'scope probes and 100MHz EMI

The other day I hooked up a LT1016 comparator (live bug prototype on a

6=94 X 8=94 piece of copper clad) to trigger on pulses from a PMT. Pulses are ~ 300ns long and ~100mV high. I sent a low level sine wave into the input (50 ohm terminated to ground) and looked at the output with a X10 scope probe. (TEK TDS2022 200MHz scope.) There was 100 MHz =91stuff=92 visible during the transition. I hung extra caps where I could and added 10 ohm resistors to the supply lines, but nothing helped. I went out and found the 100MHz probe that came with TEK2022... no change. (I've just read Linear's AN13 on fast comparators... Thanks again Jon K.) And yes I compensated the probes. I finally noticed that when I hooked the probe to ground I could see 100MHz =91stuff=92 with about 100mV of signal p-p. I then put a Schmitt trigger inverter (74HC14) on the output of the comparator and when I looked at its' output everything was fine. If I hooked the probe to the input to the inverter (output of the comparator) I could see the 100 MHz =91stuff=92 again.... Though the inverter cleaned things up considerably.

I took a new piece of copper clad, soldered a piece of buss wire on one corner. I hooked the probe and ground to the buss wire and walked around my lab/ office with it. There was 100MHz stuff most everywhere. I couldn=92t find any strong source, but there where nodes where the signal was much smaller.

So can anyone help me understand what I=92m seeing? I assume it=92s some sort of capacitive pickup. (Without the large piece of copper clad I don=92t see anything. I also put a few small coils across the probe, but could not see any magnetic pickup.)

Second do I need to look into getting a better probe? It would be nice to be able to look at the comparator output without the probe coupling all sorts of =91stuff=92 into the circuit.

Thanks,

George H.

(Oh the comparator circuit works great on the PMT pulses.)

Reply to
George Herold
Loading thread data ...

caps

t a

How long was your ground wire? One handy trick is to wrap a paperclip around the ground contact on the end of a probe to use as the "ground probe". You can then bend the paperclip to hit a ground while the probe tip sits on the point of interest. This minimizes the ground length (and loop area).

You could be picking that up in your probe ground. It will be everywhere.

me

Reply to
keithw86

I was using about a 3" length of wire w/alligator clip. I did try winding the wire around the tip, but nothing changed. The paper clip is an idea I've never tried before. Thanks. What I should really do is find the bag of changable tips that came with the probe. (They should add a 'parts bin' to plastic DSO's, ...you could keep the probes there too.) My guess is I'll still see the 100Mhz emi. So when I was just looking at the ground signal of the piece of copper clad, I cut pieces off of it and made it smaller and smaller. I cut the 6x6 to 1.5x6 and not much changed. When I made it 1.5x4 the signal was perhaps 70%, (my notebooks at work), at 1.5x1.5 the signal was below 10 mV (10%) and hard to see. I was wondering if it's some resonance in the probe/scope. I need to hit the thing with a fast pulse or step.

d

es

Yeah I ground both sides of the probe and there's 100mV at 100MHz.

Thanks, George H.

Reply to
George Herold

Probes tend to pick up ambient noise, like FM stations, bus activity, whetever. Fet probes are a lot better than passive ones.

For really clean waveforms, get a piece of coax with a BNC on one end. On your copperclad breadboard, solder the shield to ground and run your input signal into the center lead, directly or through a resistor.

If your scope can be set to 50 ohms, or you use a feedthru terminator, use a 450 (or 470) ohm resistor at the breadboard end of the coax. That will form a super-clean, super-wideband 10:1 probe. I do this well into the GHz range.

You can also solder a coax connector directly down to your copperclad. I like SMBs or SMAs for really fast stuff. The connector center pin goes through a resistor to whatever you want to probe. Then use a standard cable to run to your scope. Now you can save the breadboard without a dangling cable.

ftp://jjlarkin.lmi.net/BreadBoards.jpg

ftp://jjlarkin.lmi.net/BreadBoards2.JPG

What's a PMT cost these days?

