What are those spikes after transients?

[tggzzz]

More specifically the equivalent sample rate, whatever that might be.  Like my old TDS460 has a 100MSps rate but reports up to I think 50GSps at the highest sweep (however, it fills in DREADFULLY slowly, by random sampling).

... and is still only a 400MHz scope 

[Jwillis]

I'm not an expert on this by any means . Only just begone to delve into this area of electronics  . By the micro chip label I'm guessing It's a AVR128 series chip . 
This looks a lot like a bread board technique issue . Connections to the chip are to long with minimal decoupling . Decoupling and filtering needs to be a close to the chip as possible . There are a lot of very long wire connections that create a lot of parasitic inductance . Found this out with similar devices  . Any supply decoupling needs to be as close to the chip pins as possible to reduce current looping.

General Guidelines
http://ww1.microchip.com/downloads/en/DeviceDoc/40002183A.pdf   Page 20 Part 4

http://ww1.microchip.com/downloads/en/appnotes/atmel-1619-emc-design-considerations_applicationnote_avr040.pdf

http://ww1.microchip.com/downloads/en/AppNotes/AN2519-AVR-Microcontroller-Hardware-Design-Considerations-00002519B.pdf

Just my take on this .
I'm learning to so if anyone could chime in I would appreciate the input .

[StillTrying]

... and is still only a 400MHz scope 

The 100MHz CML+'s ETS does up to 50 GSa/s per channel, 250 samples per division, there's no difference between between dots and sinx.   BW is around 150MHz in ETS mode.

[David Hess]

With a probe configured like that even 100 MHz performance is questionable so with a 3.5 nanosecond transition time, sample rates beyond say 300 Msamples/second do not have much meaning, and in practice much lower sample rates are completely sufficient.

Before I learned better, I saw waveforms like that with breadboard construction and a 100 MHz analog oscilloscope but a 100 MHz DSO with 10 nanosecond peak detection from a 100 Msample/second sample rate (Tektronix 2232) or even 20 nanoseconds and 50 Msamples/second would work just fine, and without the need for ETS at 2 Gsamples/second or higher.

I think the sweat spot is 200 MHz bandwidth with a suitable sample rate because that is just at the limit where high impedance passive probes are still viable and probing is not too difficult, although spring clips or coaxial connections are still needed to get full performance.

Does that waveform matter?  It does not in a prototyping environment, and even in a production environment it would be unlikely to cause problems, but I would not accept it considering that two or three decoupling capacitors would fix it.  I have seen worse in real products.

[BrianHG]

If you want the absolute minimum transient:

Solder a 0.1uf ceramic cap right onto the driving IC's VCC & GND, then place the IC into the breadboard with the output pin bent out not going into the breadboard.

On your probe's GND spring contract, press it into the same hole as the IC's GND pin and touch the IC's output you wish to measure.

[uliano]

This looks a lot like a bread board technique issue . Connections to the chip are to long with minimal decoupling . Decoupling and filtering needs to be a close to the chip as possible . There are a lot of very long wire connections that create a lot of parasitic inductance . Found this out with similar devices  . Any supply decoupling needs to be as close to the chip pins as possible to reduce current looping.

I rearranged the breadboard using much shorter wire for any "signal" connection (here just UPDI & UART) while "DC" connections (supply rails, pushbutton and LED) have been left as they were.

I added 100n across each supply injection into the chip and also across FTDI breakboard (which is the supply source for the board).

I could have shortened some *but not all* supply connections, instead I went along with the (possibly very wrong) thinking "it's just DC".

All in all nothing (significant) changed since the coil barrel GND probing has been introduced some post above.

[tggzzz]

I rearranged the breadboard using much shorter wire for any "signal" connection (here just UPDI & UART) while "DC" connections (supply rails, pushbutton and LED) have been left as they were.

I added 100n across each supply injection into the chip and also across FTDI breakboard (which is the supply source for the board).

I could have shortened some *but not all* supply connections, instead I went along with the (possibly very wrong) thinking "it's just DC".

All in all nothing (significant) changed since the coil barrel GND probing has been introduced some post above.

You need to understand that a wire is not a wire: it is a component which can be modelled as series resistance, series inductance, and parallel capacitance to other wires. The implication is that there is no such thing as "ground" where everything is at the same voltage. In this case the most important element will be the series inductance.

So, what is the total length of copper in the loop between probe signal tip to pin output to pin power supply to probe shield? That will determine the series inductance. It is useful that you included a decent photo, since it appears that there may be a long blue wire in that loop.

The best probing technique is to have the probe's tip and coil barrel shield connected to the IC pins. The best ground is a solid copper plane. Both of those can (with care) be achieved using "manhattan" prototype techniques, but are more or less impossible with solderless breadboards.

