DIY Logic Analyzer Probe and Pods for Siglent (and LeCroy) scopes

[Old Printer]

Wish there was a version for the 1104x-E 
Don't really want to spend the extra $900 to upgrade to the 1204 plus, though that is a beautiful screen.

[oz2cpu]

to old printer : Siglent SDS1104X-E  dont have Logic Analyzer

---

 if we where playing with analog signals, cross talk is much more visible, and important to avoid.
but a logic analyzer is DIGITAL, and it is alot more imune,
I tried many different signals at the same time, from 10 to 500MHz at the fastest rise times my generators can create,
and I am not able to detect any cross talk.
so it is really noth worth the trouble and hard work, to use coax, it is a nightmare to solder, and much more stiff..
just take what ever thinn wires you got, and twist them.. done deal..
IMPORTANT : cut all wires to exactly the SAME length, look at the PCB's i made, all tracks are exactly the same length, and as short as possible,
and spaced as much as possible, and made so their individual capacity, inductance, cross talk, is as little as possible, and as even channel to channel as possible,
when running tests at 500MHz with 1nS rise time, it is barely visible the different response i get on the scope,
and some of it could just as well be inside the scope, it is that little.

[Old Printer]

to old printer : Siglent SDS1104X-E  dont have Logic Analyzer

It is supposed to have MSO with a $109 firmware option, which I believe is hackable, and a $329 16 Channel Logic Probe set of leads. Maybe it does not have the active components built into the scope as the 1204+ does??

[oz2cpu]

I just checked its specs, cant even find the LA connector, on that scope

[2N3055]

I just checked its specs, cant even find the LA connector, on that scope

There is S_bus connector next to USB.. It uses external device that does acquisition, that is not passive and you have to buy it... It works but is not as powerful as 2000X+ with fully integrated MSO.
If MSO is really needed, instead of buying SLA1016 for 350 €, I would put that money forward to upgrade to SDS2000X+. And if you sell SDS1104X-E  you recoup some money too.
SDS2000X+ is whole different level..... If you can  afford it, that is clear winner...

[mawyatt]

Recall somewhere that the LA for the lower level Siglent scopes had significant active circuitry, thus more difficult to DIY implement.

Agree with 2N3055, with the Siglent SDS2000X+ and with the Thomas DIY LA (completely passive) you've got a really good setup. Don't worry too much about the twisted pair or coax for the LA, both should perform well, even the Ribbon Cable we used seems to work well at lower speeds. Of course if you are trying to squeeze the best overall performance for the LA, then not only the cable type but the impedance networks in the LA base and POD must be tuned, but this is an ongoing effort we hope to get back on soon 

Just used the SDS2102X+ to show results from a mathematical solution of a coupling network (don't want to clutter this thread with details unless others are interested), what an absolute jewel of an instrument IMO. Being an old retired EE that has used about every analog scope Tek ever made, I was naive about the capabilities of these new MSOs. Since getting the Siglent, the pair of Tek 2465s haven't been used much, the Siglent is that good and right away became a very useful instrument

Best,

[Johnny B Good]

 I'm quite impressed with mine but I do have a couple of minor criticisms. The most significant being the 46 seconds boot up delay versus the 16 seconds of the SDS1202X-E and the not quite so significant yet much higher than expected increased power consumption (54W versus the 22W of the 1202).

 In view of the almost 2 1/2 times greater energy consumption, suggesting the use of a more power hungry faster processor, that significantly protracted boot up time is a rather surprising shortcoming. I can only assume the extra channels and larger feature set have added to the initialisation burden with additional time constraints that no amount of extra CPU grunt can mitigate - the price of progress in this case, perhaps? 

 I hope this isn't the start of a trend leading us back to the days of those Tektronix 500W space heaters so cunningly disguised as test and measurement kit labelled "545A Oscilloscope". 