John

Reply to
John Larkin

Your "stuff" sounds consistant, but the 100MHz figure reminded me of static ringing:

formatting link
and don't forget Part 2:
formatting link
and Part 3:
formatting link

Dave.

--
================================================
Check out my Electronics Engineering Video Blog & Podcast:
 Click to see the full signature
Reply to
David L. Jones

a caps

ut a

s

ome

Great, I'll try it and let you know what I see. (If my comparator can't put out 10mA I can always use a 1k resistor and make a it

20:1.) Say do I need to put a few pF of capacitance in parallel with the 450 ohm resistor to compensate for the ~30pf in the coax cable?

ftp://jjlarkin.lmi.net/BreadBoards.jpg

ftp://jjlarkin.lmi.net/BreadBoards2.JPG

Nice... Those are a lot neater looking than my hack jobs.

We get PMT's from Hamamatsu for ~$100 each in lots of 100. They also sell us the PMT sockets with built in high voltage supply. For about $200 each. (They kinda suck as there is a bunch ~80kHz ripple from the HV supply.)

George H.

Reply to
George Herold

ra caps

put a

s
d

es

some

ic

illo...

unusual-oscilloscope...

loscope...

=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

log.com- Hide quoted text -

Yeah, I saw those videos of yours a while ago and was reminded of it. This is some sort of capacitive pickup.. though I don't understand it. IIRC, most of your bouncing chair pick-up was via the magnetic loop formed by the probe tip and ground wire.... When you used a piece of Alum foil to short the probe tip the picked up signals decreased by

80% or so... I saw nothing with just a loop, or when I used a few turn coil across the end of the probe.

Say the boss wants a monitor of the digital pulses. Can I use a

74HC14 to drive a coax terminated with 50 Ohms? I don't care if the voltage doesn't make it up to 5 volts. Can I parallel a few of the inverters for more current? Or will that cause bad things to happen? Hey maybe I can use John L.'s divide by ten trick. (I'll put 450 ohms in series with the output).. They'll see 500mV pulses on the 'scope but that should be fine.

George H.

Reply to
George Herold

The scope end should be terminated in 50 ohms. If your scope can't do this, use a feedthru terminator. The input end of the coax will look like a 50 ohm resistor (ie, no cable capacitance will be apparent) so a 450 or 950 ohm resistor makes a theoretically perfect wideband 10:1 or 20:1 probe.

Ordinary carbon film or carbon comp resistors will be fine up to a GHz or so. Caddock makes some axial resistors that make nice 8 GHz probes if you use them this way. They plug into an SMA female nicely.

Nice price on the tubes, steep for the kinda-sucks sockets. A little flyback thing and a C-W multiplier chain would have maybe $12 worth of parts.

I visited the Hamamatsu PMT factory once. It was like a high-tech clean room in hell: gas flames on practically every bench.

John

ps - cheap LVDS line receivers make great fast comparators.

Reply to
John Larkin

ringing:

formatting link

2:
formatting link
3:
formatting link

Podcast:

formatting link
Hide quoted text -

If you want a clean step at the scope end, you can...

  1. Source terminate. Drive the coax with a 50 ohm impedance, using a
50 ohm resistor or a divider that's Thevenin 50 out. This works into a hi-z scope input, no reflections. It also works into a 50 ohm scope, at half the gain.

  1. End terminate. If the scope is 50 ohms, drive the signal into the coax through any resistor you like, forming a divider against 50 ohms. Also clean and reflection-free. Resistors above a couple of K will start to introduce overshoot from stray shunt capacitance, but most scopes are too slow to notice.

John

Reply to
John Larkin

a

re

xtra caps

th

n put a

ngs

ked

odes

s some

I

e,

I used metal film resistors. Seemed to work fine but there was a bit of weirdness on the end of the pulse response.... ringing 'gone bad'. Adding a pF of capacitance only made it worse.

So the good news is the voltage divider through the coax reduced the

100MHz 'stuff' by about 1/2. But it's still there... I think this may be a circuit screw up on my part. After the 50 ohm termination on the input line, I've got a series 300 ohm resistor and then 33pF to ground in front of the non-inverting input. There are some ~GHz switching transiients on the input pulse, and I was adding a bit of low pass filtering so as not to trigger on them. I clipped out the cap today and the 100MHz stuff went away.... Maybe I can 'buffer' the cap between two resistors? or reduce the cap some.

