Test points for medium/high-speed signals

[johnmam]

Hi folks,

I'm designing my first high-speed board for educational purposes, with both USB-2.0 high-speed and 100Mb/s ethernet using an MII bus.  Because i'm doing this for my own education, I want to be able to swap out termination networks and get a good look at my signal with a scope.  With medium-speed signals (like 50MHz MII in ethernet) or high-speed signals (like 480MHz for USB-2.0), how can I make some good test-points to poke the signal with my scope?  Because I'm doing this to teach myself a thing or two, it wouldn't be worth my time if I couldn't see the signal integrity on a scope.

My thought is to just expose the trace by removing some solder-mask, but I'm worried about a few things:

• If I probe the same trace over and over, will it affect the impedance by essentially "wearing the trace down" in one spot?
• I might have a hard time poking at a target that's 0.15mm wide

I've looked online for other suggestions, but people say that putting a big 'ol pad branched off of the trace will cause impedance issues (which makes perfect sense)

I totally understand that this is nowhere a heavy-duty application and that just removing solder mask, adding a solder bead/0 ohm resistor, or putting a pad in-line would be just fine.  I'm still curious though: if I had to design a 1300MHz DDR bus or USB 3.0 traces, how should I put down test points to measure signal integrity for those traces?

I've spent a while looking, but I wasn't able to find much information... sorry if this seems like a stupid or obvious question.

[Rerouter]

It comes down to what kind of tools you have on hand,

Generally for these speeds you would be using solder on probes with a preamp at the tip to isolate the circuit from the probe.

If your attacking it with a general probe and a flylead ground you wont have a chance.
Equally your probes impedance decreases as the frequency goes up, meaning you need really low capacitance probes once you get to 1Ghz. 

As for setting up test points on differential pairs, all you need to do is make your pad match the same impedance, so as you want a wider trace you would increase the spacing as you increase the pad side out to your test pads final size, like a crescent moon shape with a solder mask gap in the middle.

[johnmam]

Why won't I have a chance with a flylead ground?  Is the ground impedance just too much even for such a short run of wire?

i can't really seem to find "solder on probes" for sale anywhere, except for very high frequency ones that cost a fortune.  Can you give me some more keywords or example products for reasonably cheap standard models?  I'd like something under about $250.  Am I out of luck for that?

[tggzzz]

Inductance; 6" => 150nH, resonates at 100MHz with the tip capacitance.

Learn about "low impedance Z0" "resistive divider" probes, and how to make them with  resistor soldered to the testpoint with a 50ohm cable to the scope.

[joeqsmith]

It's been several years since I have done any high speed design work like this and I'm sure tools are much better now. 

Looking at digital above 1GHz, is always fun.  You may even be able to find some good used probes and positioners at a reasonable price now.  I picked up a few used ones for home for a reasonable price that are in working condition.  Deals are out there.   

I've used some of the Keysight connectors for logic signals.  No connectors on the board.  You would solder the shell to the board is all and the layout is the mating end.   You may want to have a look at their applications notes for ideas.   LeCroy also has some very good applications notes. 

Most of my books on the subject are fairly dated.  Howard Johnson's books are a good reference.  The basic principles are not going to change much. 

Envy you.  Have fun!

[johnmam]

@joeqsmith the pictures and google terms help a lot.  Thanks!  For the speeds I'm looking at I think it's overkill, but it's really nice to know what sort of tools I would need to work with if I wanted to do really high speed stuff > 1GHz.  One of the biggest mysteries of high speed for me has been "how do I test it?"  A lot of googling hasn't helped, but hearing it from you folks is helping clear up my confusion!

Oh, also, thanks for the suggestion to search for the LeCroy / Agilent app notes.  I found some really great ones that cleared stuff up for me.  this is great 

@tggzzz Would using a lead like this one (goo.gl/DG5HuZ) work for probing ground in the < 500MHz regime?

[tggzzz]

@tggzzz Would using a lead like this one (goo.gl/DG5HuZ) work for probing ground in the < 500MHz regime?

Yes, but it isn't the only way and there are some homebrew techniques; see https://entertaininghacks.wordpress.com/2015/04/23/scope-probe-accessory-improves-signal-fidelity/

But if you are looking at digital signals, a homebrew "low" impedance Z0 probe soldered directly to the testpoints can work well and is very cheap (Low in quotes, since its input impedance is significantly higher than bog-standard *10 probes.)

[alm]

If you are not willing to shell out for multi-GHz differential probes, then I would agree that a resistive Z₀ probe is the way to go. Howard Johnson and Doug Smith have written about them, and there have also been posts (1 2) on this forum. Since they are DIY and the components are cheap, you can easily solder them down. Obviously you need a scope with 50 Ohm inputs (or less ideal a feed-through terminator).

