Ping Spehro Pefhany

About that de-glitching thing, dunno how the conversation ended, and you have probably solved it by now, but I had a reason to mention HC4053 and not the simple open switch HC4016.

I have found that using those switches to 'open' a contact sucks for reasons you mentioned. So normally I try to use the 'select' mode, in the case of your example:

clock derived 2us signal (early) ------- || signal with glitch in first 2 uS---- 0 || \|| \ \ 0---------------- buffer ---------> out deglitched HC4053 | | 0 | | | |--------------- R1 ----------- | === C1 | ///

If you chose the R1 C1 time constant so it is practically fully charged in <

10uS, then switching the input to C1, and not simply only open the switch, makes sure it sees a low impedance last voltage level.

I am using something like this to insert black level pulses to get rid of noise spikes in a video signal. The HC4053 switches fast enough, and any 'charge injection' is shorted by the driving impedance or C1. Not sure if this works for you, or if C1 can be accurately charged to the required precision in the remaining >=8uS, that is simple math, but maybe it is something of use for other applications if not this one.

Reply to
Jan Panteltje
Loading thread data ...

you mentioned.

out deglitched

10uS,

noise spikes in a video signal.

driving impedance or C1.

required precision in the remaining >=8uS,

if not this one.

So the switch charge injection ends up in C1, This will raise the output voltage by Qinj/C until the next sample. So just make C large enough, nice.

There will still be a short glitch I think during the transition. Another C perhaps at the buffer input?

--

John Devereux
Reply to
John Devereux

you mentioned.

out deglitched

10uS,

spikes in a video signal.

driving impedance or C1.

required precision in the remaining >=8uS,

not this one.

Hi, Jan, thanks for the thoughts.

I think the output glitches from charge injection (q/C) will be too large. I need about 60dB attenuation- down to microvolts assuming

+/-15V supply (also a problem for HC or even 4000 series CMOS). For example, a typical high voltage analog switch spec is a couple of mV, or some tens of pC. Some are about 10x better, but still not good enough, nor high enough voltage (2pC at 5V is as bad as 6pC at 15V, plus TCVos and such like starts to rip through the error budget).

BTW, when you close the loop around the buffer that way-- it should work okay for a short periodic opening as in this question, but I worry the phase margin might not be enough.

Reply to
Spehro Pefhany

On a sunny day (Thu, 26 Jul 2012 19:41:46 +0100) it happened John Devereux wrote in :

you mentioned.

out deglitched

10uS,

noise spikes in a video signal.

driving impedance or C1.

required precision in the remaining >=8uS,

if not this one.

I think very short glitches can be low-passed. Also you could use more than one switch, so one in the feedback replacing R1.

It is an interesting problem, need some real numbers and a test setup to see hat can be done I think.

Reply to
Jan Panteltje

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.