I was trying to get this circuit to oscillate in LT spice tonight, but I'm pretty much a LTspice newbie and it seemed to work fine in transient analysis.

We might save some money by redoing the electronics on this instrument. (there is even more 'low hanging' electronic fruit.) But it's always a question of how much and what else could I be doing with my time. I've not done much HV or fast pulse stuff, so I'm going to make lotsa mistakes.

text -

Some how gas flames and clean room don't mix in my brain.

George H.

Reply to
George Herold

a
.

ere

extra caps

ith

I

en put a

d
d

ings

lked

nodes

=92s some

d I

be,

tatic

oscillo...

cilloscope...

=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

evblog.com-Hide quoted text -

Cool, thanks John I'll try the source terminated idea too. At the Vanderbilt FEL we had setups where the PMT was 10+ feet away from the electronics. And we would put 50 ohm terminations on both ends of the coax cable.

George H.

Reply to
George Herold

caps

t a

me

Hi guys, Well a bit of screwing around with the LT1016 comparator, that I thought I=92d report on.

The simple circuit now looks like this

+--R4R4--+ | |\ | Vin---+--R2R2--+--R3R3-+-+ \ | | | | >---+--->to 74HC14 R1 C1 +--- / R1 C1 | |/ LT1016 | | | GND GND Vref

Vin comes from a PMT and is a ~100mV pulse with several hundred ns width. R1 =3D 50 ohms R2 =3D R3 =3D 100 ohms C1 =3D 100 pF R4 is 50k ohms, (though this may get tweaked a bit to set the hysteresis level)

R2 and C2 form a lowpass filter that gets rid of switching crap on the input signal. Without R3 the capacitor C1 caused the 100 MHz of =91stuff=92 in the room to become coupled into the input. I assume this was due to the small stray capacitance between the output and the (+) input. If I put a few (2.2 =96 10pF) of capacitance in parallel with R4 then there would all sorts of oscillations during the transition. (1 pf in parallel seemed to make things a bit cleaner..???)

I don=92t really understand

1.) Why does the stray capacitance from output to input cause the 100MHz =91stuff=92 to couple in better? 2.) Why does R3 reduce this effect?

(I guess the real problem is I don=92t know how the 100 MHz =91stuff=92 get= s into the circuit.)

Thanks,

George H

Reply to
George Herold

Sounds like a flakey layout. Does the PMT signal come in on a nice coax? How is the shield grounded relative to the comparator circuit?

Is there a good ground plane? Are you still breadboarding on copperclad?

Is Vref bypassed close to the comparator? I'd use an RC for Vref, both close to the comparator.

Is comparator Vcc bypassed?

External hysteresis may not help. 50K certainly won't. I prefer comparators with internal hysteresis, if any.

Got a picture?

John

Reply to
John Larkin

e

tra caps

h

put a

gs

ed

des

some

I
,
m

gets

"> Got a picture?"

OK airing my dirty laundry....

formatting link

Yeah this is still the same messy piece of copper clad.

I've got bypass caps everywhere I can put them... on Vref there's a

220pF and a 0.1uf ceramics.

Since the input pulses are about 100mV tall, I wanted about 10mV of hysteresis.. which is about what I've got.

Without R3 in place the amount of 100MHz 'stuff' during the transition depends on the value of C1. Changing the R2 feeding it doesn't do much. (Well of course it changes the RC time constant and amount of hysteresis) Once I found that C1 was the issue I was just flaying around trying to understand it. What I don't get is why putting R3 in place kills of this effect. Since there is no current flowing into the (+) input why does a wee bit of resistance help things?

Perhaps I should stick this in a metal box and see if that gets rid of the 100 MHz stuff.

The circuit seems to work just fine now... But I worry that not understanding why it works will come back and bite me sometime in the future.

George H.

Reply to
George Herold

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.