Selecting an oscilloscope probe

Thread Starter

Keebler

Joined Mar 28, 2020
14
I'm measuring the flow of a hall effect flow meter using an MCU and counting pulses. The value calculated by the microcontroller program did not agree with the calibration standard. The MCU value always read higher implying that it was counting more pulses. However, when I connected my scope to measure the frequency I got a value which aligned to the cal standard. And the MCU program also then recorded the correct number of pulses. After some help in another thread it was determined that the capacitance of the probe was making the circuit work by filtering noise. Some experimentation determined that about 100pF eliminated the noise in the circuit and didn't attenuate the frequencies of interest which are <800Hz

While the current problem is solved I want to get a probe that might be better in future similar situations. At 10X my current probe has about 18pF capacitance. I'm considering purchasing a 100X probe. I see some have about 6pF capacitance whereas the more expensive 1000X probes can have about 2-3pF capacitance. Considering my use case will these probes give me a better insight into noise in my circuits. In particular is it realistic to be able to think that the spurious pulses can actually be seen on the scope? Thanks.
 

nsaspook

Joined Aug 27, 2009
8,377
At the measurement ranges of your MCU/flow counter setup conventional scope probes will be fine.

But at higher frequency ranges, this is where active or low impedance probes are useful. These types of probes can provide high bandwidth will minimal input capacitance circuit loading. Good ones are expensive and may require special scope settings, inputs or adapters.

https://www.digikey.com/en/articles/understanding-selecting-using-active-oscilloscope-probes
https://teledynelecroy.com/probes/transmission-line-probes/pp066
 

Thread Starter

Keebler

Joined Mar 28, 2020
14
At the measurement ranges of your MCU/flow counter setup conventional scope probes will be fine.
Thx for the reply I appreciate it. I get what you are saying is that at low frequencies the capacitance of the probe is not really loading down the circuit. However, it is still doing something. Pulse counting with the MCU is done by edge detection. In this case rising edge. So shouldn't I be looking for high frequency ringing which might require a probe with lower capacitance?
 

nsaspook

Joined Aug 27, 2009
8,377
Thx for the reply I appreciate it. I get what you are saying is that at low frequencies the capacitance of the probe is not really loading down the circuit. However, it is still doing something. Pulse counting with the MCU is done by edge detection. In this case rising edge. So shouldn't I be looking for high frequency ringing which might require a probe with lower capacitance?
Your circuit was broken. The problem was not probe loading, it was edge detection signal conditioning. The fact it was working with the scope connected was just a lucky condition of low-pass filtering that should have been designed into the circuit originally by circuit components or digital processing.
 

Thread Starter

Keebler

Joined Mar 28, 2020
14
Your circuit was broken. The problem was not probe loading, it was edge detection signal conditioning. The fact it was working with the scope connected was just a lucky condition of low-pass filtering that should have been designed into the circuit originally by circuit components or digital processing.
I don't know how to ask this correctly, but I guess I want to know what kind of scope/probe setup do I need to see that in fact my circuit is "broken." I want to see the waveform with (what I assume) is the ringing. Is that a reasonable expectation? I'm not looking to fix the circuit. Its already fixed. I want to see the broken circuit on the scope.
 

MrChips

Joined Oct 2, 2009
23,504
I was going to say this but spook already said it. The problem is not with the scope probe. The problem is in your circuit. The scope probe is not going to fix the problem.
 

nsaspook

Joined Aug 27, 2009
8,377
I don't know how to ask this correctly, but I guess I want to know what kind of scope/probe setup do I need to see that in fact my circuit is "broken." I want to see the waveform with (what I assume) is the ringing. Is that a reasonable expectation? I'm not looking to fix the circuit. Its already fixed. I want to see the broken circuit on the scope.
I've given you what you need in my first post to view high frequency signal edge details with minimal loading. The interpretation and analysis of is it 'broken' and where is up to the operator even with the best equipment.
https://www.ece.ubc.ca/~robertor/Links_files/Files/TEK-Understanding-Scope-BW-tr-Fidelity.pdf

The valuable lesson for most is that raw MCU edge detection is designed for clean, one transition signals at the trigger points because most will trigger on very fast signals. Ringing (reflections of signal) will always be there unless the signal energy is totally absorbed at the receiver-end so there is nothing to reflect. Matched and terminated signal paths are one way to do this but the other way is to shunt or dissipate signal energy above the needed signal trigger frequency range so any possible reflections don't affect the signal edges. Usually a simple RC filter is all you need for this.