[oz2cpu]

if boot up time is an issue Johnny,
try this : Tenma 72-8225a 50mhz 2 Channel Digital Oscilloscope With 500ms/s Sampling Rate

it takes 4 sec, from power on, THEN it is fully ready to rock :-)
I have not seen anything beat this.
Sold it, got a Rigol,
sold the rigol, got a Siglent.
Every time I got alot more bootup time.
but really not an issue for me, I used to work with R&S RTO scope, it is a few mins of bootup and a loud storm of fan noise

[2N3055]

I'm quite impressed with mine but I do have a couple of minor criticisms. The most significant being the 46 seconds boot up delay versus the 16 seconds of the SDS1202X-E and the not quite so significant yet much higher than expected increased power consumption (54W versus the 22W of the 1202).

 In view of the almost 2 1/2 times greater energy consumption, suggesting the use of a more power hungry faster processor, that significantly protracted boot up time is a rather surprising shortcoming. I can only assume the extra channels and larger feature set have added to the initialisation burden with additional time constraints that no amount of extra CPU grunt can mitigate - the price of progress in this case, perhaps? 

 I hope this isn't the start of a trend leading us back to the days of those Tektronix 500W space heaters so cunningly disguised as test and measurement kit labelled "545A Oscilloscope". 

Keysight MSOX-3104T boots 1 minute, 45 seconds.. That's 105 seconds...
 Enjoy your very fast booting scope...

[Johnny B Good]

 @oz2cpu

 Thanks for offering that perspective. I guess it's normal to not only expect more performance and features for the extra money but also more boot time and power consumption too! The thing was, I just wasn't expecting to get quite so much of either.

 Intrigued, I googled "Tenma 72-8225a 50MHz" only to find SFA information in the two page 'datasheet' provided by Farnell for their UNI-T rebrand which they're now selling the 1Gsps version for a mere £197.20 (VAT inclusive price) which is half the typical Ebay sellers' price point (the cheapest being £320). It's unusual to say the least, to find Farnell undercutting competing sellers by any amount let alone by more than 50%!

 If anyone's in the market for an 'entry level 'scope, they certainly won't be buying a Tenma 72-8225a from any of these sellers

https://www.ebay.co.uk/sch/i.html?_from=R40&_trksid=p2334524.m570.l1313&_nkw=Tenma+72-8225a+50MHz&_sacat=0&LH_TitleDesc=0&_osacat=0&_odkw=Tenma+72-8225a+2nd+hand

 Nor from this one

https://uk.farnell.com/tenma/72-8225a/oscilloscope-2-ch-50mhz-500msps/dp/2499522?mckv=soDoPBV12_dc|pcrid||plid||kword||match|e|slid||product|2499522|pgrid|1357897483847209|ptaid|pla-4588468182313071|&CMP=KNC-GUK-GEN-SHOPPING-SMEC-WHOOPS-Test725-minROAS-Step1&gross_price=true&msclkid=a71687df1b5f1abf0db7e775b3d6aa13

There's some very peculiar price variations being applied to this Tenma 72-8225A 'scope.

 I can see the appeal of a 'scope priced at just £197 to the cash strapped hobbyist whose needs are rather modest but I can't figure out the appeal when they're being priced at £334.57 and higher. Surely, anyone checking out this budget end of the 'scope market will have already seen the better alternatives from Rigol and Siglent in their searches? I mean  Who would be so stupid as to pay those prices for a UNI-T 'scope sullied by the Farnell branding?

[EDIT]

 For those whose budgets just won't stretch to Rigol or Siglent kit, there's always the  Hantek DSO2D10 available from BangGood priced at just £164.95 (plus £9.24 shipping fee to the UK from their CN warehouse). Cheaper again with 1GSa/s 100MHz BW plus a built in single channel AWG (25MHz max sine) with nicer looking on screen fonts.

https://www.banggood.com/Hantek-DSO2D10-Digital-Oscilloscope-2CH+1CH-Digital-Storage-1GS-or-s-Sampling-Rate-100MHz-Bandwidth-Dual-Channel-Economical-Oscilloscope-with-Signal-Source(AWG)-p-1765904.html?cur_warehouse=CN&rmmds=category

 I spotted this in one of today's BangGood junk mails and decided, in view of my contributions here, to take a closer look. It's actually quite well specced for the money (it even includes an extra 9 to 10 second's worth of boot up time  ). Indeed, if it had been available at that price two and a half years ago when I'd been looking to upgrade a 5MHz BW boat anchor that was already ancient when I'd bought  it some 40 years ago, I may well have purchased that instead as my very first modern day DSO.

