Cross talk on logic analyzer leads

[Back2Volts]

After having an HP54645D MSO with no digital leads for over a year, and since the OEM cables are very hard to come by, I finally decided to buy one of these custom made cable assemblies

http://www.ebay.com/itm/Agilent-HP-10089A-Logic-Probe-with-Flying-Leads-for-54622D-Keysight-2000X-3000X/252788299528?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649

I am using it to trace a PSoC 5 (CY8CKIT-59) driving an SPI LCD.   In addition to the LCD signals CS, CLK, SDA (bidir-SPI) and A0/RS (command/data), I have modified the program to add some mark signals for debugging purposes,  TRIGR, CMDWR, and DTAWR.   

The problem I have is that there is some clear cross talk between some pins.  For example, in the attached picture there is a negative pulse in CMDWR that matches the drop in A0/RS.         

All MCU pins are defined with mode STRONG-DRIVE, FAST-SLEW and SINGLE-SYNC.   
All leads have the individual grounds connected to a breadboard GND bus, which in turn is connected to the KIT-59 GND and decoupled from VDD bus with 1uF.

Because the glitches seem to appear in the marking lines, with no load by the display,  I have tried adding 4K7 to VDD, but this has had no effect at all.

How can I improve my technique?   

Would other pin drive modes be better?     

Should I Should I suspect the quality of the cable ?
     
Any comments will be appreciated.
 

[alank2]

I bought the same kit from eBay and I didn't notice this issue, but I don't recall what I was driving it with when testing.  I've always had this problem on and off with logic analyzers though, sometimes there is cross talk.  In this configuration, I think there might be a ground between each signal line which should help, but I'm not for sure.  I can test it a bit later tonight if you want.

[Back2Volts]

I bought the same kit from eBay and I didn't notice this issue, but I don't recall what I was driving it with when testing.  I've always had this problem on and off with logic analyzers though, sometimes there is cross talk.  In this configuration, I think there might be a ground between each signal line which should help, but I'm not for sure.  I can test it a bit later tonight if you want.

Good to know about someone else using the same cable.

I am tying all the (optional?) lead GNDs to a breadboard GND bus.   Is that what you mean?    I also thought about leaving unused/grounded every other lead, but I would assume the ribbon cable already alternates ground and signal wires. 

[alank2]

I'm not seeing any crosstalk on what I tested.  I have a square wave coming out of a signal generator 1MHz, 10MHz, 20 MHz.  It is set for 1.4V offset with 500mV ampltitude and I have the HP on TTL (1.4).  I've got a male BNC to banana post with screw cap on each siggen output and I am using some breadboard wires to connect up CH0 and CH1.  I did connect the grounds of CH0 and CH1 as well - did you do this?  Each channel wire has a 2 inch socket coming off of a smaller thing that also has a ground in it, though it isn't always obvious. Mine says USA on one side and GND on the other and there is a female socket in it.

I tried turning one channel on and the other off, no crossover.  I put them out of phase with each other, no crossover.  I've enclosed some pics.  Hope this helps!

[alank2]

I also tried 0-5V on CH0, no crosstalk on CH1.  I then disconnected CH1 from the ground/signal, still no crosstalk.

[Back2Volts]

Thank you for taking the time to test.

Yes, I have all grounded except the CS(CH0) which I have just noticed when taking the pictures below.
I think I need to read about and play with the HP levels, hopefully later today.    May be it is a matter of settings.   

Thanks again,
Tony

[lukier]

That this looks like an ordinary IDC ribbon cable. Remember that HP logic analyzer pods have RC network in the podlets, matching termination in the scope/logic analyzer card and the HP cable is as far as I remember resistive lossy one (like the wire in the 10x probes) so things match.

These woven HP cables are not that easy to get, but sometimes one might find an old 16500 series logic analyzer card for next to nothing and the card will have 4 cables with IDC 40 pin connectors (one straight, one right angle). In my case I got 16510A card. I see people are selling the cables separately http://www.ebay.com/itm/321527618239

[Back2Volts]

Are you talking about this card ?    Do the cables insert properly in the 645D ?

http://www.ebay.com/itm/HP-16510A-25MHz-State-100MHz-Timing-Logic-Analyzer-Module-w-Cables-TESTED-/310755468377?hash=item485a782459:g:PGEAAOxyaqlSQdLc

The cable link is obsolete.  Are you talking about a cable like the picture  below?

