While trying to find a solution my short pulse problem I came across this interesting article:

http://www.eetimes.com/electronics-...hot-pulse-stretcher-detects-nanosecond-events about using a comparator to stretch nanosecond events now this seems that it if it can do ns events then it will work for my us event!


Struggling a it with the spice model tho and then I need to change the RC constant to give a time period of >2ms I dont want to use a IC if possible but i think the pulse at 10us is and need ing to make it >2ms the difference is too large for the simple RC network Sgt Wookie has been helping me with?

 

I've corrected your circuit, and then modified it so that it appears that it would work. Since you have now indicated in the other thread that the accelerometer actively pulls the output down when it goes low, I have added a pull-down resistor, Raccel, inside the accelerometer itself.

The latch input to the LT1711 appears to draw some current when the input is high; so the 470k resistor had to be eliminated.

I do not know how accurate the modelling is for the latch input of the LT1711, or if the latch exhibits a sharp break at about 1.5v as the model does. Your mileage may vary.

If you attempt to substitute other comparators, your mileage will vary far more.

The input pulse is 15uS wide; the output pulse is ~2.2mS wide.

See the attached image and .asc.

 

Thankyou very much Sgt. when you say mileage may vary? do you mean the output will be more dependant on the characteristics of the real IC?

I wanted if possible to avoid the need for a IC as this will make the whole thing far more difficult however the LT1711 might just be small enough to fit where we would like it too.

If there is a more suitable comparator you know of then i would be happy to use one of those Ideally one i get in a DIL package to test and small SMT for use.

 

Thankyou very much Sgt. when you say mileage may vary? do you mean the output will be more dependant on the characteristics of the real IC?

Yes. Simulating in SPICE is just a starting point; to see if a circuit appears to be within reasonable parameters before starting on the actual thing. The LT1711 comparator is extremely fast, having a response time of only 4.5nS.
Here is Linear Technololgy's page on the LT1711:
http://www.linear.com/product/lt1711
Be aware that the ABSOLUTE MAXIMUM supply voltage for the LT1711 is 12.6v from +V to GND and +V to -V. It is unwise to operate an IC at or near its' maximum ratings, as reliability will suffer.

I wanted if possible to avoid the need for a IC as this will make the whole thing far more difficult however the LT1711 might just be small enough to fit where we would like it too.

See the datasheet, on the right side of the page I provided the link to. The comparator is available in an 8-lead MSOP package. Dimensions are on page 11 of the datasheet. You will need a bypass cap across the +V and -V pins.

If there is a more suitable comparator you know of then i would be happy to use one of those Ideally one i get in a DIL package to test and small SMT for use.

They don't come in a DIP or DIL package. See page 11 for the available packages.

You should not attempt to breadboard the circuit. You will have entirely too many parasitic L, C, and R's introduced, which will negatively impact performance.

The length of wiring will be rather important; the longer the wiring is, the more inductance it will have. Inductance will wreak havoc upon short pulse signals.

 

OK thankyou for the information the current "test" set up does have some quite long wires the eventual solution will hopefully be all on a small PCB very small if possible (this creates additional challenges) the PCB would ideally be 6mm x <20mm the MS-8 package is 4.9mm +/- 0.15 so may just fit on a 6mm wide board. I don't have the facilities to do surface mount and the comparator doesn't need to be as fast as the LT1711 is capable of. The ability to run from 12V is handy as that's the only supply available.....

I noticed you are using a Voltage divider to give a 6V reference for the comparator obviously this means the threshold for the comparator is 6V. Is there any reason not to use a lower threshold such as 3 Volts? 6V may aid noise rejection. Just wondering if there was a reason for choosing 6V I am not aware of

What would be the optimal supply voltage for the LT1711 if not 12V as the threshold is 6V does the supply have to be >6V I have had a bit of a play and the attached spice file is giving me pretty much everything i need a very short pulse (15us) to a nice long one (12ms)

I also added a diode to reduce the supply voltage to the LT1711 is there a better way to do this without using a regulator?

I noticed you are using SPICE command Pulse does this alternate between 0V and 12V at each time marker in the list? I would like to simulate a pulse as low 5us which is the shortest i have found in real tests

 

OK thankyou for the information the current "test" set up does have some quite long wires the eventual solution will hopefully be all on a small PCB very small if possible (this creates additional challenges) the PCB would ideally be 6mm x <20mm the MS-8 package is 4.9mm +/- 0.15 so may just fit on a 6mm wide board. I don't have the facilities to do surface mount and the comparator doesn't need to be as fast as the LT1711 is capable of. The ability to run from 12V is handy as that's the only supply available.....

I've experimented with fitting the circuit to a 6mm x 20mm board; see the attached. I'm running into space problems, as indicated by the crosshatched areas. C2 is a 1206 size capacitor; which was the smallest standard size cap that I could use and run a 10 mil trace under. I later changed to a Kemet 0805K, the layout of which has a greater space between the two pads; and it fit.

The diode is a BAT54WT by Diodes Inc; it's in a SOD523 package, which is tiiiiiiny....

I put the wire pads on there, as you'll need some way to get power and signals on and off the board. -V will have to be common to your accelerometer ground and your Schmitt-trigger input.

I noticed you are using a Voltage divider to give a 6V reference for the comparator obviously this means the threshold for the comparator is 6V. Is there any reason not to use a lower threshold such as 3 Volts? 6V may aid noise rejection. Just wondering if there was a reason for choosing 6V I am not aware of.