 Trying to track down any meaningful reviews on this Hantek led me to a how to hack it to 200MHz BW youtube video with a link to a 125 page EEVBlog thread originated way back in October 2010 active right up to the 11th of this month!

https://www.eevblog.com/forum/testgear/hantek-tekway-dso-hack-get-200mhz-bw-for-free/msg21170/#msg21170

 for anyone curious enough to take a look.

 Even at the reduced price, I can't really see Farnell selling many (if any) of these rebranded UNI-T DSOs. 

[Johnny B Good]

I'm quite impressed with mine but I do have a couple of minor criticisms. The most significant being the 46 seconds boot up delay versus the 16 seconds of the SDS1202X-E and the not quite so significant yet much higher than expected increased power consumption (54W versus the 22W of the 1202).

 In view of the almost 2 1/2 times greater energy consumption, suggesting the use of a more power hungry faster processor, that significantly protracted boot up time is a rather surprising shortcoming. I can only assume the extra channels and larger feature set have added to the initialisation burden with additional time constraints that no amount of extra CPU grunt can mitigate - the price of progress in this case, perhaps? 

 I hope this isn't the start of a trend leading us back to the days of those Tektronix 500W space heaters so cunningly disguised as test and measurement kit labelled "545A Oscilloscope". 

Keysight MSOX-3104T boots 1 minute, 45 seconds.. That's 105 seconds...
 Enjoy your very fast booting scope...

 Oh, don't worry about that. Now, thanks to both oz2cpu's and your contributions, I surely will! 

[mawyatt]

Enjoying these SDS2XXX+ scopes indeed!!

Just finished using the superb Bode Plot feature to help evaluate a few HV Op Amps, then leveraging of Thomas's excellent DIY LA Probe to trace down a bug in the Analog Devices software/DAC hardware test PCB.

For the bug tracing this required finishing up the version of the DIY LA Probe that utilizes a 3M controlled impedance twisted pair cable with connectors since using standard ribbon cable could create false "glitches" with the SCI signals and wanted to make sure we were tracing the real bug/glitch rather than a cable induced glitch. This LA Probe allowed us to find that the AD software wasn't issuing a correct LDAC command and why the DAC output voltage wasn't being updated.


Best,

[oz2cpu]

nice with a little bit LA user-case-stories

please people, do tell when you use LA, and what it did for your project
post pictures, and screen shoots :-)
we love to hear about when the Logic Analyzer feature is so fantastic.

[mawyatt]

Ok, here's a few images of the setup and the LA Probe using the twisted pair cables with connectors on each end. The component values for the POD are input R of 250 ohms followed by 90.9K in parallel with 8.2pF, the BASE is 360 ohms series R to scope. The shunt elements in the base are not installed and these are for "tuning" the response for maximum BW and good pulse shape, but this must wait until time permits.

The cases are 3D printed but still learning this process, so not as nice as could be

Anyway, works nicely with our "enhanced" SDS2102X+ and really like the display and SPI serial bus decoding. Agree with Rob at tautech, having a few more analog traces along with the 16 bit digital traces of the LA should prove valuable when we get to the multiple DAC outputs of the present project.

This is my first use of a LA and must say hats off to Thomas and Siglent for making this experience worthwhile and useful

Best,

[Slartibartfast]

However, we can verify that Thomas (oz2cpu) LA Probe works well with standard 10 line ribbon cable (8 active, 2 ground), the twisted pair cable we have is a little too wide so used ribbon cable instead. Don't have a high speed pulse generator, but did verify that using a sine wave input from a AWG the Probe works to ~100MHz with ribbon. You do "see" a small amount of crosstalk, but that's expected, with twisted pair the crosstalk should be eliminated.

