schmitt buffers

Oops, typo. But was only 60dB off ... well, then there's super-caps but they may not last forever and they cost a bundle. 10F should get you to around 25uV :-)

Yeehaw!

Every time I do things like that I toss and turn around in bed the next night, thinking "there has to be a method that costs 5c less". You could take a quad comparator and rent out the other three sections. Problem is, you've got to find renters for those.

Have you considered a cmos comparator in a SO package?

On a quick look.

That one has push/pull outputs and only 200ua shoot through current from what I can see.

Just a thought.

Jamie

More like a few hundred ohms. But it is a nifty part and I sure wish they had released something similar in the 74HC and 74AUP series. Problem is, by the time AUP came out hordes of good analog engineers began heading towards assisted living, so no market :-(

LVDS do, but I doubt you'd be happy with their drive capabilities. Also, they require strict RF layout guidelines to be followed. Else they can dance a tarantella.

Nice part. But I am afraid it'll soil your rail as well when you are inside the +/-100mV range.

ramps.

Yeah thanks for that.

I'll put those right next to the 10H ferrite beads :)

I didn't say but that is a heavily signal-averaged display. But it brings out any synchronous component beautifully, which is what I am interested in here.

I was *literally* doing this last night, sad git that I am. And I don't even have a real application!

Note to self: must get out more...

ramps.

No! Don't do that. Stay in your lab where it is safe and you are surrounded by your instruments. They'll keep you warm and comfy.

Hi Jamie,

Yes something like that might work. I tried LMV7239 already early on (before the 74LVC1G17). But I found their saturation voltage was too high for this. I think the key may be a part that is micropower but nevertheless has relatively high current outputs. Perhaps like this one.

Thanks,

John

Adding a small cap fom the output back to the input and a series resistor from the input to the source will speed up the transition time thru the linear region. Art

Hi Art,

Don't see how that can work here I'm afraid. The linear region is pretty big. Worse, my circuit has to keep it there "by definition" (i.e. within the upper and lower switching points).

Thanks,

John

Since I couldn't find the shoot through/cross over current in the spec or I didn't look hard enough, I ran LT's test fixture on it and as close as I can see, it looks like ~ 400ua. Not bad.

Jamie

A year ago I went through HELL trying to figure out why a circuit using a number of 74LVC parts was misbehaving. I finally discovered there was HUGE shoot-through current during transitions. I estimated the current at about 2 A for about 2 ns. I replaced all the components with

Now, my problem seemed to be more related to the output stage than the input, but tweaks to the transistor thresholds might benefit both. The new circuit runs on 3.3 V, the old one ran on 5 V.

Anyway, the AUP family has dynamic power consumption ratings that are, IIRC, 20 times lower than the LVC.

Jon

Yep. My poor-man's approach: 2@ 74HC04 in sequence, cap from last output to first input, add the series R and you have serious "snap". Also makes a nice delay ;-) ...Jim Thompson

That will do nothing to reduce the front-end shoot-through current as the input voltage approaches the snap point, which was the OP's issue. It behaves, DC-wise, like an HC14 sort of schmitt with a lot more shoot-through current. And may oscillate on the transitions.

It *will* slam the front-end ESD diodes and make the output transitions look funny.

Adding external hysteresis to a gate or a comparator can have all sorts of unexpected consequences.

John

for

em?

s.

or

linear

put to

a nice

=A0 ...Jim Thompson

The more recent (recent as in last 15 years!) battery voltage monitor comparators cured this problem quite well- and specifically compare the *cross-conduction* current near threshold to the '14 type schmitts showing 1000:1 improvement. As you might expect, the voltage monitor is the most stressing application since the input can hover arbitrarily close to threshold for ages. A cascade of one of these with his supa-dupa buffer would satisfy any reasonable performance spec.

If you RC-decouple the comparator's supply from the rest the non-averaged plot should look pretty good as well.

Might be time for an eye on town and a nice pint of McEwan's Heavy :-)

Careful, a lot in their models contains behavioral equivalents of circuitry. You can't necessarily deduce supply spikes from those models.

This is not to say behavioral models are bad. They do allow speedy simulation of DC-DC converters, for example. They are just not so suitable for the task at hand.

How ?

While ago I tested static cross conduction of HCT04 gate powered at +5V. The worst case was about 4mA, at 0.9V at the input.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Don't listen to them, just do it :)

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

From the output stage. They have to cut it close with shoot-through avoidance. Also, some capacitance needs to be muscled around. It's usually shorter than 2nsec though but depends on trace length, measurements setup and so on.

That's why the non-inverting buffer and the two resistors won't work well in John's case. Because the cross conduction is going to be of similar magnitude.

in the datasheet for comparator I linked to, it says the it has a few tens of nanosecond break before make on the output "to maintain micropower operation"

-Lasse