Difficulty when using oscilloscope to observe Ethernet signal

[houkensjtu]

Update: For those of you who may interested in the result, I have actually successfully captured several sections of a Ethernet packet, each was shifted by utilizing the "Delayed time base". And by gluing them altogether, I actually saw the waveform of 56 bits of preamble and 8 bits of SFD.
Here is the link : https://goo.gl/photos/3EQojfnYpNPd4fWz7
Hope you enjoy it too!

Hi all! I've been trying to use my old digital oscilloscope (Tektronix TDS 340AP) to observe an Ethernet signal.

Basically what I have done is only to force the NIC (Network Interface Card) to work at 10Base-T (10Mhz). Thus the signal is basically 10Mhz and encoded by the so called "Manchester Code", which is pretty easy to interpretate. You can check the wave form I got from the below link:

https://goo.gl/photos/3EQojfnYpNPd4fWz7

The problem is, as described in Wikipedia (https://en.wikipedia.org/wiki/Ethernet_frame), every ethernet packet has a 49bit 56 bit "preamble" header. No matter however I change the setting of trigger function, the oscilloscope will always trigger at the first slope of the preamble header signal. Since the preamble is supposed to be simple repeat of 0 and 1, I really want to "skip" the preamble part and trigger at a later timing. More importantly, since the memory depth of my scope is limited, it's impossible for me to capture the whole signal of a ethernet packet.

My question is, is there any setting or trick can make it possible to delay the trigger timing?

[jboard146]

I've not used that particular scope, but what you are looking for is "Trigger Holdoff". According to quick google search for the users manual on Pg 3-13 at the very bottom of the page. I've attached a screen shot of the crappy copy of a copy I found.

[nctnico]

Hold-off is not what you want because that is a forced dead time between 2 triggers. What you want is delayed trigger which should be possible by just scrolling the section you don't want off screen using the horizontal position knob.

[kripton2035]

I just noticed that the scopes that trigger on ethernet signals are very high end keysight scopes ... like the one actually in the contest
so may be it's quite difficult to achieve ?

[stmdude]

I just noticed that the scopes that trigger on ethernet signals are very high end keysight scopes ... like the one actually in the contest
so may be it's quite difficult to achieve ?

That's a whole different kind of triggering. That's triggering on specific contents in the ethernet packets, so it does demodulation and protocol decode for the trigger.
I.e, you can (for example) set up a trigger for all packets with an ethernet source-address of BA:DC:AF:EB:AB:E0.

[nctnico]

If you want to look at the packet contents then you are much better off with software tools like Wireshark. I'd never use an oscilloscope to do ethernet decoding.

[stmdude]

If you want to look at the packet contents then you are much better off with software tools like Wireshark. I'd never use an oscilloscope to do ethernet decoding.

Wireshark cant look at layer1 stuff. A scope can.
Usefull if you're, for instance, designing network gear

[Keysight DanielBogdanoff]

Maybe you could rig up something on the breadboard that sends a trigger-in to the scope after the preamble?

[nctnico]

If you want to look at the packet contents then you are much better off with software tools like Wireshark. I'd never use an oscilloscope to do ethernet decoding.

Wireshark cant look at layer1 stuff. A scope can.
Usefull if you're, for instance, designing network gear

That is a valid point but it would mean you'd be involved in designing semiconductors so it is very unlikely there isn't a budget for specific tools.

[houkensjtu]

Hold-off is not what you want because that is a forced dead time between 2 triggers. What you want is delayed trigger which should be possible by just scrolling the section you don't want off screen using the horizontal position knob.

You are totally right. But as I said, the memory depth is very limited on this scope, I can only get the very begining part of a packet, and sadly most of them are occupied the preamble. That's why I want to delay the trigger.

[houkensjtu]

If you want to look at the packet contents then you are much better off with software tools like Wireshark. I'd never use an oscilloscope to do ethernet decoding.

Wireshark cant look at layer1 stuff. A scope can.
Usefull if you're, for instance, designing network gear

Yeah that's the point! Except that I am not designing network gear, I was just curious

[nctnico]

Hold-off is not what you want because that is a forced dead time between 2 triggers. What you want is delayed trigger which should be possible by just scrolling the section you don't want off screen using the horizontal position knob.

You are totally right. But as I said, the memory depth is very limited on this scope, I can only get the very begining part of a packet, and sadly most of them are occupied the preamble. That's why I want to delay the trigger.

What happens if you scroll the horizontal position off-screen (in run mode!) ? I'm not sure about the TDS300 series but I think you should see the rest of the packet due to delayed triggering.

[borjam]

Wireshark cant look at layer1 stuff. A scope can.
Usefull if you're, for instance, designing network gear

Still, if debugging physical problems, an eye pattern is likely much more useful than capturing whole packets. Capturing whole packets makes of course would help detect stuff that an eye pattern won't, like packets cut short, etc.

[Jeroen3]

If I recall correctly Manchester code guarantees an edge. So you just set your trigger to trigger after 22 triggers. 