[lukier]

Yeah, this type of card and cable. I don't know if it will fit 645D as I don't have one - I bought the cables and HP pods for a DIY logic analyzer, so the fact that one side of the cable is right-angle didn't matter.

I can check later if the POD connects well to the right angle (black connector) side, so then I guess the straight (grey) connector could be used on the deep MSO port.

[abraxa]

Does the "custom made cable assembly" even have line drivers on the PCB? It looks to me like it's just a simple connector adaptor PCB. In that case, I wouldn't be surprised to see all kinds of signal integrity issues.

That said, I don't think you need to set the pins in STRONG-DRIVE and FAST-SLEW mode. the slew rate and drive strength (= current) can be adjusted precisely for this reason: setting them too high is good for stable signals but bad for SI/EMC.

[Back2Volts]

Does the "custom made cable assembly" even have line drivers on the PCB? It looks to me like it's just a simple connector adaptor PCB. In that case, I wouldn't be surprised to see all kinds of signal integrity issues.

You are correct.   It is a rather thick piece of PCB with copper stripes between the soldered connectors.   No drivers or resistor networks.

[AndersJ]

Perhaps your analyzer is telling the truth.
Your circuitry may need better decoupling or grounding.
Have you verified your signals with a analog scope?
Are your analyzer input threshholds set properly?
Is your power supply good, no inappropriate current limiting?

[alm]

I have seem similar crosstalk issues with cheap logic analyzers using standard fly leads. That was solved by replacing the fly leads with coaxial leads. If you have connected the ground lead for each channel, then I would not expect crosstalk due to the cable. If you check the pin-out of the 40-pin pod connector (see Keysight Probing Solutions for Logic Analyzers datasheet page 71/72), every even wire is connected to a ground connection (assuming you connected the ground connection for each channel). So any crosstalk would happen in the connectors or adapter boards (which look like double sided boards without ground planes). I doubt that the edge rate is high enough for crosstalk to happen in those few centimeters. The fact that the impedance of the wire is different probably will not make that much difference at this length, unless you are pushing the limits of the pods. I believe these pods were used with logic analyzers up to 800 MS/s.

I suspect that this may be ground bounce, especially since you are using a solderless breadboard. If you search for the term than I am sure you can find a more detailed explanation, but in short ground bounce happens when the impedance of the ground lead is too high compared to the currents that are flowing to this. It happens, for example, if you have an IC with two dozen signal pins switching at very high speed but only one ground pin. The current of all those signal pins will have to flow back through the ground pin. This current flowing through the inductance can induce a voltage drop, raising the effective ground potential for that device. The way to solve this is to lower the impedance to ground. This is why fast parts, especially those with many pins, will tend to have many ground pins. And why the pod allows for one ground connection per channel, even though HP/Agilent/Keysight specifies that you may only need to use one ground wire every four channels or so if your edge rates are sufficiently low (so more inductance can be tolerated).

The single, probably not very low inductance/resistance, ground connection on your solderless breadboard may be responsible. Try lowering the inductance of the ground leads, for example by directly connecting them to the demo board, or at least connect more wires from demo board to the ground strip on the breadboard, one after every two or four wires to the logic analyzer pod.

The original passive pods were designed to work with much longer cable, but they were twisted pair, hence a well-defined impedance. I do not believe the woven cables are resistive, only the leads plugged into the pods are. I do not own an HP MSO, but from the pictures it looks like a 40-pin 100 mil connector on the scope end. That means that the cables for the 16500-series logic analyzers will not fit. See this auction for pictures of that cable. And even within that series there were multiple models matching different logic analyzers with differences in the housing of the connector. I believe the original cables+pods for the MSOs are quite rare. It is not unusual to find a MSO + leads for less than the leads alone.

[lukier]

The original passive pods were designed to work with much longer cable, but they were twisted pair, hence a well-defined impedance. I do not believe the woven cables are resistive, only the leads plugged into the pods are. I do not own an HP MSO, but from the pictures it looks like a 40-pin 100 mil connector on the scope end. That means that the cables for the 16500-series logic analyzers will not fit. See this auction for pictures of that cable. And even within that series there were multiple models matching different logic analyzers with differences in the housing of the connector. I believe the original cables+pods for the MSOs are quite rare. It is not unusual to find a MSO + leads for less than the leads alone.

