I needed to remove some noise ontop of a digital

That's an overcomplicated version of the age old RC low pass where the R is paralleled by an antiparallel diode pair making the overall circuit high f requency cutoff for large signal and low frequency cutoff for small signal. Just guessing, but going on 60 years minimum it's been used.

The signal I am cleaning is isn't really digital, it's sort of looks like that at times.

It has a slow rise time, then flattens at some fixed analog level and when it shuts off, it drops to 0. This peak flat level varies depending on external conditions but remains constant once up there.

I needed any signal more than or close to ~ +/- .5 to over ride the charge at the gate for fast update.

I tried the double PN diode method but forward V (Vf) is too much, I tried Schottky, but the charge is too high in those. The bjt method works a treat.

What I am seeing atm is reverse Leakage(BE) on the bottom NPN and I need to scope that out to make sure it isn't zenering, not sure why the top NPN is not doing it, test signal is symmetrical.

I could use Ltspice I suppose but I live to do real live testing on something that I plan to use right away.

Jamie

R is paralleled by an antiparallel diode pair making the overall circuit hi gh frequency cutoff for large signal and low frequency cutoff for small sig nal. Just guessing, but going on 60 years minimum it's been used.

Nothing like putting digital specs on an analog prototype to guarantee fail ure. You do understand there is an analog memory element being used to make the determination of the +/- 0.5V that is erased once you let the gate cap acitance charge?

My comment was it's digital in nature but the amplitude level varies when it's in the (1) state! This means I was looking for way to filter the small AC that was sitting there but not inhibit the ability for a fast rise and full out side this window, which is what I needed.

I am now adding this to the PCB as part of a conversion interface application to something much bigger.

It works perfectly, I get the fast squaresh looking pulse with variable amplitude on the top with clean steady state signal once on.

Jamie

So, your post was not a request for help but for recognition as a circuit designer and, perhaps, a request for approval (pat-on-the-back) from the professionals here? You poor, lonely soul.

No, my post was about the commonality of using the circuit, idiot.

I was curious as to how popular it was among those that actually understand it's working and maybe side effects.

It seems no one stepped up to the plate with anything definitive not use it. Blogs did lay comment but I don't know how far I would take his opinion on it, he maybe a little like Phil, but he isn't in the dark like many here, he just has a hard time retracting.

--- putting your idiot comments aside ----------

I still do detect an abnormality with reverse leakage (maybe avalanche) on the bottom NPN via the BE via dt probes. I also stated that but obviously no one cared or thought it was an issue.

I don't need a pat on the back, I already know what I can do, I don't need to prove myself. I do however like to get comments from those I feel could offer up something tangible, you don't fall on that list. And when I do fall back in the woods, I have no problem asking and if that fails, I'll hit my book shelves. Lately google hasn't been very forthcoming.

Jamie.

On a sunny day (Sat, 31 May 2014 16:47:32 -0400) it happened "Maynard A. Philbrook Jr." wrote in :

What is wrong with the good old hysteresis circuit: ------- R2 ------ | | in ---- R1 ----- + | opamp out --------- out Vref --- -

?

That would be great, but as I said later to clarify it better, the signal is digital in nature but not fixed. when on, it'll stay at a fixed level and then turn off to 0. The problem is this ON level isn't fixed! The on state level maybe at different levels each time it comes on, but stay there once on.

So this has to be something that tracks and I needed to remove the other signals that are sitting on top of this signal..

Jamie

On a sunny day (Mon, 2 Jun 2014 20:56:09 -0400) it happened "Maynard A. Philbrook Jr." wrote in :

OK, untested:

------- R2 ------ | | in -------------- R1 ----- + | | opamp out --------- out | ------- - |--|>|----| | | --|

A variety of ways...

...Jim Thompson

failure. You do understand there is an analog memory element being used to make the determination of the +/- 0.5V that is erased once you let the gate capacitance charge?

If the circuit is digital then you fix amplitude noise feedthrough with...w ait for it...saturation of the output stage. This seems to have held digita l electronics technology in good stead since its inception. Work on quietin g the voltage source against which the output is saturated rather than nonl inear filtering of the input signal.

(shdjlwwumj=========__1401842717)

First off, your gif is not what you posted in ASCII above.. you are using 2 zeners in series, but you showed 2 diodes back to back.

Secondly, I thought i made it clear, I had to reproduce out the output close to what I have on the input?

I put in 1.5V pulse, I expect t see about that on the output.. the same goes with using a 5V pulse, I want to see ~ 5V on the output.

The example shown was using a ref at the crittle level, 1 to 1.5 is the bottom end, any higher input produces a near perfect output.

My circuit is far from complicated and I used 2 npns in place of the diodes that drive the fet for the actual circuit that i used. Transistors make better diodes in this case.

I've been reading this thread and not gotten a lot out of it. People have been all over the map in their opinions of what this circuit will do including the OP. I certainly am no expert in analyzing unusual circuits, but I think this one is not so complex.

I would say this is an source follower preceded by a couple of back to back diodes. The diodes combined with the high input impedance of the gate result in something similar to hysteresis but is also a peak detect. While hysteresis is usually a feature of digital functions, in this case the diodes are biased off until the input signal is outside of a range of the gate voltage making hysteresis. But meanwhile the gate retains the previous peak voltage until that hysteresis range is crossed. This capability should work in both directions.

So I'm not sure it will do exactly what the OP wants. If there is truly noise on the input signal, the output will show the initial value of the "digital" portion of the input signal, but then each time the superimposed noise exceeds the previous value the output will jump. So in a "real" signal I would expect to see something like a staircase waveform superimposed on the digital signal at the output.

In any event the output will reflect the sum of the digital signal and the "small" signal superimposed, not the digital signal alone.

Bingo, finally some one gets it... :)

The output signal does not have to be perfectly flat, but we can't have higher freq ripple on this signal and the way the system works it's actually better to have the on state with a slight ramp and this just happens to work out in my favor. I used transistors instead of diodes on the gate because I have found in some cases they make better diodes. from what I can see I am getting less leakage.

I have altered the circuit with a op-amp (non-inverting) and a couple of diode feed back to the (-) to drive the gate transistors. I did this to make up for loss over the diodes and V(th) voltages. We are trying to maintain an average gain 1 starting from 1.5 volts up to 15volts.

The signal i am conditioning is a form or staircase pulses.

Jamie

I seem to recall from reading about a crystal radio that the writer found transistors connected just the way you are using them to be more sensitive than the diodes usually used for such sets. I'm not sure that means lower leakage necessarily. I expect a lower forward voltage to be what determines the sensitivity of a crystal radio. Anyone know for sure?

I also read a page where someone was testing various materials including razorblades. I think they use a carbon brush from a motor as the "whisker". Continuing in that vein they tried a hack saw blade which worked even better (once the paint was removed).