I wonder ...

At RF, inductive, capacitive and even radiative coupling are so much worse than at AF (audio frequency), that RF guys frequently neglect thinking about galvanic coupling. Usually that can be justified if a massive, low-impedance ground plane can be used. Here, we do not have such a ground plane. In fact, the two concepts discussed in this thread have TWO separate ground planes (one in the pod, one in the scope adapter), connected by a set of ground lines. I'm not sure whether that concept is a smart idea, I think a significant part of the crosstalk may be due to galvanic coupling.

What happens if a logic signal edge travels along signal line, say, #3? Some current will flow in the line, and some return current will flow in the ground line. Which ground line? After all, all ground lines of a pod connect the same two ground planes. Well, the ground line that is associated with the signal line, i.e. ground #3, will carry the majority. The reason is that this ground line out of the eight forms the lowest inductance loop with the signal line. But in the same way as a 10k resistor parallel to a 1k resistor also carries a smaller part of the current, the other ground lines will also carry a part of the return current that ought to travel via ground #3. I'd expect that part of the return current to cause crosstalk to the signals on lines #2 and #4, and, to a lesser degree, to the other signals.

For this reason I think it may be smarter to keep the grounds of the various signals separate inside the pod. IMHO this would reduce the crosstalk that you measured. The final purpose of a logic analyser however would in most cases have the grounds connected inside the DUT anyway, so I'm not sure how much difference it makes in terms of the real-world application of the device.

Cheers  Peter

[Slartibartfast]

Mike,

I really like your concept with the detachable pods. I'd envision that I'd more often use the LA with the less than 8, rather than more, channels. Having only one cable lying around would be useful.

Maybe I missed it, but did you post your design files?

Cheers  Peter

[ElectronicsHobbyist]

Hi Peter

All design files are on thingiverse (See first post on this topic).
Im about to order them right now by the way. Thanks very much to oz2cpu and all the other developers for developing and publishing this great piece of equipment.

[mawyatt]

However, we can verify that Thomas (oz2cpu) LA Probe works well with standard 10 line ribbon cable (8 active, 2 ground), the twisted pair cable we have is a little too wide so used ribbon cable instead. Don't have a high speed pulse generator, but did verify that using a sine wave input from a AWG the Probe works to ~100MHz with ribbon. You do "see" a small amount of crosstalk, but that's expected, with twisted pair the crosstalk should be eliminated.

I wonder ...

At RF, inductive, capacitive and even radiative coupling are so much worse than at AF (audio frequency), that RF guys frequently neglect thinking about galvanic coupling. Usually that can be justified if a massive, low-impedance ground plane can be used. Here, we do not have such a ground plane. In fact, the two concepts discussed in this thread have TWO separate ground planes (one in the pod, one in the scope adapter), connected by a set of ground lines. I'm not sure whether that concept is a smart idea, I think a significant part of the crosstalk may be due to galvanic coupling.

What happens if a logic signal edge travels along signal line, say, #3? Some current will flow in the line, and some return current will flow in the ground line. Which ground line? After all, all ground lines of a pod connect the same two ground planes. Well, the ground line that is associated with the signal line, i.e. ground #3, will carry the majority. The reason is that this ground line out of the eight forms the lowest inductance loop with the signal line. But in the same way as a 10k resistor parallel to a 1k resistor also carries a smaller part of the current, the other ground lines will also carry a part of the return current that ought to travel via ground #3. I'd expect that part of the return current to cause crosstalk to the signals on lines #2 and #4, and, to a lesser degree, to the other signals.

For this reason I think it may be smarter to keep the grounds of the various signals separate inside the pod. IMHO this would reduce the crosstalk that you measured. The final purpose of a logic analyser however would in most cases have the grounds connected inside the DUT anyway, so I'm not sure how much difference it makes in terms of the real-world application of the device.

Cheers  Peter

Don't think this is a galvanic coupling issue, it's simply signal edge coupling to adjacent lines due to dynamic fields. With twisted pair or coaxial lines there is little EM field outside the pair or coaxial line, and thus little coupling to adjacent lines. A few simple tests proved this once the twisted pair LA was assembled and tested, no coupling whatsoever to any lines under any signal conditions we could generate.

We are using this twisted pair LA now without any issues, coupling or otherwise. It works beautifully with either the SDS2102X Plus or SDS2104X Plus (both enhanced) without issue

Best,   

[mawyatt]

Mike,

I really like your concept with the detachable pods. I'd envision that I'd more often use the LA with the less than 8, rather than more, channels. Having only one cable lying around would be useful.

Maybe I missed it, but did you post your design files?

Cheers  Peter

I didn't't post my design files, they are different than Thomas's files. I use the single 8 bit input mostly, in fact have yet to find a need for the full 16 bits use, and could even use the single bit lines now since I'm sensing SPI lines.

Best,

[Slartibartfast]

I use the single 8 bit input mostly, in fact have yet to find a need for the full 16 bits use, and could even use the single bit lines now since I'm sensing SPI lines.

That is exactly what I anticipate for my usage scenarios too, hence my preference for detachable pods and wiring.

So I'll do my own design then. Should not be much work given all the information on the circuitry you have provided in this thread.

You have correctly pointed out that oz2cpu's in line resistors at the base do not terminate, but it seems in your design you also only use in-line resistors, instead of parallel-connected terminators. Is that work left for the future, or is there some background to it that you did not explain in the thread?

[mawyatt]

I use the single 8 bit input mostly, in fact have yet to find a need for the full 16 bits use, and could even use the single bit lines now since I'm sensing SPI lines.

That is exactly what I anticipate for my usage scenarios too, hence my preference for detachable pods and wiring.

So I'll do my own design then. Should not be much work given all the information on the circuitry you have provided in this thread.

You have correctly pointed out that oz2cpu's in line resistors at the base do not terminate, but it seems in your design you also only use in-line resistors, instead of parallel-connected terminators. Is that work left for the future, or is there some background to it that you did not explain in the thread?

I did some simulations of what I thought the scope input looked like, and was able to get a close pulse response to what Rob at tautech had shown (measured, see earlier in this thread). First off the twisted pair is not terminated with the characteristic impedance (100~130 ohms depending on the cable) on either end, this would load the input probe too much and attenuate the signal too much at the scope end. The cable is AC terminated with an RC network on each end that allows a reasonable input impedance as "seen" for the device being probed and doesn't attenuate the signal too much for the scope to sense. The network I did allows this with select Rs and Cs. However, I have not had time to create the optimum network RC values just yet, will do so later when time allows, so for the beginning just included the series resistors and series capacitor.

Here's what the schematic and PCBs look like. I got 2 types of twisted cable off eBay for ~$25 each, the one I've used is like the 3M type mentioned in the thread, the other is a larger cable but should also work. Also have a few PCBs somewhere, but really don't have the time now to pull all this together (in middle of a major project which is benefiting from this LA), later in a couple months. PM if you want more details or files if I can get them easily.

Best,

[mawyatt]

Few more.

Best,

[Slartibartfast]

here you go, i think TK was the one to first provide the pinouts, he shared this file attached,

There is something wrong here. TK seems to have messed up the side designations.

PCIexpress standard documents state the "lower side" of the board (i.e. "solder side") is "Side A", while the  "upper side" ("components side") is "Side B". This coincides with the orientations of the PCBA sides when looking down the PCIe card onto the PCIe socket, and placing it such that the 11-pin section is at the left, with pin counting from left to right (in a PC, this amounts to placing the I/O shields to the left). (edit: used to be with PCI but incorrect for PCIe)

TK has called the lower side "Side B" and the upper side "Side A", in contradiction to the standard. Depending on whether the table was meant to give the layout looking into the socket, or onto the edge connector, either could be right. So, in the two header lines in the pin table posted by TK, which is the correct line? The one denoting A vs. B, or the one lower vs. upper?