OK, so I went to the attic and got the cables. The cables are resistive (signal carrying ones), as mentioned here:
https://www.eevblog.com/forum/testgear/dsox2000-and-3000-series-licence-have-anyone-tried-to-hack-that-scope/msg880066/#msg880066
and I've just measured that as well (~188 ohm).

16550A is a different card and yes it uses different connector. It is not beyond the wit of man to try to make one cable with IDC on both sides from two cables - in case of my cables the connectors are riveted which can be a problem.

Unfortunately, the short right angle connector (mating with the 16510A card) is too short to reach the pins of the pod (and I guess similarly the connector on the scope). In this case it could mate with this silly adapter PCB - not ideal but impedance mismatch wouldn't be that bad I guess.

I've attached the photos of the cable and how it mates with the pod - these are 40 pin 0.1" connectors on both sides, maybe with slightly different notches on the narrow sides.

[alm]

OK, so I went to the attic and got the cables. The cables are resistive (signal carrying ones), as mentioned here:
https://www.eevblog.com/forum/testgear/dsox2000-and-3000-series-licence-have-anyone-tried-to-hack-that-scope/msg880066/#msg880066
and I've just measured that as well (~188 ohm).

I stand corrected, thanks for doing the legwork.

16550A is a different card and yes it uses different connector. It is not beyond the wit of man to try to make one cable with IDC on both sides from two cables - in case of my cables the connectors are riveted which can be a problem.

I know that there were like half a dozen different part numbers for those cables depending on the analyzer, but I was not aware that the old ones used standard 40-pin connectors. That would make it viable to use them with an MSO with some minor adjustments (like the adapter board).

However, I am not sure that it would make that much of a difference in this case. The cheap flat cable seems to be much shorter than the original cables designed to go to the back of the big boat anchors. There is still a ground lead between every signal lead. Noise immunity and interference will be worse, and reflections will be worse due to both the (likely) impedance mismatch and the lower damping.

A possible test would be to connect the same number of channels, but spread them across the pod. I.e. use only channel zero and all even channels. Leave the odd channels floating. That should increase the distance between channels (and thus their capacitive coupling), while keeping the ground current the same.

[Back2Volts]

At this point I am thinking that alm is correct in pointing to ground/power supply.    I am feeding the PSoC KIT-59 from the host USB and the display from one of those cheap 3.3V/5V PS units that plugs at one end of the breadboard.   

Whenever I have a chance I will reconfigure the KIT-59 to not use the USB (have to remove a diode) and feed MCU and LCD from the same supply.

[alm]

I was mostly talking about the ground inductance between logic analyzer port and micro. Inductance of the power supply leads should be mostly compensated by the bypass caps on the devboard.

[Back2Volts]

...

That said, I don't think you need to set the pins in STRONG-DRIVE and FAST-SLEW mode. the slew rate and drive strength (= current) can be adjusted precisely for this reason: setting them too high is good for stable signals but bad for SI/EMC.

Before changing the PS I gave it a few more tries:

- I added a couple more lines from other GND MCU pins to the breadboard GND
- Increased decoupling by the LCD
- Set the mode of all the signals going to the LCD to RESISTIVE-PULL-UP-DOWN and SLOW-SLEW.

The first two resulted in partial improvement.    The mode change was the final fix.

I may still try going with single PS for MCU and LCD.

Thank you all.   This forum is great!
Tony
 

[rstofer]

How wide is that negative pulse?  It seems as though it might be a few us or so.  Are you absolutely certain there isn't an actual signal there?

[alm]

I would put a good 10x probe from the scope as close as possible to the MCU pin with a short ground wire also directly connected to the devboard (which I assume has a solid ground plane). Possibly even a spring ground clip around the ground sleeve of the probe. That is why you get a mixed-signal scope, right ? That should exclude any signal integrity issues on the breadboard and in the logic analyzer cable.

[Back2Volts]

I would put a good 10x probe from the scope as close as possible to the MCU pin with a short ground wire also directly connected to the devboard (which I assume has a solid ground plane). Possibly even a spring ground clip around the ground sleeve of the probe. That is why you get a mixed-signal scope, right ? That should exclude any signal integrity issues on the breadboard and in the logic analyzer cable.

thinking...   will a $18 (pair) Chinese probe fit that description ?

[Back2Volts]

I changed the pin drive mode back to strong/fast to capture the glitch pin in the analog channel.    Unfortunately I only get a slow rising pulse with no corresponding glitch.

The analog channel corresponds to the CMDWR signal