How fast of an O-scope do I need?

I have been spending late nights for the past week, reading about DSOs, industrial communications, etc trying to figure out how fast of a scope I need. Every time I think I've almost got it figured out, I read something that sends me back to square one.

I am going to use the scope for checking the electrical integrity of comms busses and VFD output. I figure the comms busses will be more demanding in terms of speed, so that's what I'm focusing on. I've convinced myself: 1. that o-scope manufacturers' published bandwidth specs are almost meaningless and I should focus on single-shot sample/second and 2. that I need a scope with 50X more samples/second than the bit rate of whatever comms bus I'm going to scope. I arrived at that little thumb rule by figuring that I need 100 samples per wave period in order to get a good single-shot look for glitches and ringing, and that there's a maximum of 1 wave period per 2 bits. BUT then there's various encoding techniques that can squeeze far more than two bits into a period, meaning I could get by with a lesser scope. I just don't know which Industrial comms schemes use which encoding, and what actual baud rates (my thumb rule would be more applicable to baud rates, if they were easy to find for today's tech) I will be looking at.

The highest bit rates I expect to see in the field are 12Mbits/s (max) for Profiibus DP and 10/100Mbits/s for ethernet. What's faster than profibus but slower than ethernet? Is anything faster than ethernet? According to my thumb rule I would need a 5GSa/S scope for the ethernet, but I suspect that's wrong.

Please tell me if I'm being stupid, and while you're at it, please tell me how fast of a scope I need.

Thank you.

Are you going to be designing the communication chips or just using them in purchased industrial equipment?

If you will just be using them, I'm not sure there is much need for a scope at all. Might be good to verify that a baud rate is set properly but definitely not necessary.

Now, having a good scope is a great thing but if you need it to debug a comm bus of industrial equipment, then I think you need to get your money back from whoever made the equipment.

I echo Norm's sentiments. Unless you're designing, I'm not sure you need a really high speed scope.

If I recall correctly, I believe the 2 points per waveform is the Nyquist theorem used to ensure that sufficient data is available when converting to the frequency domain, like a Fast Fourier Transform does. 2 points per waveform tells you almost nothing about distortion, noise or the actual waveform. For that, your 50 or 100 points per waveform shows a far better picture. For waveforms at line frequencies you get great images with even moderate speed scopes.

For power line measurements, the more important feature is isolation, to make sure you don't fry the scope input.

For communications, there's two diagnostic tools, an oscilloscope for wave form and a protocol analyzer for issues regarding content.

I have to ask, do you really need scope bandwidth for ethernet?

Ethernet is transformer isolated at the NIC card/port, which is one reason it's gained such enormous popularity - the absence of ground loops within its 100m/link range. I suspect that an analyzer tool like wireshark could be used to see that the data is garbage and deduce from that that there is such high noise that the data is distorted, without necessarily needing to capture a 100Mb waveform at 5GHz.

Think about what happens when someone runs into the problem of excessive noise on ethernet. Do they hire someone to establish what the noise looks like on the waveform and try to track down the source? Not in the plants I've been in. They first try shielded cable and if that doesn't work, then they go to fiber optic. It seems to me that the empirical method, shielding or FO, works because after all, what does a screen shot of a distorted ethernet wave form really buy you? The presence of noise is already an established fact - bad or really slow comm due to noise distortion. How does it help to get a picture of what the noise looks like? The picture doesn't tell you where the noise comes from or how to minimize it. I suppose a clean waveform can show that an ethernet problem is network related, not signal related, but won't looking at network traffic in wire shark show lots of repeated packets due to the noise?

The one or two people I've run into who do Profibus use a dedicated Profibus network analyzer, a handheld unit. Apparently, it incorporates a time domain reflectometer of sorts that can analyze segment lengths, a critical measurement for Profibus DP, which can run up to 12Mb baud rates. Profibus PA (for field instruments) runs at something just less than 40K baud. The one guy told me that the analyzer can locate unterminated segments. I'm not sure how you'd use a scope as a time domain reflectometer.

I recall several threads on this forum about using a scope on RS-485 comm links.

Several discussions about scopes on the Eng-tips forum about motor/drive work mention 200MHz and 50MHz digital scopes. One thread in particular warns about handheld Fluke issues with data storage and communications.

danw said:

I echo Norm's sentiments. Unless you're designing, I'm not sure you need a really high speed scope.