It sets the comparator to trip at 1/2 of +V. Simply grounding the inverting (-) input would make the circuit unreliable, and VERY susceptible to noise.

What would be the optimal supply voltage for the LT1711 if not 12V as the threshold is 6V does the supply have to be >6V I have had a bit of a play and the attached spice file is giving me pretty much everything i need a very short pulse (15us) to a nice long one (12ms)

Well, a 5v-6v supply would be OK, as your Schmitt-trigger input is only looking for voltages in the ~1.2v-3.5v range or something like that.

I also added a diode to reduce the supply voltage to the LT1711 is there a better way to do this without using a regulator?

Well, if you could find a Zener diode in SOD523 or similar size package, and use that in series with a resistor to +V, you could get cheap regulation. Don't know offhand if there is room to do so; space is already at a premium without adding a couple more things.

I noticed you are using SPICE command Pulse does this alternate between 0V and 12V at each time marker in the list?

If you right-click on the symbol for V2, you will see the parameters I've given it; the Help file explains in some detail what the parameter mean.
I really don't have the time to explain things that are documented elsewhere; and it is not fair of you to ask me to do so.

I would like to simulate a pulse as low 5us which is the shortest i have found in real tests

It should be self-explanatory how to change the pulse width after you right-click on the symbol for V2.

 

Sgt you are a star, thank you so much for time and input.

It looks like the PCB is borderline size wise so I will see what can be done about that to give a bit more room.

Hopefully the manufacturer will give me the required detail on the input stage this week.

I'll have a good read of the help files and check the documentation.

As you seem to be the only person who has answered my questions is there anyway I can formally thank you for your assistance?

EDIT: Seems suitable Zener Diodes may be available EDZ6.2B http://www.rohm.com/products/databook/di/pdf/edz5.6b.pdf if they can dissipate the power. I have some maths to do.... (never my strong suit)

 

The manufacturer got back to me and advised i see page 65 of the technical manual see attached for the info. Doesn't seem to help though?

GPin1 is what is being used.

 

Well, no wonder the passive circuit didn't work - the input needed 8mA current flow.

It would have saved a lot of time if you had showed it to me when you first started asking about it.

You'll have to use an active circuit.

[eta]
There's another problem I just noticed; the latch input has an absolute maximum input of 7v. This means you'll have to have the supply to this particular comparator at under 6v, and the supply current will need to average in excess of ~10mA.

 

I only just got that information its in a different manual to main manual!

 

With the comparators' supply voltage being reduced, that also means that the accelerometer needs to get its' power from the lower voltage supply, or it will likely damage the comparators' input.

 

Hi Sgt, I have made some changes (the voltage reg is now sorted ) Sorry,

I also rearranged it a bit too (to make the PCB design a bit more obvious) the terminals in the relevant places etc.

Struggling a bit with Diode choice (D2), any recommendations on a suitable surface mount Zener diode?

EDIT: Updated the model

 

I've just added a Zener, changed the 220 Ohm resistor to a 200 Ohm to improve the transient response in the circuit, added a 10mA load to the comparator to simulate the output device (it should only draw 8mA, but I'm wanting some finagle factor here).

Anyway, have a look.

[eta]
Here's a 6.2v 300mW Zener in a SOD523 package:
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=568-6293-1-ND

You had some high-voltage Zener in your circuit. 6.2v is about as high as you can go.

 

Thank you again Sgt, I did wonder how one simulated the load on the output. With your help I think that's a pretty professional looking design.

Don't tell me your day job is at the Cape?

 

Well I have had a design for the layout done and sent it off should get the first ones back next week, then we can test them for real.

 

OK got the first one back from test (well 2 actually) one works perfectly the other not, so still have some issues to work through. But very pleased so far.

 

44mS wide output pulse? Gee, that's considerably longer than I expected; it should've been ~15mS if the input was a very short pulse. Of course, multiple input pulses within that 15mS time limit would increase the output ON time.

Did you change the size of the cap C1 from 33nF to 100nF? That would also explain it. Or, they may have swapped C1 and C2 when assembling it.

The board looks nice and compact. Can't read the numbers on the smaller resistors.

Try documenting what you observe as far as voltages, signals, etc. on the working prototype vs the non-working prototype.

 

That test was a hand shake (~2ms input) test so input pulse would have been longer, the C1 value is 33nf the real test will be done on monday where the input pulse will be shorter (circa the real 10us) and i will be able evaluate it better, i didnt keep the recordings of the none working one but it seemed to latch and not unlatch until the next pulse working more like an inverter than pulse stretcher. I will try and take some better pics of the board that can show the values

 

All right, in that case the drain from C2 via the LT1711's latch input leakage was about 1/3 that of the LTSpice model - certainly less than I expected.

Perhaps it was poor judgement on my part to not include a bleeder resistor across D1; as if the latch leakage is really low, you'll see the output latched for a much longer time than expected.

You might take that prototype that tested faulty, and solder a 1MEG resistor across D1.

Also, it's hard to tell by the photo, but the soldering on the end of the cap closest to the accelerometer (on the right) looks like it might not have flowed very well. They're so small that you really need to look at it under a microscope to be certain - at least *I* do.

 

SGT will the time of the output being high be relative to the input pulse length, as well as the Capacitor value? I thought that once triggered it would be the value of C1 that would determine the output period? regardless of the length of the input pulse?