Since we're dealing with a scope that clearly defines directions up and down I suspect I can trust the "Upper" vs. "Lower" designation, and "Side A" vs. "Side B" are chosen arbitrarily (opposite to standard). Can somebody confirm, or state what is correct?

Edit: Just realised the fact that in a typical tower case PC the PCBA is mounted upside down (component side pointing down) may be the source of the confusion. If that is true then indeed TK assigned A and B correctly, and with "upper" and "lower" he just referred to the sides as referring to the scope's "up" and "down", different from the PCIe standard which has the component (upper) side down, and vice versa.

https://web.archive.org/web/20111102042843/http://www.adexelec.com/faq.htm#pcikeys

[oz2cpu]

if you look at the pictures of the pcb layouts in post 1
you will see what is what,
also the pcb is impossible to flip in the 3D printed case, and the edge connector and case is impossible to stick into the scope wrong.
you are welcome to reinvent it all, once again,
but you can also just download the free available ordering zip, send it to JLC or similar place
3D print the shells, and be done in a week,
it is a well tested design, we are many users now on this design.
here is a picture from last week, when i designed a PID system for a robot motor controller

[AlexDavidson]

Thanks to the great work by oz2cpu and the boards he sent me, I’m now able to use the digital channels on my SDS2000X+. I had no problems printing the cases and assembling everything, and it all seems to work as intended. A note to others using this solution: the wires used for cabling shouldn’t be more than 0.75mm diameter otherwise you will have trouble soldering to the boards and fitting into the cases. I used twisted pairs removed from HDMI cable as suggested earlier in this thread.

The main reason I wanted to enable the digital channels was to assist in debugging communication problems in a project with 6 daisy-chained SPI devices (AS5048A encoders). Using the LA probe I was able to connect 4 digital channels to the master SPI, freeing up the analog channels to look for noise etc along the chain & power supply rails. I used both pods so I could set 2 different thresholds, because some of the signals were 5v & others 3v3. At one point I also connected digital channels to the MISO/MOSI interconnections along the daisy chain to check that data was being shifted along correctly, but this is not shown in the capture below.

For those interested, the trigger for the capture below was MISO bit 14 set, which happens to be the AS5048A error flag bit. Decoding was set to use clock timeout because CS only changes every 6 packets. S1 & S2 are both decoding the same data - one is set to hex display, the other to binary to make it easy to spot bit patterns. The analog channels were connected to CS & CLK as a cross-check for glitches, and C3 was connected to an MCU output pin used for debugging.

Following the trigger, which in this case came from the 4th encoder in the chain, code in the project sends a UDP packet to a host PC, which is why SPI data stream pauses for about 200 us before transmitting the 5th and 6th packets. Next the code sends 0x4001 six times to clear the error registers in the AS5048As, before resuming sending 0xFFFF, which is the read angle command.

It took a while, but I’m happy to report that I did eventually sort out the comms problems with all those extra channels to play with. (Those AS5048A encoders seem very fussy when it comes to output loading, and very susceptible to noise on the clock, which is a bit annoying since they are likely to be used in a noisy environment, e.g. with stepper motors as in my case.)

I did however notice a few issues with the Siglent along the way, mainly to do with SPI triggering and decoding, which didn’t always seem consistent. After changing something, say the SPI trigger conditions or channel mapping, I would sometimes have to play around with thresholds or cycle the SPI settings to get it working correctly. Sometimes the decoding was clearly incorrect (not just the last bit issue mentioned in the bug thread), and sometimes it would trigger on the incorrectly decoded data; other times not. Part of this might have been due to the rats nest of connections to the project or me, however I didn’t have the same issues on my Keysight MSOX3000 series scope when I tried connecting it in parallel with the Siglent (making an even bigger rat’s nest).

Yes it would be nice if the digital channels didn’t have these issues, but we can’t ignore that the Siglent is excellent value for money, even more so with the DIY LA probe.