[uliano]

So, what is the total length of copper in the loop between probe signal tip to pin output to pin power supply to probe shield? That will determine the series inductance. It is useful that you included a decent photo, since it appears that there may be a long blue wire in that loop.

There are many "points" that can be assumed as ground: Earth, AC plug ground, PC PS supply ground, the - coming out from the serial breakboard, the blue line at a side of the breadboard, the (two) pins connected to gnd of the MCU...

To which ground should I minimize the path from coil tip? To the ground of the chip? To the "digital" one (the one that is in the upper middle in the picture) as the output pin is in the "digital" domain?

The best probing technique is to have the probe's tip and coil barrel shield connected to the IC pins. The best ground is a solid copper plane. Both of those can (with care) be achieved using "manhattan" prototype techniques, but are more or less impossible with solderless breadboards.

I'm reaching awareness that I'seeing a probing epiphenomenon, and won't certainly step up the prototyping technique for a simple MCU circuit.

[sicco]

uliano,

You’ve still not disclosed to us how long the wires are… Share a picture of the entire setup. So a picture that shows also where the other end of your grey ribbon cable sits.

Share a scope screenshot also of the same setup after disconnecting all of those grey ribbon cable signal wires. As in no longer physically connected to the digital outputs that you are looking at.

Your scope likely has more than one channel. Measure with the second channel what your chip GND (VSS) pin looks like. Same bandwidth, same V/DIV, same everything. GND probe clip on the same spot as your first channel.
If you have more scope channels, then do the same for the VDD pin of the chip.

[uliano]

the gray ribbon is the UPDI connection with the ICE, maybe 15 cm long

https://www.microchip.com/en-us/development-tool/ATATMEL-ICE

which is connected to an USB hub toghether with the USB cable coming from the red breakboard with the FTDI chip.

[tggzzz]

So, what is the total length of copper in the loop between probe signal tip to pin output to pin power supply to probe shield? That will determine the series inductance. It is useful that you included a decent photo, since it appears that there may be a long blue wire in that loop.

There are many "points" that can be assumed as ground: Earth, AC plug ground, PC PS supply ground, the - coming out from the serial breakboard, the blue line at a side of the breadboard, the (two) pins connected to gnd of the MCU...

None of which will be at the same potential. Even the earth won't be at the same potential as the earth: consider the case where someone is standing with feet apart on the ground and there is a nearby lightning strike. There can be sufficient potential difference that current flows up one leg and down the other!

To which ground should I minimize the path from coil tip? To the ground of the chip? To the "digital" one (the one that is in the upper middle in the picture) as the output pin is in the "digital" domain?

Whichever one matters for that part of the circuit

For outputs you consider the current flow, so in this case it wll be the digital 0V associated with the output.

For inputs you consider the points at which the receiver will interpret the analogue voltage[1] as a digital signal. For "simple" signals that will be the digital 0V associated with the input. For differential signals it will be the difference between the two inputs, but you can't put a scope's shield on either of them and still have to put the shield on the digital 0V associated with the input.

[1] ignoring using currents to convey digital signals

The best probing technique is to have the probe's tip and coil barrel shield connected to the IC pins. The best ground is a solid copper plane. Both of those can (with care) be achieved using "manhattan" prototype techniques, but are more or less impossible with solderless breadboards.

I'm reaching awareness that I'seeing a probing epiphenomenon, and won't certainly step up the prototyping technique for a simple MCU circuit.

It probably is a probing problem, but are you sure  All that matters is what the other components in the circuit "see" and how they "interpret" it

[uliano]

None of which will be at the same potential. Even the earth won't be at the same potential as the earth: consider the case where someone is standing with feet apart on the ground and there is a nearby lightning strike. There can be sufficient potential difference that current flows up one leg and down the other!

I know, I was just making fun of myself.

To which ground should I minimize the path from coil tip? To the ground of the chip? To the "digital" one (the one that is in the upper middle in the picture) as the output pin is in the "digital" domain?

Whichever one matters for that part of the circuit

For outputs you consider the current flow, so in this case it will be the digital 0V associated with the output.

I tried to shorten the wire to ground but, alas, to no avail 

The wire is coming out directly from the digital gnd.

[Jwillis]

Just a novice idea.
Did you configure the I/O pin output drive strength of the FTDI ? Disable  the high drive level on the UART and CBUS I/O pins . High drive currents can induce over and under shoot . Also try adding a resistor between the FTDI and AVR128 .

[uliano]

The pin I'm probing comes out of the MCU and goes into the FTDI not the other way around.

I gave a quick look at the datasheet and, if I didn't miss anything, for output pins I can only configure a slew rate limit (of a whole port).

