Electronics > Projects, Designs, and Technical Stuff

Generating extremely fast rise-time square waves for probe compensation

(1/3) > >>

Gridstop:
I bought some 300mhz pomona 10x probes, and in addition to a regular low-frequency compensation adjust in the probe for the 1khz calibrator on the scope, they have 3 adjustments for rising edge, mid freq, and low freq. For calibration they recommend a 1Mhz clock signal with a <=1ns rise-time.

For generating such a signal it seems a number of canned oscillators are capable of such a thing, particularly it seems MEMS oscillators. (The frequency accuracy is not a major concern, so any plain ol' accuracy clock is fine.)

A number of them support ~1ns avg 2ns max on a CMOS style output (no resistive load, up to 40pF capacitive), which should be fine for driving a 10M, ~15pF probe input.  If I put some good low-ESR supply caps on it, will I really get a nice square wave, with flat top & bottom out of it? I'm trying to think if I should do a PCB-mount BNC connector to use with the probe's BNC adapter, or if that would add excessive capacitance.

Lastly, my scope also has much faster (1Ghz) 50 ohm inputs, but to drive that would require a PECL output clock, unless someone knows of a 50-ohm coax driver/buffer that can run off the clock's CMOS output and thus provide a second output for driving coax. It'd be nice to be able to compare the 1Ghz 50 ohm input to the 200mhz/300mhz probe input to see what it's 'supposed' to look like. But I guess it'd be virtually impossible to actually SEE the same signal, since typically the 50-ohm drivers can't handle the capacitance of the passive probe, and the CMOS drivers that can handle the capacitance can't handle the resistive load of 50 ohms, so there's no way to hook both up simultaneously, unless someone knows of a really amazingly good 50-ohm coax driver.

alm:

--- Quote from: Gridstop on January 15, 2012, 05:56:32 pm ---I bought some 300mhz pomona 10x probes, and in addition to a regular low-frequency compensation adjust in the probe for the 1khz calibrator on the scope, they have 3 adjustments for rising edge, mid freq, and low freq. For calibration they recommend a 1Mhz clock signal with a <=1ns rise-time.

--- End quote ---
At least with the 300MHz plus bandwidth Tek probes I know, they usually only need adjustment when some part of the probe (eg. hybrid or cable) is replaced. I assume you're mostly doing this for educational reasons, since the time, effort and costs are likely much more then the advantage gained by adjusting the probes. Not that there's anything wrong with that.


--- Quote from: Gridstop on January 15, 2012, 05:56:32 pm ---For generating such a signal it seems a number of canned oscillators are capable of such a thing, particularly it seems MEMS oscillators. (The frequency accuracy is not a major concern, so any plain ol' accuracy clock is fine.)

A number of them support ~1ns avg 2ns max on a CMOS style output (no resistive load, up to 40pF capacitive), which should be fine for driving a 10M, ~15pF probe input.  If I put some good low-ESR supply caps on it, will I really get a nice square wave, with flat top & bottom out of it?

--- End quote ---
I have never tried, but I wouldn't expect a low distortion square wave unless specified. There's no reason for an oscillator to produce a very nice square wave, for EMI reasons it may be better to have it slightly overdamped. There may also be ringing in your circuit.


--- Quote from: Gridstop on January 15, 2012, 05:56:32 pm ---I'm trying to think if I should do a PCB-mount BNC connector to use with the probe's BNC adapter, or if that would add excessive capacitance.

--- End quote ---
At least make sure to use a very short ground lead, a few millimeters is better than a few inches. The best ground connection is coaxial, like the dedicated probe tip connectors some vendors cell, or a BNC connector. Second best would be those spring-like connections you slip around the ground sleeve near the tip, and various other attachments that attach to the tip. The 4" ground lead with alligator clip is only suitable for lower frequencies and if fidelity is not critical. A square wave with a 1 ns rise time contains a wide spectrum of high frequencies, so it's very likely that one of them will resonate in the LC circuit formed by the inductance of the ground lead and the capacitance of the probe. See the link to Howard Johnson's article for more information. I believe the Tektronix app note ABCs of scope probes also talks about this issue.