Never expect a perfect signal. Design your circuits and software to accept the widest possible range of crappy signals while outputting a near perfect signal.

Maybe you could talk deal with this guy for a capable system on the cheap. https://forum.allaboutcircuits.com/...d-help-before-i-smoke-it.176396/#post-1596692
 

jeffl_2

Joined Sep 17, 2013
44
I would like to suggest that if your circuit has been detecting spurious edges, most likely the correct resolution would be more like passing the circuit through a fast buffer first with a reasonable amount of hysteresis (>0.2V?), instead of "analog" filtering. (These MCU counter inputs are famous for doing as little "signal processing" as they can get away with.) You want to craft a circuit that "by definition" is as insensitive to input ringing as possible, to minimize having to worry about how it works with different cables and such in the first place.
 

Janis59

Joined Aug 21, 2017
1,332
I have a frequent need to meaure oscillo of about 3 kV on less of 100 MHz signals. So, first I used 1:1 with DIY resistor divider. But it was working deadly weak about form accuracy (genuine delays). So I took 1:10 and then resistors may be smaller. At last I bought 1:100 qualified for 5 kV and after 2 seconds it was dead. I bought more, some 5 different producer probes - all was dead immediately. Looking what is happening - their inner resistors are melting the outher layer, thus the Focault effect. Their capacitors have exploded thus the reactive current problem. Just those cheap end probes 1:100 ARE NOT be used for RF frequencies (for 20-50 USD per piece). They are NOT designed for that even if there is written in datasheet they are. Long about I am wishing to make my own probe what is optoisolated by glass fibre and capable up to 3 GHz, but all the time another more urgent works are breaking this idea down. If the 100 MHz of such cost in markets about 6000 per piece, would be good business, heh??
 

Janis59

Joined Aug 21, 2017
1,332
To appreciate the idea above: BFU-910 plus BF-998 based cascode LNA in input, feeding in constant current mode the LED emitter diode. The problem is that up to 1 GHz still such emitter exist, but over exist only digital not an analog. In receiver end - again the cascode LNA and passed transimpedance cascade to oscillo input. So, this moment the main stopper is proper open-fiber capable ultra-high speed analogue LED or laser transmitter for less than few hundreds USD. Receiver I have found.
 

andrewmm

Joined Feb 25, 2011
1,456
So
a 100:1 probe, will likely have lower capacitance, and higher bandwidth as you have seen,
but all the signals will be smaller.

The spikes that your MCU can see could be sub ns wide,
especially if your using edge triggering on the MCU and no s/w filtering in the MCU,

It sounds like your source is relatively high impedance, hence why it is affected so much by your probe,

This is about the hardest situation to measure , and get real results,
Your scope would need to have a many M ohm and low capacitance input , and very fast response on your scope,
your probably talking 5G plus sampling and GHz band width ,

This is one of those cases where as a designer, you design conservatively, as you know is a real pain to measure.

As its a high impedance source, you would us differential / twisted pair wiring, well away from other sources of noise, possibly with a screen, low capacitance wiring is critical.

As you know its high impedance, and low voltage, you know you are going to pick up spikes, thus one would either design in a low pass filter and / or use the low pass filtering in the MCU input if present,

some things are just to difficult to measure at a reasonable cost.
 

Ian0

Joined Aug 7, 2020
2,162
Oscilloscope input amplifiers are not always the quietest - they have a difficult job to do, so you can't always expect low noise as well, so if you attenuate your signal too much, you might lose it amongst the noise.
My suggestion would be not to look for a better probe, but to program you microprocessor to output the input signal on another pin. If the input was on a comparator, then you can often map the comparator output to a pin, then observe the voltage on that pin. Any jitter that the MCU has found and interpreted as a logic signal will then be a 3.3V signal that you can really see.
 

