DESIGNING EXTERNAL RETRACE-BLANKING

OK, I realize that this probably isn't terribly interesting or exciting, to most of you. BUT, just in CASE someone might actually be interested, at least someday, here's a Progress Report::::

---------

I have reworked the External Oscilloscope Blanking-Pulse Generator design (for an X-Y-Mode (XY) scope's "External Z-Axis Input") (an add-on for my existing Curve Tracer, for use only when the Sawtooth Ramp sweep-waveform-type is selected, i.e. not when either the Sine or Triangle Ramp Device-Under-Test (DUT) Sweep Waveform type is selected).

And... I made significant improvements in performance and robustness, to the point where I THINK it might now be "Not Crap", and, dare I say it(?!), MAY even be "Good Enough", OR, EVEN, possibly, (gasp!!)... "Good" !!

(But OH how I more than half-expect to be "corrected", about that! ;-) )

I decided to just "bite the bullet" and "add another opamp", to boost the differentiator's output (by 101x), before sending it to the comparator, which did the main trick.

I also "corrected" the differentiator's component-value ratios, ditched the weird ac-coupling attempt, and used much-faster opamps. And, I inserted V-sources to simulate worst-case offset-voltages, for temperatures from 0 to 100C, and then set resistances to make sure that the comparator's input and output would stay below the proper thresholds, for all expected sweep/input, temp, and offset conditions, except for negatively-sloped sweep-voltage portions, which are where the blanking of the scope's trace is desired.

The new blanking-circuit design's LT-Spice files are (or, will soon be) at:

[Note: No other user-created files are needed, to run the simulation with the files given above. You would Open the "blankng2.asc" file, and run it. The "blankng2.plt" file would then be automatically used (just to set up the "Plot Settings" that I was using, for the waveforms' display window).]

I will probably try to change at least the first (buffer) opamp (U1) from an LT1364 back to an OP275 (or maybe something else, actually), to save $.

I am also seriously considering just using a JFET and a couple of resistors, as the input buffer, instead of an opamp, especially if it will save board real estate.

And... I would *LIKE* to find a widely-available opamp that has a slew-rate that's SOMEWHAT-similar to the LT1364's 1000V/uS (in a Dual 8-DIP), except cheaper (by a LOT, I hope). And, I'm guessing, a significantly-slower opamp would probably work just about as well, anyway (and maybe better, especially considering the 1364's offset voltage drift versus temperature). (Besides, I just "threw in" the LT1364, because I wanted to try a faster opamp and a) I have some on hand and they are used elswhere in this unit, and, maybe mainly in this case, b) just because it was in the built-in LT-Spice component library and I was in a bit of a hurry.)

That reminds me: I'd also like to find a cheap, widely-used comparator, in

8-DIP or smaller (through-hole only), to use in place of the LT1011 comparator, which I used in this design and simulation mainly just because it was already available in the built-in LT-Spice component library (and, because I was *hoping* that its performance would be quite similar to that of many "run-of-the-mill" comparators that I might eventually specify for this circuit).

Thanks for wading-through all of that! SOrry to have blathered-on for so long!

If anyone has ANY opinions or comments, suggestions, ideas, etc, about anything related to this circuit, I would REALLY appreciate hearing them!!

Thanks, very much.

(On your marks.... Get set... GO!!) :-)

Regards,

Tom

tomg at fullnet.com

-------------------------------------------

Well, I don't have LTspice, so I haven't seen a schematic, but from your description, it sounds like you're trying to derive a blanking pulse from your drive signals. Why not get the blanking earlier in the circuit?

Good Luck! Rich

----------------------------------

Rich,

Thanks much, for the response.

There isn't much of anything upstream that I could use, except for the ramp generator, itself. And I definitely COULD use the pulse from the opamp/comparator, there. It's almost exactly what I need, already. But I'm trying to come up with a mod for the existing (crowded) pcb(s), and I think that that pulse would be extremely-difficult to route to where I need it, i.e. without *major* re-work. But I'll take another look at it, since, exept for the routing issue, that seems like it would almost certainly be the best way to do it, and would probably then only require a single comparator and a few resistors and caps.

Meanwhile, I took another quick look at the circuit I was working on, which derives the retrace-blanking pulse from the sawtooth ramp itself. And, for those who don't have LT-Spice, I made JPEGS, which are at:

(the circuit) (plots, 60 Hz) (plots, 22 kHz)

The new LT-Spice files are at:

I saved an opamp by using a 2N7000 and one resistor as the input buffer (Output offset is kinda large. But all I need are the slopes, anyway.). And I got rid of the negative output spikes by connecting the comparator's negative power supply pin to ground, instead of -V, which also eliminated a diode, a resistor, and a decoupling cap.

So, anyway, the circuit now needs TWO dual opamps. (Bummer.)

One opamp is used as the differentiator. Two are used as two stages of amplification of the differentiator's output. And I came up with what I think is a very good use for the fourth one:

I used the fourth opamp, and a low-pass filter, to force the comparator's NEG input (which was formerly tied to ground) to TRACK the signal going to the comparator's POS input, but with a small positive OFFSET added to it; about 1mV minimum (added mostly by the opamp itself, which also buffers the signal). So, now, when an 11mV pulse arrives at the pos comparator input, the low-pass prevents the neg input from following fast-enough and the comparator is triggered on. With that setup, the circuit should be more robust for handling the small pulses produced by the differentiator, for the lower-frequency sweeps. It also helps in minimizing the response-time/delay, since the comparator's inputs are held closer together while waiting for a pulse. (I also added a series RC in parallel with the LP's R, to help out with the higher-freq response).

