Matched diodes

The diode curves will be soft enough that clamping a 10 mv signal is pretty much meaningless.

John

Whats the bandwidth? Have you tried any active rectifiers (i.e. with the op amp absolute value circuits)? Even there to get any accuracy you'll need some fairly "precision" opamps - i.e. low offset voltage.

HTH...

Not sufficient, the matching only happens at equal currents. Would there be any problem if you just AC-coupled the signal with a +10mV bias?

Diode matching isn't tested in most mass-production, but is likely within 5 mV. It might change (significantly) when soldering is complete, or if there are package stresses. Metal-can parts are best, and non-surface-mount is preferred, if you really want matching.

You don't say if this 10mv is sub level of a larger DC, nor do you say if it's DC or AC ?

So for the moment, I'll assume it's a 10mv DC base line that you want to squelch out

Pass the 10mv into something like an op-amp first stage pre amp.

A gain of 10 would then bring this signal to 100mv. The output would then be a reference to use on a comparator which can be used to switch something on/off..

A comparator, being that the output on most that i've seen are open collector, will let you easily integrate this into a analog switch.. For example, a T bar Diode switch can be used where the comparator output (via a R load) would supply the center bias for the switch via

Freq response time, max input and things like that has to be known..

P.S.

On the preamp stage, you wrap the loop back gain with a zener to avoid over drive on the input.

It's seems to be a trivial problem, but maybe there are some constraints you are not mentioning here?

Jamie.

Thanks folks for the responses.

It's a dc input, which has to be measured down to 10mV, to an MCU successive approximation AtoD. However the signal can go negative by a diode drop, say -0.6V. The input has a schottky diode holding it to

-0.2V.

However the MCU gets the measurments wrong when the input goes to

-0.2V. I don't know the mechanism and at the moment I don't know if it is this input or another input that is screwing up.

I do know the MCU makes the wrong decisions. If the input is clamped to 0V when at -0.2V everything works OK. The board is in China so I was looking for a cludge they could try to prove the point.

I didn't think matched diodes would work but it was worth asking the question. John gave me an answer that told me to look at other solutions. I've gone for a 2 transistor circuit driving a FET to clamp the pin when the input goes negative, and removed the schottky.

I'll know by breakfast whether that has worked.

I am curious how -0.2V can screw the Samsung MCU AtoD. It's not enough to forward bias the catch diodes but if it got into the substrate it could change the voltage range. I wondered if the comparator in the AtoD was doing a phase reversal due to the negative input but it seems unlikely.

It's an interesting challange sorting out the Chinese, it's a shame it has to be done under time pressure.

Source impedance? ...Jim Thompson

You could use an open collector comparator to clamp the input to ground when pulled below zero. ...Jim Thompson

I suppose it's possible that it's outside the common mode range of the comparator, but that seems unlikely.

Are you sinking a lot of current with the micro? Has it been confirmed by measurement that the input is actually at -200mV or are you surmising that it "must be"?

Best regards, Spehro Pefhany

You need to elevate your ADC input with an offset!

It's common practice to do this, You will of course loose some dynamic range when doing this. If you bring the input up to lets say .5 volts, and in the software you account for this offset, you'll be fine.

Jamie

Right. In other words, an active clamp.

John

How about a 5V op amp running with a gain of 500 followed by a 500:1 voltage divider?

Cheers

Phil Hobbs

It worked.

The micro spec says 2 Meg input impedance.

The actual measurement was - 190mV.

SNIP

The MCU AtoD gets a positive reading when the input is -0.2V. The adjacent AtoD channel is unaffected. So a negative input voltage screws the successive approximation conversion.

Sure, I meant on other pins.. outputs.. if there is enough current flowing then you could have more than 200mV to the internal level of Vss.

Curious.

Best regards, Spehro Pefhany

Why diodes, just use a piece of wire, that will clamp to 0 volts every time.

Troll...