On Wed, 18 Sep 2002, Olin Lathrop wrote:

*>> I do what I preach. Remember when I wrote about a clamping circuit and an
*>> amplifier ? Here goes (attached).
*>
*>I think this circuit would actually work, but it looks like a rather
*>hairbrain scheme.  Q1 and Q2 look like a sortof confused current mirror.

The circuit is tested and works perfectly as shown. Output is composite
sync with Voh >= 3V, Vol < 0.3V with 10pF load, with flanks shorter than
300nsec (and some compressed video above Voh). By changing the feedback
resistor to a higher value the output becomes rail to rail into 10pF load.

*>Since the Q1 emitter is grounded, its sole purpose seems to be to bias the
*>base of Q2 so that its emitter is close to 0V at a current of about 29uA.
*>The net effect is for the emitter of Q2 to present a low (a few KOhm,
*>dependent on the gain of Q2) impedence to signal levels below ground and a
*>high impedence (47KOhm) to signal levels above ground.  After several

One cycle, max two. The current is 29uA when VeQ2 == VeQ1. When the
emitter of Q2 is taken 10-50mV below that voltage the current becomes very
significant. You can consider that C1 is being held to 0V though 2 ohms or
so for low enough currents (<100 mA). I suggest a modeling with SPICE,
using a PWL input source that mimics a few (2-3) TV line signals. I used
BC182 transistors with beta about 250 at Ib = 100uA.

*>cycles, this will bias C1 so that the sync tips are just a bit below 0V and
*>the rest of the signal above 0V.  OK, that's useful, although it could have
*>been done with a cap, diode, and resistor (Just like in an old fashioned TV.
*>Gee, maybe those old geesers did know something.  Hmm.)

The trick is to clamp to 0V so a dc amplifier referenced to GND can do the
rest. If a transistor amplifier would be used it would be useful to clamp
to some other voltage, perhaps 0.5V.

*>Now comes the really strange part.  An LM393 is a low offset comparator with
*>common mode input range to ground and open collector output (hence the need
*>for R5).  R4 and R3 attempt to provide negative feedback around the
*>comparator for reasons I can only begin to guess at.  I doubt it will
*>actually operate nice and linear with a gain of about 10 as implied by R3
*>and R4.  Maybe this was an attempt to kludge around the fact that the
*>previous stage provides a signal that needs to be thresholded at 0V.  I
*>think there is a serious chance of hash on the output with large variations
*>in the video signal during the visible scan line phase.  Positive feedback
*>would have been the right thing here, but there's that problem of the input
*>signal already at the limits of the common mode range...

The circuit was tested and did not show problems with violent changes in
video level (black/white alternating fields from generator).

*>I would replace this whole message with a cap, two resistors, and a PNP
*>transistor, although the output would be inverted.

You can use whatever you want but this circuit was 'designed' by me
precisely to obtain a clean rail to rail output with usual components and
no adjustments (I already used the other half of the 393 in my circuit).
The output is rail to rail and clean regardless of the supply (3V to 12V
was tested) with the higher gain option. For some strange reason the
output was still ok at 1.4V. Hm, LM 393 is not specced to 1.4V. It is also
stable over temperature (the current mirror does that if Q1 and Q2 are
mounted together as they should be).

I understand that you analysed the circuit 'cold' or with SPICE. I did
some testing with SPICE and then built the circuit on a breadboard and
tested it.

The opamp need not be 393, with a 358 one obtains vertical sync pulses
(the 358 is not fast enough to follow the hsync), a 311 behaved much like
the 393, etc. The opamp must include ground in the input voltage range.

Peter

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads