opamp clamp

Merry Christmas!

...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 | |

Would that be clean at 100 KHz? U2 and U3 will wind up hard.

Thanks to you and Speff, but I really need the clamp to be in the feedback loop of an inverting amp. Nico's suggestion of finding an amp that inherently clips clean (MC33503?) from 0/5 rails is interesting, but I think it would also need an input clamp to keep the summing point from winding up, which wouldn't be bad. I'll try that.

Save room for dessert! We have reserved seats for the Sweeney Todd movie... just before dinner! At least we aren't having meat pies.

John

That little detail is left as an exercise for the student ;-)

What does 100KHz mean? Sine or square?

I was thinking of seeing "Charlie Wilson's War"

We had our Christmas "Eve" celebration on Sunday so that the individual families could do their thing at home.

So we're done. Whew!

...Jim Thompson

Could be anything; square wave or pulse from a conditioned fuel-flow sensor; direct signal from a variable-reluctance pickup, mv to 100 volts as speed changes; 120 vac from an alternator. 100 KHz is probably overkill; a APU turbine is about the fastest thing we'll ever see, maybe 45 grand, with a pickup on a gear with 32 teeth... 24 KHz tops.

Well, I got...

A truly awesome stainless-steel potato masher

A jug containing 25 assorted bungee cords

Two Donna Leon mysteries, about the Venetian detective, Commissario Guido Brunetti.

A custom-thrown-for-me, custom-fired ceramic bowl with flat bottom and steep sides, specifically for making whipped cream.

The latest volume in the Tales of the City saga.

A little butane fire-starter wand, to be used only to ignite Bananas Foster. It fired up first click, 10 out of 10 tries, unusual for these gadgets.

Chocolate.

The Brat got an enormous Samsung LCD tv and a blu-ray player. Mo got a new car.

John

Well, I am not on salary with the parts manufacturers, so here is the clamp:

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

If you can afford to waste some current, Widlar's trick using BJT saturation is really pretty good at ~100 kHz speeds.

Vr 0--RRRR--\\ >----RR---+ ----- | | | | | +---RRRRRR--+ | | Vin | | | |\\ | 0----RRRR----+---|-\\ | | \\____| | / +----|+/ | |/ --- \\ / V

You pick the resistors so that the BJT saturates at the clamp voltage, so suddenly the emitter resistor appears in parallel with the feedback resistor. You get a nice sharp-edged change in gain--which is not exactly clamping, and is limited by the beta of the transistor, but you don't lose as much linearity at the edges of the swing.

Cheers,

Phil Hobbs

Aha! The real truth erupts ;-) You are only interested in frequency. So clipping/clamping can be any ol' thing, then run a "charge-pump" for the tachometer effect.

...Jim Thompson

snip Happy Santa stuff

Just done something similar for a biphase mark signal, on a single rail TLC272 witha phantom center ground. I used a LT1006 for the simulation, and a 5V rail

Virtual earth config, 680K Rfb

The i/p R was split into two, 10 K to signal source, followed by 2

The LTspice sim looks ok when signal in goes to a 100V. I didn't try it out, since my PC sound card doesn't give a 100V out (yet).

The only thing I wondered about was when a diode is hard on, does this actually increase the gainof the VE config to 680K/(220R+diode) ?

I tried putting a couple of zeners across the Rfb, straight from the LT lib, but the HF sucked at low levels

Merry Christmas to the other side of the pond

martin

Analog? Heaven forbid. Downstream, there's switchable lowpass filters, ac/dc coupling, and an insertable integrator for conditioning velocity-type sensors. Then two comparators, each with its own threshold dac. Then into an rs flipflop in an fpga, giving us programmable trigger levels and hysteresis, then to a semi-slick period measurement algorithm.

The fpga timestamps rs flipflop rising edges and also counts edges since the last time it was polled. The uP checks the fpga once every millisecond. If no edge, skip; if one edge, subtract timestamps to get period; if multiple edges, subtract timestamps and divide. So above about 1 KHz, it automagically begins to average multiple periods. Period data is available every edge, or every millisecond, whichever is slower.

John

So why the clip/clamp?

You haven't fallen into the "do everything with a uP" crowd have you ?:-)

...Jim Thompson

No, we let fpga's do the heavy lifting, and uP's mostly just move the results around. Need a half-dozen 32-bit multipliers and a dds sinewave synthesizer? No problem, they're free.

One of my guys wants to digitize the raw input waveform and do *all* the subsequent processing in the fpga. I'm not sure I'm ready for that.

Hey, this ain't bad:

ftp://66.117.156.8/Clipper.jpg

Worked first try, as calculated!

John

I like Spehro's version better ;-)

...Jim Thompson

But it's not inside the opamp loop.

John

The virtues of a clamp circuit are the good linearity and the accurate edges. This one is neither very linear nor very accurate.

VLV

Depends on what you want. This one is very much sharper than a diode, and can be much more accurate too, depending on the voltage drop across the emitter resistor.

The OP specifically said that the voltage range was important, but the waveform much less so. Engineering is about making things that work for the intended use, no?

Merry Christmas to all.

Cheers,

Phil Hobbs

What's your fixation on putting that clamp in the feedback of a relatively high voltage OA? Why not something real simple: View in a fixed-width font such as Courier.

. . . . . +5V . | . | . .--[100k]-|--. . | | | . | | | . | |\\| | . sig >-[100k]--+--+-------|-\\ | . | | | >-+->

. - v .-|+/ . ^ - | |/| RRO . | | [51k] | . | | | | . | | | | . '--+-----+---+ . | . --------------------+----- . com . .

Only to keep the opamp running closed-loop. Open-loop recovery is seldom specified.

When the signal goes positive, the opamp output will saturate at ground and go open-loop. The summing point will charge to +0.6 and internal opamp nodes will wind up. How it recovers from that depends on the amp, but most opamps will take a while to get linear again... milliseconds to seconds even in some cases. There's a similar problem when the input swings a bunch negative.

Capacitance at the -input node can be a problem too, since it charges as much as +-1 junction drop when the opamp rails.

Are there any r-r opamps that are known to recover gracefully, and quickly, from being railed?

John

A REAL specification would be helpful.

From some of your postings it looks like you want to deal with really small signals as well as ones that would over-drive???

...Jim Thompson

John,

Aren't you trying to be overly complicated? (Your usual method ;-)

Why not...

...Jim Thompson