I tried inserting a resistor (I had no clue so I randomly choose 47 Ohm, if you have your preferred value I can test) between MCU and FTDI and again nothing really changes: 13%overshoot 18% undershoot.

[T3sl4co1l]

Heh, looks worse, wonder if that's because of common mode.  Hard to say of course as these are nanosecond things, it's hardly a pixel at this scale.

Tim

[uliano]

Heh, looks worse, wonder if that's because of common mode.  Hard to say of course as these are nanosecond things, it's hardly a pixel at this scale.

Tim

sorry, I didn't post the zoomed-in view. Worse here means half a microsecond! 

[tggzzz]

Heh, looks worse, wonder if that's because of common mode.  Hard to say of course as these are nanosecond things, it's hardly a pixel at this scale.

Tim

sorry, I didn't post the zoomed-in view. Worse here means half a microsecond! 

Looks like ~10MHz/100ns "ringing", which is a relatively long period.

If you have a clean ~1ns risetime source, it would be interesting to see what that shows.

[T3sl4co1l]

Most likely stuff with that kind of time constant, is couple-meter wires from PS to board, scope probe, uh USB if this thing is connected, etc.  I'd expect bypass caps to be slower, and immediate signal quality things to be faster (~ns, and probably not visible on the scope at all, or at least very well).

It also keeps ringing for a while, which suggests a high Q, low loss -- if it's bouncing between cables and boxes, it's doing so for a fairly long time.  In that case, a ferrite bead somewhere will definitely cut down on the duration, though not so much the peak amplitude right after the edge (except if it's CM up the probe, in which case it may).

Tim

[uliano]

Look also at reply #16 where i didn't pull out the wire gnd directly from the pin but from breadboard rail.

There the time constant was quite shorter.

[iMo]

Do not expect perfect edges with your setup.
You work with a breadboard with 3.5pF parasitics.
Your blocking capacitors are foil - not suitable for decoupling in your case - you have to use ceramics instead.
Also mind 10mm of a straight wire is 10nH.
Considering that your ringing seems to be normal..

PS: foil capacitors are good for audio or "low" frequency apps.
For decoupling of fast transients in digital circuits you want capacitors with low parasitic inductance - the best are smd ceramic capacitors (like 10nF, 100nF), or through-hole one..

[sicco]

Is it an ordinary probe, 10x, 10 Mohm (1M after the scope’s BNC, 9M in the probe), and a long coax transmission line between scope and probe head?

If yes, then the transmission line (the probe’s coax) has an open end (1Mohm is not what the coax characteristic impedance can be…). Whatever was propagating from probe to scope would bounce back from scope to probe. And in the probe that 9M is another open end, so all that bounced at the scope again bounces back. And that’s what you’re seeing here quite likely. It’s not really ringing as in an LC circuit. It’s echoing, it’s standing waves.

Do your spikes disappear when you use a 1x probe, and switch to 50 ohms input for the scope channel? (add a small C in series because your chip output probably collapses when DC loading with 50 ohms)
Does the ‘ringing’ period change if you add a BNC extension between probe BNC and scope BNC?

To confirm, connect the same probe and scope to a perfect square wave generator (50 ohms source, terminated, probe at the termination 50 ohms). Do you still see ~10% amplitude ‘ringing’?

see also https://phet.colorado.edu/en/simulations/wave-on-a-string and give a unit step with the wrench. Wrench is your probe input. Other end of the string is selectable (fixed, free, no end) which translates to, short circuit, 1M or 50 ohms termination at your scope’s BNC input.

[perieanuo]

those spikes measured with the barrel mod are small enough for that spider pcb, with careful routing you're fine

[sicco]

http://jahonen.kapsi.fi/Electronics/DIY%201k%20probe/

on how to get a better probe.

[tggzzz]

Is it an ordinary probe, 10x, 10 Mohm (1M after the scope’s BNC, 9M in the probe), and a long coax transmission line between scope and probe head?

If yes, then the transmission line (the probe’s coax) has an open end (1Mohm is not what the coax characteristic impedance can be…). Whatever was propagating from probe to scope would bounce back from scope to probe. And in the probe that 9M is another open end, so all that bounced at the scope again bounces back. And that’s what you’re seeing here quite likely. It’s not really ringing as in an LC circuit. It’s echoing, it’s standing waves.

You don't understand how 10* high impedance probes work! It isn't an ordinary transmission line, but is a very lossy transmission line for that reason. Having said that, if the probe is excessively badly manufactured, I suppose they may have skimped on that part.

(In addition, a 1* probe is inappropriate for this)

For further details, see the references I gave earlier in this thread.

[sicco]

Ah, tggzzz has alternative facts…

So what then is the characteristic impedance of your ‘very lossy transmission line’ in your probe cables? 1 Mohm??