .as I got rid of the gross issues that it had, and could focus
on doing a careful HF step response, I noticed that no matter what I would do,
be it on the Vert. Output Amplifier, or on the Vertical Plugin, I couldn't get
a clean step response in the whereabouts of 3.5ns.
If I would push too much on the HF compensation adjustments, I would
eventually get 4.5ns at best, but with a lot of ringing on the step corner.
If I would adjust it to tame the ringing, the step response would worsen even
more to - say - 5.5 to 6ns (this is a 60MHz scope, not a 100).
But what was puzzling me was that the step was not really like a slow
transition, but it was really fast up to about 2/3rds of the amplitude, and
then it would turn into a really rounded shoulder... So, I thought there were
2 superimposed time constants at play and that maybe, only maybe, I would be
able to spot which stage was not keeping up, by probing the vert amplifier at
It crossed my mind, for instance, that local power supply decoupling could be
lacking at some spot and that the slower last third could be due to a
decoupling having depleted its stored charge during the first part of a hi
speed transition and then, the stage would be running on "lean" mixture and
So I started looking at places where there should be a more or less steady
voltage value, either the power to each stage, or the biasing networks.
So I went on to probe the +15V (dcpl) that feeds the Q2405 and Q2505 (this is
the NPN pair that precedes the PNP 151-0271-00 tranny I found guilty) and also
the bias to the bases of the very stage where the 151-0271 trannies were. This
is a common base stage, that has about +8V bias, coming from the Thermal
balance adjustment voltage divider, and decoupled by a 1nF capacitor.
While probing the bases of the transistors looking for the base pin (I didn't
know it by heart), I eventually probed both emitters (of the two PNP
transistors) and noticed they were markedly different... much different than I
would expect from two complementary sides that were being driven with a nearly
symmetrical square wave, swinging in AC around the center point...
The signal at the base of one of the transistors had something like x mVpp,
while at the other emitter, if was like 4x.
Well, the bases are not tied together, but from the bias voltage, through 100R
resistors into each base, so I thought that the markedly different signals
would also be different from one base to another... but not... The bases had
pretty much the same signal (almost no signal in fact, since it's a decoupled
bias)... only the emitters were remarkably different.
So, I thought... well, either this is caused by the transistor themselves, or
it's the signal coming into them...
So I swapped the transistors, and the signal at the emitters swapped along...
So I knew the transistors were the cause for the difference... leaving only to
discover if they were just too different (un-matched) or if one of them was
So I ran them through my AVR transistor tester, which is far from being a
curve tracer, but would be able to tell me of any gross differences they might
While one of the transistors tested like a normal PNP transistor, with about
40 of gain (hfe), the other, despite showing an increased hfe of about 70
(almost double), tested like if it had an anti-parallel diode connected across
the emitter-collector, and this anti-parallel diode with a Vf of 3.5V... the
whole thing being completely unusual for a regular silicon transistor (not
being a mosfet, a darlington, or a power device).
Testing them with my DMM in diode test mode showed no differences between
So, despite they test equal on the DMM, I concluded that transistor could only
To confirm my theory, I needed a pair of small signal transistors that were -
at least - reasonably fast and I was lucky enough to find a couple of BFQ241
in my parts bin, which are high voltage, 1GHz TO-92 transistors used as video
amplifiers of CRT computer monitors.
In the moment that I shuffled them in, not only the signals at the emitters
became similar (as they should), but I immediately got a strong overshoot on
the square wave...denoting that the HF response of that stage started
contributing a great deal to the overall response.
After readjusting the HF compensation trimmers, I managed to get a risetime
about 2ns faster than it was, with a nice top almost free of ringing or
At this point, it's almost at the performance we would expect from a 100MHz
amplifier but, I decided to bring the subject to the group, and see what could
A fellow member of the group contacted me off-list to tell he has one original
151-0271-00 in spare and was so kind of offering it to me... so I`m now
awaiting for it to arrive, which will still take a while, as I`m in Brazil.
Hope this illustrates well how I got there...