--- Quote from: Gridstop on January 15, 2012, 05:56:32 pm ---Lastly, my scope also has much faster (1Ghz) 50 ohm inputs, but to drive that would require a PECL output clock, unless someone knows of a 50-ohm coax driver/buffer that can run off the clock's CMOS output and thus provide a second output for driving coax. It'd be nice to be able to compare the 1Ghz 50 ohm input to the 200mhz/300mhz probe input to see what it's 'supposed' to look like. But I guess it'd be virtually impossible to actually SEE the same signal, since typically the 50-ohm drivers can't handle the capacitance of the passive probe, and the CMOS drivers that can handle the capacitance can't handle the resistive load of 50 ohms, so there's no way to hook both up simultaneously, unless someone knows of a really amazingly good 50-ohm coax driver.

--- End quote ---
This is the strategy I would employ. Observe the signal via both the 50 ohm input (should be cleaner) and the 10x probe. I would use a resistive divider probe connected to the 50 ohm input. Some links:
Probing High-Speed Digital Designs (Howard Johnson)
DC to 1GHz Probe Construction Plans (Doug Smith)
A wideband 1:21 DIY 1 kOhm oscilloscope probe (by forum member Jahonen)

Of course now you have the problem how to characterize this probe. You can either assume it to be much better than a 300 MHz Hi-Z probe (probably a fair assumption), or connect it in parallel with a 50 ohm cable to a 50 ohm line driver. Since the capacitance should be much lower than a passive probe, the capacitive load should not be an issue.

Gridstop:
If, as you say, a regular old clock is unlikely to be a nice square wave (and I certainly don't disagree). I still have to use like... a comparator or something that's meant to drive a 50 ohm load and produces a nice square wave output. Then I could use coax to the 50 ohm input on my scope to see what it's 'supposed' to look like, and try tapping it with a BNC T-adapter, then sticking the passive probe into that using the BNC adapter. I could easily see on the 1Ghz 50 ohm input if the probe was distorting the waveform, since the wave would change on the 50 ohm input.

I could just try with a PECL clock output, which I would GUESS has much less ringing than a CMOS output, and are specified with risetimes as low as 250ps, but ONLY for capacitive loads up to 2pF or so. And tapping the coax for the passive probe would certainly raise that somewhat. But again I could see how much it changed on the 50 ohm input, but if it ruins the rise time, I've wasted some money. Although even if the risetime goes up to all of 5-10ns, it should still let me dial in the passive probes MUCH closer than with my 1khz scope calibrator, which has a risetime in the microsecond range, not nanosecond.

A small set of PCB's from itead/seeed, a PECL clock, and regulators, BNC connector, etc..etc.. shouldn't cost more than $20-30.

Rufus:

--- Quote from: Gridstop on January 15, 2012, 05:56:32 pm ---But I guess it'd be virtually impossible to actually SEE the same signal,
--- End quote ---

Stick a signal into one side of a BNC tee connected to a 50 ohm input on the scope and the probe with BNC adapter into the other side. The scope and probe are going to see the same signal.

Compensate the probe so they look the same (barring delay from the probe lead).

For the same problem I made a buffer from a 74FCT541 (fastest thing in my junk box), dead bug, 7 buffers driving the output in parallel, one buffer speeding and squaring up the signal to drive the other 7. Just fed it a kHz or something from a signal generator. Displayed rise/fall time was 2-3ns but its only a 200MHz scope.


bfritz:
Making something for this purpose, without having another probe or equipment with known characteristics will make this task very difficult.  It is a tricky task, even if you had all the correct equipment.

There is some pretty decent test equipment for reasonable prices on eBay for what you want to do.  What you want, is a pulse generator.  The most common use, is for calibrating probes.

http://www.ebay.com/sch/Pulse-/97197/i.html?_catref=1&_dmpt=BI_Signal_Sources&_frs=1&_trksid=p3286.c0.m359

The Agilent 8012B and Phillips PM5715/11 come pretty close to your specs if I remember correctly.  I think with a high impedance output they have a rise time of about 5ns.  (Please check the number, as that is what I remember, about 90% confidence.)  They seem fairly reasonable used at about $100 for Buy it Now prices.  I typically have good luck offering 85-90% of Buy it Now, and having people accept my offer.

I'm sure there are other devices there, likely for a bit more money, that can achieve the rise time you want.

Navigation

[0] Message Index

[#] Next page

There was an error while thanking
Thanking...
Go to full version