7

I am a power electronics engineer in R&D and often measure fast switching waveforms, specifically high bandwidth MOSFET gate signals. When connecting oscilloscope probes in the past I have found that the ground clip often induces unwanted HF content due to the inductance of the loop and the very high dI/dt and dV/dt in the area (switching 800 A at 900 V).

To mitigate this I was always taught to create little ground hoops with a wire and slot my probes in. However, due to smaller and smaller designs, the access is becoming increasingly difficult.

I was considering incorporating some of those miniature 2 mm RF connectors (U.FL) on the next PCB design and running very small diameter coax cable out of the device to monitor accurate waveforms on an oscilloscope. However, I am not experienced with RF and cable transmission specifics and was wondering how the connector impedance might affect/load my real signal or the measured signal on the oscilloscope.

Could anyone shed some light on this idea?

Peter Mortensen
  • 1,676
  • 3
  • 17
  • 23
Jono Moran
  • 81
  • 8
  • 2
    The connector impedance might be irrelevant. If you plan to directly connect a coaxial cable to a FET gate signal, the cable presents a load which adds capacitance and depending on if the measured signals are fast enough compared to how the signal travels in the cable, you get transmission line effects and the cable will look like a 50 ohm load (assuming it has 50 ohm characteristic impedance) and to reduce reflections it needs 50 ohm termination at the scope end. That is a huge extra load for the FET gate driver. – Justme Feb 18 '21 at 11:58
  • well, sounds like a job for high-ohmic probes – Marcus Müller Feb 18 '21 at 12:01
  • 1
    I feel like I should point out that U.FL is a connector completely unrelated to BNC, and it really isn't what you could call a "miniature BNC connector". – Hearth Feb 18 '21 at 16:34

3 Answers3

10

You might consider adding an attenuator probe to your PCB consisting of a single series resistor. The 50-ohm coax cable to your oscilloscope can be any length - the oscilloscope input at the far end of this cable must be 50-ohm-terminated, not left to its default 1MEG

schematic

simulate this circuit – Schematic created using CircuitLab

The series resistor Rs will load the MOSfet gate driver only when the coax to the 'scope is connected. This resistor, combined with 50-ohm 'scope termination is a wideband attenuator - its attenuation should likely be made an easy-to-calculate 20:1 or 50:1. You should be able to arrange the short PCB path from Rs -to- coax connector to have 50-ohm impedance to PCB ground plane.

glen_geek
  • 23,591
  • 1
  • 22
  • 50
  • Of course, a small trimmer capacitor in parallel with Rs would probably be a good idea. – Hearth Feb 18 '21 at 15:45
  • @Hearth Yes, if you try for more-extreme attenuator values (using a larger-value Rs) then a parallel compensation capacitor is correct. Calibrating it is a problem, requiring a known fast, clean edge. As OP suggests, ground loop paths in this probing example would likely be a bigger problem. – glen_geek Feb 18 '21 at 16:02
  • Probably not necessary here, fair. I've been working on a circuit that involved similar attenuators lately and I did need compensation capacitors, but I needed a bandwidth in the hundreds of megahertz and a larger (100:1) attenuation. – Hearth Feb 18 '21 at 16:05
  • I've used this approach many times and it works quite well. If you have space for a connector, you might consider using a HDBNC (also called micro-BNC) with the resistor also designed in. HDBNC are quite small and easy to source. Their Z0 is 75R so you'll need a T and 75R terminator at the scope input instead of 50R and you'll have to adjust your scaling factor accordingly. – rsbonini Feb 18 '21 at 16:31
  • 2
    This should work. With 1k resistor, the ratio will be 1:21. As some scopes may not allow 1:21 or custom probe ratios, sometimes suggested values are 950 ohms to get 1:20 or 450 ohms for 1:10 ratio, but it depends a lot how much extra load you are allowed to put to the gate driver. – Justme Feb 18 '21 at 16:56
  • All very useful information here, @Justme your comment on the original post about the 50Ohm oscilloscope mode loading the FET driver is exactly what I was worried about as every joule counts in this game. The gate rise time generally appears to be in the 100nS range with 10's of MHz artefacts and oscillations and I generally use a x10, 200 MHz probe with 200Mhz Cleverscope CS448, HV isolated channels (which I can highly recommend) I'm not sure this oscilloscope can do 50Ohm input mode though, would I need external termination for this suggestion? – Jono Moran Feb 18 '21 at 19:28
6

I like to use this sort of BNC to scope probe adapter, which frequently is included in the accessory kit you get with the probe. You put it on the tip of your oscilloscope probe, then plug it into a perfectly ordinary BNC connector. Since the probe has a high resistance (9 MΩ for a 10x probe) positioned right at the tip, there's minimal loading from this method.

enter image description here

(source: Amazon listing for Cal Test scope probe adapters)

BNC connectors are physically fairly large, however. You can run traces out to one that's further away, but if you need minimal capacitance, which is likely if you're dealing with high slew rate gate signals, you can also use a small connector like MMCX and a BNC to MMCX adapter, which can be found cheaply on ebay, though I'm not sure how much to trust the quality of such an adapter.

The image below, from the ebay listing linked above, provides a good comparison of the sizes of BNC and MMCX:

enter image description here

U.FL is still smaller, but I wouldn't recommend using it for this purpose for a few reasons: It's not rated for more than a few insertions (I've seen connectors rated for less than five!), and it's not as simple to connect an oscilloscope probe to properly.

Additionally, if you don't have a resistance very close to the point being probed, your 50 Ω impedance line will probably add a few tens to hundreds of pF to the gate node--which can be significant if you're using IGBTs or SiC MOSFETs, or other devices with particularly low gate capacitance. This method of using a scope probe provides that resistance, so you only need to worry about the capacitance added by the connector and traces themselves.

Hearth
  • 27,177
  • 3
  • 51
  • 115
0

The U.FL is a 50 ohm connector and has no attenuation or compensation (scope probes usually have). If your scope has 50 ohm input or you use a feed termination you have a 50 ohm system and everything is fine. However, you are putting a 50 ohm load on your circuit, and it can't necessarily handle it.

If you need an high-impedance probe, you could use an U.FL-to-BNC cable and then use a BNC-to-scope probe adapter (usually is supplied as an accessory with the probe). Unless you are doing megahertz signals, the coax pigtail mismatch should be negligible (you do care about the probe loading, usually about 10 megaohm and 8 pF).

Peter Mortensen
  • 1,676
  • 3
  • 17
  • 23
Lorenzo Marcantonio
  • 8,231
  • 7
  • 28
  • 5
    when observing edges of high-powered MOSFET control signals, "MHz signals" is definitely not sufficient. You're still confusing frequency with bandwidth! And steep edges have very high bandwidths, and exactly these are of interest here. – Marcus Müller Feb 18 '21 at 12:58
  • Yes agree with that. I read that with the new GaNs there barely exist scopes that can see the edges. However if you are not strictly interested in edges I found that simply a good low ground impedence is sufficient and 10cm of pigtails do no harm (at least with a 500MHz) scope. If you need higher spec you'll be using active probes and Tek (for example, but probably other brands too) is happy to sell you probing modules to *solder* on your board – Lorenzo Marcantonio Feb 19 '21 at 06:54
  • @LorenzoMarcantonio That's not quite true. You need a pretty expensive scope, but it's not hard to see the edges. Now if you care about the edges of the *current* waveform, that's another story altogether, and even the best of the best current transformers don't quite have a high enough bandwidth for that. – Hearth Feb 19 '21 at 13:59