If I recall correctly, I believe the 2 points per waveform is the Nyquist theorem used to ensure that sufficient data is available when converting to the frequency domain, like a Fast Fourier Transform does. 2 points per waveform tells you almost nothing about distortion, noise or the actual waveform. For that, your 50 or 100 points per waveform shows a far better picture. For waveforms at line frequencies you get great images with even moderate speed scopes.

For power line measurements, the more important feature is isolation, to make sure you don't fry the scope input.

For communications, there's two diagnostic tools, an oscilloscope for wave form and a protocol analyzer for issues regarding content.

I have to ask, do you really need scope bandwidth for ethernet?

Ethernet is transformer isolated at the NIC card/port, which is one reason it's gained such enormous popularity - the absence of ground loops within its 100m/link range. I suspect that an analyzer tool like wireshark could be used to see that the data is garbage and deduce from that that there is such high noise that the data is distorted, without necessarily needing to capture a 100Mb waveform at 5GHz.

Think about what happens when someone runs into the problem of excessive noise on ethernet. Do they hire someone to establish what the noise looks like on the waveform and try to track down the source? Not in the plants I've been in. They first try shielded cable and if that doesn't work, then they go to fiber optic. It seems to me that the empirical method, shielding or FO, works because after all, what does a screen shot of a distorted ethernet wave form really buy you? The presence of noise is already an established fact - bad or really slow comm due to noise distortion. How does it help to get a picture of what the noise looks like? The picture doesn't tell you where the noise comes from or how to minimize it. I suppose a clean waveform can show that an ethernet problem is network related, not signal related, but won't looking at network traffic in wire shark show lots of repeated packets due to the noise?

Click to expand...

I do use wireshark and I agree it is a very useful tool, but it only works for ethernet. There are other busses I'm concerned with. I think my troubleshooting capacity, even in regards to ethernet, can only be increased by the addition of a scope. Yes, noise is something that can cause comms breakdown, but not the only thing. If I'm connecting to a comms bus with a scope, I'm looking for:
1. signal amplitude - too little terminations resistance results in too high voltage and ringing. Too low voltage indicates too many devices on the bus or too many terminations resistors on, etc.
2. Ringing/reflection - ringing can be caused by loss of termination resistance, spur lines, etc.
3. Noise, as mentioned.
4. Missing data - better captured with software, but still a bullet point.
5. Other stuff I don't know about yet.

It should be noted that all this stuff I'm spewing out is theory, gained from reading on the net. I've never actually connected a scope to a comms bus and therefore I wouldn't say I actually know what I'm talking about. I could be off base.

danw said:

The one or two people I've run into who do Profibus use a dedicated Profibus network analyzer, a handheld unit. Apparently, it incorporates a time domain reflectometer of sorts that can analyze segment lengths, a critical measurement for Profibus DP, which can run up to 12Mb baud rates.

Click to expand...

I've used one of these Softing Profibus Testers before and it was awesome. Vendor let me borrow it for a week. It was a really cool tool, but too expensive IMO to work on only one specific bus. I can't afford to buy this one alone, much less a similar device for each industrial comms bus. I need a more general purpose tool that can give me some of the same functionality. They give two specs for the softing unit; 192Msa/S and 8/16 samples/bit (16*12M=192M). So I guess that's a good number to start with. If I dealt with Profibus daily, this would be a worthwhile investment, as the software was also very good for troubleshooting the data layer.

danw said:

Profibus PA (for field instruments) runs at something just less than 40K baud. The one guy told me that the analyzer can locate unterminated segments. I'm not sure how you'd use a scope as a time domain reflectometer.

Click to expand...

check this out, slides 20, 21, 22. If I can get the hang of doing TDR measurements on the scope, I should be able to troubleshoot any of the busses (on the physical layer) without their individual kilo-buck testers. Together with a general purpose logic analyzer and possibly an array of software tools, I should be able to troubleshoot the data layer too. I'm looking at the Bitscope 325, which is a USB Mixed signal scope with 8 logic analyzer channels. It seems ideal in every way except the max sample rate is 40MSa/s, and I'm basically trying to figure out if that's going to be sufficient before I make the purchase.

danw said:

I recall several threads on this forum about using a scope on RS-485 comm links.

Several discussions about scopes on the Eng-tips forum about motor/drive work mention 200MHz and 50MHz digital scopes. One thread in particular warns about handheld Fluke issues with data storage and communications.

Click to expand...

I'll look those up. Thank you for your input.

How fast of an O-scope do I need?

It needs to be fast enough to capture the pulses of the fastest bus you want to work with. If you plan to monitor a Gigabit Profinet bus you will need one about 2 Gigahertz or better, preferably 10GHz.