schmitt buffers

John Devereux a écrit :

The added current is mainly from the input stage and depends on the supply voltage WRT the input mosfets Vt. If you lower your supply rail it diminishes and there's a point where that added current will be negligible.

If you have space (and your noise margins allows) you can use several in a row, 2 or 3, with graded supply voltages. The first one will give provide you with low current and a fast step to drive the second one through the annoying zone fast enough.

For this nice gate figures seems to add up neatly, and it's also allowed to be driven over its supply rail which comes handy for your first gate low supply rail.

1.8V first stage and 3V second stage would seem to work OK.

--
Thanks,
Fred.

They call it "Additional Supply Current" or "Delta ICC", and for the

74AUP1G17 its typically 40uA cross conduction.

You might look at the NL17SV16XV5T2 or NC7SP04, in which the current is not specified (is it due to a better device or a worse datasheet, I don't know)

Regards

Klaus

Hi Klaus,

I assumed that was what it meant, though they never seem to spell it out. Looks like its 40uA *max*, so could be good. Only "4mA" outputs though.

Worse datasheet I expect, but will take a look.

Thanks,

--

John Devereux

That's interesting to know, thanks,

I'm already doing something a bit similar using an opamp as the "first stage" driving the output buffer. I was hoping to find the perfect jellybean part as an all-in-one solution.

It's a good idea, might be able to run the whole thing off 1.8V for some applications.

John

--

John Devereux

As Fred and Klaus wrote, it's cross conduction in the first stage. Unavoidable. Unless you only need the 10ohm outputs but not the speed, then you can hang another Schmitt from a slower series in front of the

74LVC. Maybe a 74HC part.

Another trick is to supply that separate "per-Schmitt" via an RC. Meaning there is a big capacitor of 1uF or more from VCC to ground and then a resistor from there to your rail. That'll take the edge off the supply spikes.

--
Regards, Joerg

http://www.analogconsultants.com/

But unfortunately won't work with slow inputs...

--
Thanks,
Fred.

Make-yourself schmitt of a non-inverting gate and two resistors ?

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Any cmos gate will have input stage cross-conduction if the inputs aren't close to one rail or the other. Schmitts tend to be a little lower than non-schmitts, because they are intended to operate midway between the rails.

We ran some NC7NZ34s at Vcc=5 and drove their inputs from a 3.3 volt FPGA. When the inputs were high, they got very hot.

You could use slower parts, which tend to have smaller fets inside and won't get as hot. Or precede the cmos parts with a low-power comparator, to slam them through the linear region.

You can make a comparator with hysteresis from two low-power comparators and a cmos RS flipflop.

Maybe a micropower cmos opamp would work.

John

I don't really need the speed. I hardly care at all about the propagation delay (few hz circuit), and anything below ~1us output transition time would be OK. But I *am* interested in a low resistance output (though not high current as such).

I tried some high output current CMOS opamps but they don't saturate as well. I think their static current drops too much voltage at their bond wires or something...

It's not the fast output transition spikes that are the problem. I have a slow ramp feeding the input, and it is the synchronous supply current changes that are are the problem. They are not spikes, they are slow ramps.

Too slow to filter with an RC.

Trace 3 shows the supply current variation, trace 4 shows the effect on the supply voltage. (Hmm, might have got the current probe the wrong way round there...).

--

John Devereux

Hi Vladimir,

Hmm, from other comments I get the impression this would still have the same problem.

--

John Devereux

Why not? Cross currents can be much more moderate with older technology Schmitt parts. That's basically how my frequency counter does period measurements. You can hang a millihertz sine wave to the input and it'll show you the frequency or period.

--
Regards, Joerg

http://www.analogconsultants.com/

If you want ultra-low Rdson in the outputs how about this puppy?

Those aren't too great for a digital output. If you want the driver from hell that can muscle a stiff load you'll need a gate driver like the one above. The cross currents of those are pretty low but you can always hang a pre-Schmitt in front of it in order to dodge the cross current.