Thread Starter

Keebler

Joined Mar 28, 2020
14
This is about the hardest situation to measure , and get real results,
Your scope would need to have a many M ohm and low capacitance input , and very fast response on your scope,
your probably talking 5G plus sampling and GHz band width ,
Thx. That was the analysis I was looking for in the first place. Yes some things are just too hard.
 

Thread Starter

Keebler

Joined Mar 28, 2020
14
Oscilloscope input amplifiers are not always the quietest - they have a difficult job to do, so you can't always expect low noise as well, so if you attenuate your signal too much, you might lose it amongst the noise.
My suggestion would be not to look for a better probe, but to program you microprocessor to output the input signal on another pin. If the input was on a comparator, then you can often map the comparator output to a pin, then observe the voltage on that pin. Any jitter that the MCU has found and interpreted as a logic signal will then be a 3.3V signal that you can really see.
I get the basic gist, but I am not familiar with comparators. Can you give me a bit more detail please? I'm counting pulses. How do I know if any one particular pulse is noise? Thx.
 

Ian0

Joined Aug 7, 2020
2,162
I get the basic gist, but I am not familiar with comparators. Can you give me a bit more detail please? I'm counting pulses. How do I know if any one particular pulse is noise? Thx.
Are you counting in software, or using a counter-timer?
Is the input dierctly on GPIO or does it go via an different peripheral? Which MCU are you using?
 

Thread Starter

Keebler

Joined Mar 28, 2020
14
Are you counting in software, or using a counter-timer?
Is the input dierctly on GPIO or does it go via an different peripheral? Which MCU are you using?
Directly on the GPIO as rising edge that triggers an ISR which increments a counter. Currently I'm using an ATMEGA2560, but will be switching to an AVR128DB on a new board. I call a routine that uses an elapsed timer to convert the pulses counted every second into flow. The ISR counter is reset. Its not 100% precise (eg elapsed time is slightly more than 1 second) but excepting noise its good enough based on our calibrations.

I should mention that the output of the flow sensor is open collector and I pull it up to 5V. Interestingly, a different sensor from the same manufacturer with a higher flow range and PNP output doesn't seem to have noise issues. Also, as I said at the beginning I did solve the noise issue with a ~300pF cap. I'm just looking for techniques to troubleshoot this and similar problems in the future. Thx.
 

Ian0

Joined Aug 7, 2020
2,162
For the purposes of troubleshooting, could you get your counter to divide by two and output that signal on another pin?
They you will see immediately when it has mistriggered.

Does your GPIO have the option of hysteresis on the input? If so, switch it on. Do you use the Atmega's internal pullups?
A lower value of resistance might work better.
 

Thread Starter

Keebler

Joined Mar 28, 2020
14
For the purposes of troubleshooting, could you get your counter to divide by two and output that signal on another pin?
They you will see immediately when it has mistriggered.

Does your GPIO have the option of hysteresis on the input? If so, switch it on. Do you use the Atmega's internal pullups?
A lower value of resistance might work better.
I'm using an external 10K. Interesting. Can pull-up resistors pick up that much noise? Maybe my assumptions about how the mistriggering occurs might be wrong. I was thinking its like ringing where the pulse goes low but not to zero. Possibly to below V(IL) but higher than V(OH) No-mans land as it is called. In that case a stronger (ie lower value) pull-up might help?

I was considering a Schmidt trigger for the new board design but wanted to keep the complexity and cost down. I will see if the AVR128DB has this feature.

What does dividing by two do? I can certainly trigger a pin at whatever count I like. Thx.
 

Ian0

Joined Aug 7, 2020
2,162
What does dividing by two do?
The output will change state every time it picked up what it thought was a valid pulse, so you will be able to detect any mistriggers immediately on the scope. For a steady flow rate, you should see a nice squarewave, 50/50 mark-space. You will see instantly if a mistrigger is happening because it counts a pulse twice, or whether it is being triggered by other events where you can't see a pulse
Can pull-up resistors pick up that much noise?
10k is about what I'd use, possibly 4.7k. Where does the sensor get its 0V connection from?
 
Top