If you want a sturdier variants, you can sometimes find used Tek/HP Z₀ probes for under $100 on eBay. Some part numbers would be Tektronix P6156/P6158 (note that it comes with the tip and that the P6156 has multiple tips selecting attenuation) and HP 54006A. These are not made for soldering down; they were designed back when circuits were through-hole.

The logic analyzer connectors are mostly to cram many channels in a small space, but they are good to look at for termination and minimizing stub length. Note the suggested routing for some of them that has the traces running under the connector, creating a minimal stub.

Not sure what the cheap solution would be for differential probing. Probably a used Tek/HP/Agilent/Lecroy differential probe. I can not see CH1-CH2 work very well at these frequencies.

Also keep in mind that you might need much more than 500 MHz bandwidth even for a 50 MHz digital signal. It is all about the rise time of the signal and the probing system. See application notes by Tektronix and Keysight with details.

Videos about low inductance probe connections:

[tggzzz]

If you are not willing to shell out for multi-GHz differential probes, then I would agree that a resistive Z₀ probe is the way to go. Howard Johnson and Doug Smith have written about them, and there have also been posts (1 2) on this forum. Since they are DIY and the components are cheap, you can easily solder them down. Obviously you need a scope with 50 Ohm inputs (or less ideal a feed-through terminator).

There are balanced homebrew Z₀ probes, but I haven't personally tried one so I can't attest to their performance. See the references in https://entertaininghacks.wordpress.com/library-2/scope-probe-reference-material/ , especially http://emcesd.com/pdf/cd94scr.pdf

[David Hess]

Solder in or leave a spot for a printed circuit board to probe tip adapter.  These look like the old Tektronix printed circuit board Peltola coaxial sockets.  I know at least Tektronix and LeCroy sell these but you can also make your own with a collet socket pin which fits the probe tip and some bus bar wire.

The alternative that I sometimes use is to just build an active probe into the circuit test points which will drive a 50 ohm coaxial transmission line back to the oscilloscope inputs.

[joeqsmith]

Solder in or leave a spot for a printed circuit board to probe tip adapter.  These look like the old Tektronix printed circuit board Peltola coaxial sockets.  I know at least Tektronix and LeCroy sell these but you can also make your own with a collet socket pin which fits the probe tip and some bus bar wire.

The alternative that I sometimes use is to just build an active probe into the circuit test points which will drive a 50 ohm coaxial transmission line back to the oscilloscope inputs.

>1GHz active probe and driver on-board, I'm very interested!  I have looked into rolling my own DC coupled active probes with GHz plus BW for digital work.  I've never get past the paper and pencil stage.  Sounds like you are using readily available parts to pull it off.   

[David Hess]

>1GHz active probe and driver on-board, I'm very interested!  I have looked into rolling my own DC coupled active probes with GHz plus BW for digital work.  I've never get past the paper and pencil stage.  Sounds like you are using readily available parts to pull it off.

It just amounts to implementing a simple source or emitter follower to drive the 50 ohm coaxial cable back to the oscilloscope.  There is no real need for the unity or x10 attenuation that an active probe would support so just having a double termination which yields x2 attenuation simplifies things.

A complementary pair of bipolar RF transistor emitter followers is the easiest way and this is commonly used by active probes to drive the coaxial cable back to the oscilloscope.  Some "low" impedance active probes, like 10 kilohms, use bipolar input devices.  Alternatively a pair of stacked RF JFET or RF depletion mode MOSFETs can be used.

There used to be some suitable hybrid buffers which essentially implemented all of the above in one package but I am not aware of any modern integrated parts which have the same performance.

[joeqsmith]

Sadly your simple source follower seems out of my grasp.  If you have a layout for a driver that you like using  along with a parts list, I would be interested in seeing the details.   Any sort of data that you have where you proved it out would also be helpful. 

[David Hess]

Sadly your simple source follower seems out of my grasp.  If you have a layout for a driver that you like using  along with a parts list, I would be interested in seeing the details.   Any sort of data that you have where you proved it out would also be helpful.

I usually end up air wiring the probe into place as needed with whatever transistors are handy so I lack a standard layout.

The way Bob Pease implemented it is shown below in the first photograph.  The JFET stage is only needed for high input impedance; when probing a low impedance digital circuit, the base of the first bipolar transistor can be connected through a low value resistance to the signal.  The voltage drops of the two emitter followers roughly cancel.

The second photograph shows the old Tektronix P6202A active probe which is the same basic design and shows the changes necessary to drive a 50 ohm transmission line.  All that you need for an embedded version is Q22 and Q26.  Lower value resistors are used to allow driving the 50 ohm transmission line.  If extra high impedance is needed, then the JFET stage is added.