It seems to work fairly well, in LT-Spice, even with worst-case offset tempcos and from 0c to 100c. But, for a 6.7us sawtooth fall-off period, there's still about a 250-300ns time-lag, after the tip of the sawtooth, before the blanking pulse reaches 5V. That's for a 22kHz sweep, which, unlike the 60 Hz case, has a much larger idling offset between the comparator's inputs (188mV; and the ways I've tried, that lowered it, tended to make the 60 Hz case break when the tempcos were near worst case. I also then left a safety margin). So: about 60ns of that delay is from waiting for the LT1011 comparator's + input to rise to the level of the - input. Then for some reason, as the difference between the POS and NEG inputs continues to grow, it still takes another 150 ns or so before the comparator's output even BEGINS to change (by which time the comparator's inputs are 2.2v apart). And then it takes another 65ns or so for the comparator's output to slew up to 5V. I keep reading about things like "7ns Comparator". I guess the LT1011 must be one of the older, slower comparators.

P.S. Rich, LT-Spice is "too cool"! If you haven't tried it, you're missing out. It's a free download (less than a meg when I got it), from

. I usually fight the urge to tell people what they "ought" to do. But you ought to download it!

Regards,

Tom Gootee

------------------

Here's an idea for you...

you can effectivly blank the retrace by using the Y (verticle) instead of the Z input.... you just apply a large enough voltage to the Y input to deflect the beam all the way up or all the way down... it may not be perfect blanking but it does work and it is much easier compared to connecting to the Z input of most scopes... Mark

Mark,

Thanks. I can see how that might work.

But, in this PARTICULAR case, I don't think it would work quite well-enough. I'm using the scope in X-Y mode, with the X-Y inputs coming from my curve tracer. So, the beam, in general, isn't doing repetitive left-to-right sweeps all the way across the screen, and basically never goes past the edge of the visible part of the CRT.

For example, for each "flight" of steps of the staircase waveform applied to a transistor-under-test's base or gate, the beam would sweep from somewhere on the screen to somewhere else on the screen, for each step. And depending on the switch settings on the curve tracer, the horizontal input might sweep mostly right to left, or mostly left to right, and "retrace", to a different starting point, after each stairstep. And the vertical can also go in either direction, similarly. So, in general, each "sweep"'s trace can start anywhere and end anywhere.

Your idea could work very well if every sweep went off the right edge and was supposed to re-enter from the left edge after retrace, for example. It also would probably work acceptably, even with the beam staying in the viewable area, if the scope's bandwidth was low compared to the speed at which the beam was pushed off the top or bottom of the screen.

In my case, it actually has BEEN working, acceptably, so far, ALMOST like that. I didn't even force the beam to the top or bottom. The "retrace" speed was fast-enough that the lines were usually very dim, to the point where they were easy-enough to ignore.

But now I want to get rid of them, *completely*.

Anyway, I'm not sure why you think that deflecting the beam out of the viewable area is "much easier" than just using the External Z-Axis input to blank it. For one thing, in my case at least, I'd STILL have to have the circuit that knew WHEN to do that, and it would then have to be able to provide a larger pulse than it would for the Z-axis input (and different scope attenuator settings would affect its performance, then, too). And I'd then ALSO have to somehow SUM (or at least switch) that larger, fast deflection-voltage pulse with my normal signal voltage that's going to the scope (which is basically directly connected, for both X and Y scope channels, to the output of a high-precision instrumentation or differential-type output-amplifier subsystem).

That sounds like at least a little more trouble than simply providing another BNC connector (which is actually already on these units), and a cable, to go to the scope's External Z-Axis input, and providing a 5 to

10 volt pulse to that completely-separate BNC.

Well, OK. Maybe it wouldn't be much harder, re-design-wise, your way. And it might also be cheaper, in production. But, since I already have a BNC output available (formerly designated for a 500ns 5V "sync" pulse output, which occurs at where the leading edge of the new blanking pulse WILL be, but which had "no specific intended purpose"), and since I really want to totally BLANK the beam, during the fall-off portion of the sawtooth sweep, I think that I'll have to stick with providing a true blanking pulse, for use with the External Z-Axis input.

Thanks again.

Regards,

Tom Gootee

tomg :at: fullnet.com

-------------------------------------------

You're not giving us a lot to go on here. The 400:1 in ramp-up rate is not so important as the ramp-down rate. Is the ramp-down always on the order of 5us regardless of ramp-up or is it proportional, say 10%? And what is the quiescent DC of ramp drive when it's not ramping? You should be able to improve the differentiator gain by a factor of x100. Then you can get away with a single dual 8-pin part. For low cost, quite a few of the precision LT parts are available from TI at 30% the cost but with less precision and performance that may not be important.

Fred,

Thanks for responding! Sorry to have not provided enough information.

The ramp-down is proportional to the ramp-up. The proportion increases slightly with frequency, but is always in the 12% to 15% range.

I would love to get the gain of the differentiator up by 100x. I might not even need the comparator, then, as I think you mentioned, or alluded to (or else I misunderstood). But I don't know how to do that, especially for the lower frequencies, without slowing its response so much that the resulting pulse's 5-volt crossings are no longer synchronized well-enough with the beginning and end of the ramp fall-off, at every frequency of interest. I'd probably be happy-enough if they were always within 200ns of being synchronized.

Quiescent DC of ramp drive when it's not ramping? It's always ramping. But it does suffer from what seems to be a typical "problem" with these sorts of oscillators, i.e. as the frequency is increased, the output amplitude grows. This one's amplitude grows more for the negative swings than the positive ones, too. e.g. It ramps from -3v to

+3v at 60 Hz, but from about -6.8v to about +3.3v at 22 kHz. (That's taken care of, downstream, where the amplitude and offset are servo'd to precise desired values; "user-selected" for the amplitude and "zero" for the offset.)

Thanks for the tip about the TI parts, too!

Regards,

Tom

--------------