My designs run towards the elegantly simple. But too simple may invoke constraints that turn into nightmares later; you've got to leave enough knobs to turn.
Because I want to do downstream filtering and integration, and need a roughly 5-volt signal range so I can program the final comparator thresholds with cheap multichannel unipolar serial dacs.
Something sort of like this:
ftp://66.117.156.8/Tach1.JPG
John
Any offset issues?
...Jim Thompson
Not too serious. The integrator is ac coupled at its front end, a few Hz, so it's ok. The dacs are 8 bits, good enough to set trigger thresholds for what is a fairly gross signal anyhow.
Tachs can get interesting. A variable-reluctance pickup has voltage proportional to frequency (at least until cable capacitance kicks in) so needs integration to work reliably over a wide range; some peak at
100 volts or more. Some signal-conditioned pickups are constant-duty-cycle pulses, some constant-width. An alternator could be a few volts to 140 RMS. Everything has to clip/clamp/recover cleanly.Gears have non-uniform tooth clearances and shaft runout. Some people use a "short tooth" setup where one rotation will make, say, 15 nice pulses and one runt/missing pulse per revolution.
John
I just tried an MC33503, a r-r cmos opamp whose datasheet claims all sorts of internal anti-sat and clamp stuff. In an inverting (g -1) configuration, it takes several microseconds after an overload for the output to break loose from ground, and then it jumps and rings a bit. It's ugly at 30 KHz, and at 100 KHz or so, if it hits the rails it pretty much quits working.
Clamping the summing point with a schottky doesn't help.
John
I think r-r in the datasheets simply means no phase inversion occuring.
- Henry
I like that one the most.
That's not a representative cmos opamp. ON Semi has made all sorts of special efforts to get this MOS- input opamp to operate rail-rail at its inputs with a miserly 1V supply. That's without two input stages, one each low side and high side, as is usually done. Why not treasured and preserve the mc33503 for its special capabilities.
As for snipped-for-privacy@sushi.com's claim that CMOS opamps tend not to "stick" by design, I dunno. Perhaps he can give us chapter and verse. Let's see names named!
You'd think that if people make single-supply r-r opamps whose common-mode range includes ground, or even below ground, they might address what happens when the outputs slam into ground, and how/when they recover.
The 33503 does show explicit anti-sat and clamp stuff in its "schematic", so I figured it was worth a try.
Oh well, things are slow this time of year, so maybe I'll fool with this some more.
John
Oh, I only know of what I have designed. I know anything with bipolars in the output stage tends to stick. It is a matter of base charge storage.
Then again, it seems to me an op amp of a few MHz GBW that takes a uS to recover is doing pretty good.
Just so happens I have a few CMOS op amps right beside me... Here's some numbers maybe to try...
LMV822 EL5120 OPA356
D from BC
I don't know the condition of your test, but it's easy to imagine a case of integrator runup. Note how the antisat circuit is on the stage _after_ the internal capacitor node.
Consider, a BJT opamp has a base to absorb the continuing error currents from the input diff-stage, during output saturation. OTOH, amplifiers like the mc33503, with a MOSFET gate responding to the equalization-capacitor node, suffer from that node charging to a voltage far from its normal operating voltage, during output saturation, etc. If the mc33503 designers had returned the anti-sat control to the capacitor node, the problem could have been avoided.
The high SR stuff in CMOS should do fine. You can also do a bi-level clamp keeping the amps closed loop but it's a mess of parts and we know how you don't like that. You say the tach output hits 100KHz, and these are pulses, so I assume you need 500KHz analog BW minimum.
No such animal. The "sticking" is due to overcharging the compensation capacitor, and the recovery time thereof.
...Jim Thompson
The more I ponder this problem it is clear that Spehro's approach is best... NOT in a feedback loop, NOT subject to ringing, NO problems with bandwidth.
...Jim Thompson
-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |
And it WONT work in my circuit topology.
Hey, this is cute:
ftp://66.117.156.8/Clamp3.JPG
but it has some problems.
John
And WHY won't it?
You're not clever enough to center it up, or is it just NIH ?:-)
...Jim Thompson
Well, try some! A simple inverting amp, 50K Ri and Rf, opamp powered from +5 and ground, ni input grounded. Drive it hard with a sinewave and see what happens.
John
How about...
Analog expander/5V digital clipper ---------------------------------- opto | +----/ \\---+ | | | | | | |
----R---+---|\\ | // | \\----+-----[fill in the blank ]--->|---+ com +---|/ LED | Com
I haven't thought about what goes in the blank yet..nor thought about the stability.
D from BC
Well....I only have the SMD versions :P And I don't have a break out pcb board. But...I could make a break out pcb in less than 1 hour... But...that's not fun..
Would be nice to have a networked DSO to post the oscillographs. :) But don't have one :(
D from BC
Because I need front-end gain and common-mode rejection, and clamping downstream of all that won't keep that stuff from winding up.
And I want to clamp a couple of hundred millivolts inside the rails, so the following stages stay linear.
John
How much?
How much is "a couple of hundred millivolts" ??
...Jim Thompson
Have something to add? Share your thoughts — no account required.
Ask the community — no account required