DIY X1 High BW Passive Scope Probe

[StillTrying]

Just an experiment.
Both probes in parallel across the same 100 resistor. Of course if you can't decide which channel is the DIY and which is the proper scope probe that's a good start!

[tron9000]

can we have some more detail about what this is about? I'm going to be doing some test that will require measuring ripple on a DCDC converter (I have seen daves video on this) and am thinking about this.

Whats your test setup? what cables are you using? whats your target BW?

[macboy]

Define "High BW".
If you mean 100 kHz, then fine, maybe. If you mean 100 MHz, then you won't succeed.

The problem with a passive 1x probe is that you have no means to isolate the DUT from the high input capacitance of the scope (>= 20 pF typically) or the very high capacitance of the coax cable that you use to connect the probe to the scope (>=  30 pF per foot). This is perhaps the single most important reason why all high bandwidth passive probes are at least 10x attenuating, often higher. They add series impedance at the probe tip to isolate the DUT from the downstream capacitance. Just a few pF becomes a very small impedance at 10's of MHz, and this loads the circuit and alters the signal to the point that what you see isn't what you would have seen if the probe wasn't connected. Adding series impedance necessarily means attenuating the signal, but we live with that.

Even a "pure" coax transmission line probe, which you could create with a series 50 ohm series resistor, 50 ohm coax, and a 50 ohm termination at the scope, will be 2x attenuating, and it will load the circuit with 100 ohm plus maybe one pF of capacitance to ground. However, this might work well to observe GHz signals at nodes that don't mind driving that 100 ohms. Passive "Low Z" (as opposed to typical High Z) probes are also somewhat common, with impedances of 500 to 1000 ohm (plus ~1 pF) and attenuation factors of 10x (500 ohm) to 20x (1 kohm).

The only way to make a true high-bandwidth 1x probe is to make it active. This is why expensive FET probes exist.

Try this: Disconnect one of your probes. Observe signal change on the other, save that trace. Reconnect, then disconnect the other. Observe and save that trace. Compare the two saved traces. Are they still so similar? Now try a typical High-Z 10x probe, and compare that trace to the other two. How do the different probes affect the signal?

Suggested reading is Tek's "Probe Fundamentals". Latest pdf I found was 2009.

I hope you don't feel like I'm poo-pooing your idea. Never stop thinking, experimenting, and discovering.

[tggzzz]

Just an experiment.
Both probes in parallel across the same 100 resistor. Of course if you can't decide which channel is the DIY and which is the proper scope probe that's a good start!

... or it is so crappy that its poor performance has completely slugged the source

[StillTrying]

"Define "High BW".
If you mean 100 kHz, then fine, maybe."

Well I'm already up to 30MHz, actually twice that if you ignore the terrible ringing above about 45MHz.

"or the very high capacitance of the coax cable"
I'm getting around that by not using a coax cable. 

"I hope you don't feel like I'm poo-pooing your idea."
Not at all, I've read masses of the probe stuff over the years, I know it's a daft idea.

"... or it is so crappy that its poor performance has completely slugged the source"
LOL
CH1(yellow) is the scope probe switched to X1 for comparison.


pcab.gif
My 0.5m of 'special unshielded cable' surprisingly doesn't seem to have any worse noise pickup than a 1m scope cable on X10.


px10amp.gif
Is mostly self explanatory, With the scope probe on X10, CH1 is the highest I can go without the 20MHz BW limit coming on. I think there's quite a bit of 40-70MHz ringing adding and subtracting from the true CH2 amplitudes.

I'm afraid you'll be disappointed to know I'm just using a hand plaited 0.5m 3x 7/0.2, 1 signal and 2 gnds. It's something I've been going to try for years over twin core or twisted pair, it seems to have good possibilities for a X1 probe up to about 35MHz.

[alm]

As you are already reporting, without any termination or lossy cable, I would expect severe ringing and resonances at higher frequencies.

The obvious feature is the very short (0.5 m) cable, which will limit the versatility. On my bench, for example, the scope is already about 0.4 meters above the surface, so the test point would have to be pretty much directly under the probe input. Obviously a shorter cable makes for better performance, especially for 1x probes. Tektronix used to make the P6101 in three lengths:
- 1m, with 34 MHz bandwidth and 32 pF of loading.
- 2m, with 15 MHz bandwidth and 54 pF of loading (this was the standard version).
- 3m, with 8 MHz bandwidth and 78 pF of loading.

Extrapolating from this and noting that bandwidth and capacitance are mostly linear with length (as expected there is a little capacitance in the probe body), a 0.5m probe would have about 60 MHz of bandwidth and 20 pF of loading. So I wonder if you would not get even better performance if you got a good 1x probe (I would not use a switchable one), shorten the lead to about 0.5 meters and solder it back (though soldering probe cable requires some fairly aggressive flux).

To characterize I would measure the input impedance, do a sweep from DC to well beyond its -3 dB point (both to determine bandwidth and look for resonances in the pass band), and test the response to a pulse much faster than the risetime of the probe. This should give you a decent idea of the probe performance. The current pulse may just have a limited spectrum over which the probe response happens to be fairly flat.

Finding a conventional probe that is a fair comparison will be tricky because of the short cable, so I would either make a DIY probe with a longer cable (to match the commercial probe) or hack a commercial probe. For hacking the older Tektronix probes (P6101, P6101A) might be easier than modern probes since they came apart easily.

[StillTrying]

I usually work(LOL) in front of the scope on the desk so 0.5m is just about long enough 80% of the time. I've often thought about shortening a probe, and got a few old X1/X10s which could be 'upgraded'.

I don't have anything near enough fast enough for proper tests.
Here's an 74LS pulse, each probe has it's own LS output, Blue is the scope probe on X1, Yellow is the scope probe on X10 and Purplely is the DIY, the X1 DIY's tip is just a 39R resistor, its bandwidth and loading look quite good to me!

[macboy]

Looks quite good 
What is the grounding method for the x10 probe? big old ground clip? How much better is that one with a short spring tip ground?
I don't suppose you have any 74F or 74ACT kicking around, those can give a fast edge (~2 ns). 74LVC can get <1 ns.

[StillTrying]

What is the grounding method for the x10 probe? big old ground clip? How much better is that one with a short spring tip ground?

With the short clip directly on the IC's pins the only noticeable change using the X10 scope probe was the -0.5V dip reduced to -0.1V dip.  springclip25nsdiv.gif

You supposed right, but I can create fast edges mechanically, I'm sure the DIY X1 will be very close or even exceed the scope probe's rise time that way.

Try this: Disconnect one of your probes. Observe signal change on the other, save that trace. Reconnect, then disconnect the other. Observe and save that trace. Compare the two saved traces. Are they still so similar? Now try a typical High-Z 10x probe, and compare that trace to the other two. How do the different probes affect the signal?

I tried this with the X10 scope probe and the X1 DIY,  I didn't measure the rise times, but there's very little visible change when the other probe is added to the same LS output, a 0.1V drop in the very peak of the waveform maybe, but visibly you'd barely notice that the other probe had been added.
I got these all on one screen, but the rising edges of the 4 waveforms didn't match up, the X1 DIY arrives 5ns before the X10 scope probe but EXTTrig ignored the 5ns skew.

Finally, I went to great trouble to produce a pic, get your  s and your s ready. 

[StillTrying]

Shall I try a flat 2 or 3 core before I buy enough extra-thin extra-flexible to make a pair of these things?
Or should I give up this daft experiment?