Nah, old American saying: If the capacitor ain't big enough then use a bigger one :-)

If you let us know the requirements there may be other options. For example a shunt regulator that burns off some power to even out the surges. Kind of like driving on ice with the hand brake slightly pulled for a "poor man's" differential lock.

So how many milliamps per division does that represent?

A few options: You could try a CD40106 as a "pre-Schmitt", don't remember its cross current but it was miniscule. Or you could hang a regular comparator up front and add a 3rd resistor for hysteresis. Then you can even set the amount of hysteresis yourself. This stuff would go before your big driver.

--
Regards, Joerg

http://www.analogconsultants.com/

Use a real comparator instead, and then follow that with your big buffer.

--
Regards, Joerg

http://www.analogconsultants.com/

how about something like this:

-Lasse

Yes I was looking at mosfet drivers.

Well I already put on a 1000uF one to test that theory. Turns out even a just a few mA ripple does not get filtered out at all, down at a few hz. It was counterintuitive to me. dV = dt. I/C = 50ms . 5mA / 1000uF =

0.25V.

Really I am just playing to learn here, although there is always the possibility of real-world application somewhere.

Like a TL431 (rather than my 3.3V LDO)?

It's 10mA per division, like it says. What, your scope doesn't label the current probe trace? :) I suppose that photo could have been better.

I already got it working well with a comparator made from a CMOS opamp as a "predriver". I posted because the 74LVC1G17 on its own is otherwise so nice, sort of a zero-cost perfect jellybean building block. It seems a shame to have to add a load of extra cruft.

--

John Devereux

We have some ceramic mil spec 54AC14DMQB up on eBay for less than a dollar each,

Not sure what their behavior is, but the military paid big bucks for them. Supposedly they will source or sink 50 ma.

Data at <

>

About 200 remaining. US sales only.

--
Many thanks,

Don Lancaster                          voice phone: (928)428-4073
Synergetics   3860 West First Street   Box 809 Thatcher, AZ 85552
rss: http://www.tinaja.com/whtnu.xml   email: don@tinaja.com

Please visit my GURU's LAIR web site at http://www.tinaja.com

[...]

Then a 10,000uF cap gets you to 25uV :-)

Yup. There's also the LMV431 and various TLV431 (but watch max cathode voltage on those). The low power versions can live with around 100uA of cathode current so you won't have to waste so much. If your max ripple current is, say, 500uA you can set it so the shunt regulator "eats"

800uA or so when nothing is switching. That gives it enough headroom. The when the switching begins and the voltage wants to sag it'll ease off and evens it out.

Mine usually doesn't either. I think there is a way of sending such stuff to the display via the computer, in a dozen languages, but I keep forgetting how because I never use that remote control scheme.

Well, LMV431 would reduce the effort :-)

--
Regards, Joerg

http://www.analogconsultants.com/

CMOS logic doesn't include any real resistors, so current limiting is a weak point. There are three solutions you might employ.

First, the old CD4007 is a kind of CMOS construction kit part: you can add resistances and build your own Schmitt trigger with CMOS inputs and outputs, and any current specification you want.

Second, the amount of current depends on conductance and bias voltage, either low voltage power supplies or older (lower-drive) logic families, or both, will improve matters.

And third, you can go to a logic family (or amplifier family) that has better power modulation characteristics: differential ECL receivers might be excellent, for instance. You can, too, still buy SN74LS14, or MC1489A...

No, it doesn't! It gets me to 25mV. Told you it was counterintuitive! :)

It's a factor of 1000 away from where I want to be - which is indeed in the region of 25uV or better of *synchronous* (to the output) ripple.

Here's the result of using an opamp-comparator "predriver" stage

Note the vertical scale is now 10uV per division (greenish noisy trace).

But oh, the expense...how can you suggest such an extravagance?!

--

John Devereux