[tautech]

My question is, is there any setting or trick can make it possible to delay the trigger timing?

Not sure how complex your trigger suite is but if you had them:
Pattern trigger
Interval trigger
Pulse trigger

Any or all could be of use to help overlook the preamble and then trigger on subsequent packets of interest.

[rf-loop]

My question is, is there any setting or trick can make it possible to delay the trigger timing?

Not sure how complex your trigger suite is but if you had them:
Pattern trigger
Interval trigger
Pulse trigger

Any or all could be of use to help overlook the preamble and then trigger on subsequent packets of interest.

With ancient TDS300 series scope. No Way.  Afaik, it do not have any advanced trigger ffunctions. Also it have old analog trigger system.  Also need note that it have up to  1000 points (yes 1k) sampling memory.

[houkensjtu]

With ancient TDS300 series scope. No Way.  Afaik, it do not have any advanced trigger ffunctions. Also it have old analog trigger system.  Also need note that it have up to  1000 points (yes 1k) sampling memory.

Seems that's the truth. Too sad if I have to give up here...

[houkensjtu]

If I recall correctly Manchester code guarantees an edge. So you just set your trigger to trigger after 22 triggers. 

I'm afraid that there is no such functionality on my TDS340AP. Maybe plugging in some external trigger source can do the trick, but I still have no idea.

[macboy]

If I recall correctly Manchester code guarantees an edge. So you just set your trigger to trigger after 22 triggers. 

This would be the solution, if this old Tek didn't try so very hard to be just like an analog scope. It has no capability to trigger after N trigger events. It also has no simple trigger delay to start the capture N divisions or microseconds before or after the trigger event. It just has basic slope and level controls, plus holdoff. I have an old Philips digital scope from around the same era (probably slightly earlier) and it does have the event/N trigger and delayed (or advanced) triggering. It does not have holdoff, which I think is absurd. Back then, adding features like that usually required additional hardware, not just software, so I guess they had to pick and choose.

One way to achieve delayed trigger is to split (or tap) the signal, and send part through a delay line before entering the external trigger input of the scope. The "delay line" is nothing more than a coax cable, cut to appropriate length to give the required delay. This is more commonly used to pre-trigger (sending the signal directly to trigger input, and through the delay before the channel input), but it works fine the other way too. The speed of propagation of the signal through the coax will be approximately 2/3 that in air, or 200 000 000 m/s. You want about 6 us delay (about 60 bits at 10 Mb/s) so... about 1200 meters of cable. Oh well, it was a thought anyway. For this large amount of delay, you might be able to design and build a delayed trigger pulse circuit of some kind.

[houkensjtu]

One way to achieve delayed trigger is to split (or tap) the signal, and send part through a delay line before entering the external trigger input of the scope. The "delay line" is nothing more than a coax cable, cut to appropriate length to give the required delay. This is more commonly used to pre-trigger (sending the signal directly to trigger input, and through the delay before the channel input), but it works fine the other way too. The speed of propagation of the signal through the coax will be approximately 2/3 that in air, or 200 000 000 m/s. You want about 6 us delay (about 60 bits at 10 Mb/s) so... about 1200 meters of cable. Oh well, it was a thought anyway. For this large amount of delay, you might be able to design and build a delayed trigger pulse circuit of some kind.

This seems a little bit difficult for dumb like me, but I got the idea!
So I need to research for a while how I could delay a digital signal for exactly several micro seconds.

[Fred27]

Would a logic analyser be a simpler and more flexible choice for the job?

[PA0PBZ]

Just AC couple the signal, rectify it and charge a cap. Then trigger on the voltage level of the cap using external trigger or CH2?

[nctnico]

This is more commonly used to pre-trigger (sending the signal directly to trigger input, and through the delay before the channel input), but it works fine the other way too. The speed of propagation of the signal through the coax will be approximately 2/3 that in air, or 200 000 000 m/s. You want about 6 us delay (about 60 bits at 10 Mb/s) so... about 1200 meters of cable. Oh well, it was a thought anyway. For this large amount of delay, you might be able to design and build a delayed trigger pulse circuit of some kind.

With all this effort it would make more sense to get a different oscilloscope with at least pre/post trigger and perhaps much more memory. Ethernet signals up to 100Mbit have a frequency content not higher than 125MHz so a reasonably priced scope will do fine.

[femtohertz15]

Just browsing a service manual, I think TDS340 has Delayed Sweep capability.   Should do the trick.   Looks like a zoom tool in the Tek menus. but special:  Trigger, wait <x> amount of time, capture. 

[siggi]

Just browsing a service manual, I think TDS340 has Delayed Sweep capability.   Should do the trick.   Looks like a zoom tool in the Tek menus. but special:  Trigger, wait <x> amount of time, capture.

Yups, looks like the 340 has delayed time base. Tap the horizontal menu button, select "Time Base" from the soft key below the screen, then pick "Delayed Only" from the side menu. You'll then be able to dial in the delay from trigger event to capture.