As a firmware developer without a background in electronics, this video was very insightful to things I have seen before but did not understand. Most excellent!
I work in repairs and run into exactly same problem with replacement ATmega chips, as that company with Z80 boards. I kind of workaround it with some capacitors, but would never understand it if not for you (i don't have much of the scope, basically i just supposed to turn screws in and out). Now i try to calculate proper terminators for lines in question. Many thanks and definitely a thumb up!
Well done, Jack. I was just thinking you should mention that a poorly grounded probe can make it look like your board has a problem, when really it is nothing more than the way you are probing it. But it sounds like you are going to cover that in your next video. What I would like to hear you talk about--and you may well have already covered it--is what sort of things you do when you have to push beyond the limits of your equipment. If you had to debug a DDR3 system running at a couple of GHz with your 500 MHz scope, you would have to do some smart things in your embedded firmware to debug the system. In my industry 10 Gbps communications are the norm, but 16-28 Gbps single-channel communications are coming quickly and boy are the test instruments and fixtures expensive. Doing more with effectively less capable test equipment from the perspective of embedded firmware will always be a hot topic in my book. Again, well done.
You can't please everybody! I can see that from many earlier comments. I'd have to actually WORK to try and figure out who's right about what! I came here from watching your interview with David Jones in Sydney Australia on one of his EEVblogs. I used to read your articles all the time, so it's nice to reconnect in a new way. Oh yes, A. R. Jasso reminds me, I knew I could never do an EE job, because I was wayy too slow doing the "simple algebra" we got when solving a differential equation (not sure of the order 8-)) in controls class. I was thinking of Fourier series / transform right from the start. That quick rise time was more like an impulse than the slower one.
Oddly enough you made me remember that I have actually encountered the same problem with a poorly laid out PCB and fast logic IC transitions you described. It must be quite a common one by the looks of it. But we did not redesign anything, instead we ended up using a somewhat dirty trick to overcome the issue. When I undrstood what was going on the board I decided to just cut the traces right near the output pins of the ICs and inserted 470 ohm 0402 resistors in order to just confirm that it would fix the issue. And of course combined the capacity of the traces it was well enough to slow the signal down and pretty much eliminate any of the "ringing". The proper way would be to correctly redesign the board with modern components and technology in mind but it was a very old and obsolete soviet era digital device with very limited use so we wanted to get over it with as little time and effort as possible so we just documented my "solution", added the pads for the resistors and called it a day. If you have an excessivly high performance logic output then it is not a problem to slow it down. Going the other way though is nowhere near that easy.
One thing I'd add to the question why software people should be aware of this topic: see, in many cases the pin's slew rate can be controlled from software. For example, the STM32 microcontrollers have an option in GPIO control registers, which allows to choose pin's speed grade. So programmer is really half-step away from physical effect. :)
You're not alone. I studied biology and chemistry. My day job is in a medical lab but on my free time I fix electronics. Thinking about doing some RF to millimeter stuff myself. I don't like limiting myself to one thing in life.
There is a "tale" about a magic switch in some computer (it might be retold in the Hacker's Dictionary, I am not sure). One day a switch is found glued to the side of a minicomputer. It has to labels, Magic, and More Magic. It is set at More Magic. Only a single wire goes from the switch into the computer. The guy who found it flipped it - and the computer crashed. It was later discovered to be connected to the ground (of the metal frame).
I enjoyed your video. I'm not an EE, but I could follow this video some what. Could you explain the two resistors (220 Ohm and 270 Ohm ) at time 9:51. How does that work? What does it do to the impedance? Thank you.
This sounds like something useful to look at even with GPIO pins, like on a Pi or an AtMega setup. Especally on the Pi that is down in the CMOS range.....The books looks like an essential read too....
Hey know, some software/firmware guys still know about EM theory and FFT's ;). The new MDO's out there are king for helping with these termination mismatch problems. It'll never replace a true SA but those (*edit* MDO's /not/ MSOs -- you want real analog)Tek's and Agilents (or Keysight or whatever they're called now) are darn good in these situations
I knew it. There it is.... "transition time". There is no discontinuity, discrete increments anywhere. The line is solid being analogue, square, trapezoidal or whatever. Yes time is involved. You can't avoid moving through time axis no matter what. You can have horizontal line but you are not allowed straight verticle. If it was allowed, the line will not be there at all.
It's relatively easy to destroy a board or circuit you are probing if project's power supply, scope or signal are not mains insulated. especially when you connect your probe ground to other a reference point on a circuit other than its common ground. I am just a hobbiest but this was an interesting video. I haven't thought about fourier series in a while.
Interesting. In my case I'm about to make an untuned wide band loop antenna with an LNA. Would love to find a neat way for all frequencies to "see" the same impedance through all the frequencies of interest. This would improve cross modulation?
The formula shown at 1:50 is wrong in many ways. Jack suggest replacing the B on the right-hand side with E. So far, so good. But that's only the curl of B in free space when there is no current around. What's missing is the current density over e-naught times c^2. To get the whole formula, find the Maxwell Equations in "The Feynman Lectures Volume 2", Chapter 18. It's publicly available at: www.feynmanlectures.caltech.edu/II_18.html 73 de DL5ULF
How does electromagnetic wave in electric energy form propagate in DC power supply since there are no oscillations. What does an oscilloscope show, one that is really sensitive? Can photons propagate without wave form?
Oops, sounds like you let the smoke out. I've done it. I was working on a high end $$$ transistorized amp, and when I put my scope probe in the circuit, it killed it. No smoke, just no amp.
is it a common stereotype that engineers aren't familiar with the mathematics and physics side of things? i definitely harbor a love for the science, too.
Lol. Love your stats on that claim. My dad is 50 and re-did his law degree in the UK and emigrated there. How old do you think senior researchers are? What is the average age of a publishing academic? Please send us a link to your paper: "How The World Works" Haha
Yes. In fact, the older my professors, the better they are. There is a reason why older engineers are paid an order of magnitude more than graduates ...
Nice video. i love to lelarn, butt its hard. Jusst like you sead it in Australia, at dave`s plase : it takes to long time to whatch videos. butt im loving it, so thaks and god bye here from Norway. im a New Fan. Thanks
Think of it as a variable resistor. You must first know your probes impedance. Use short lines and try to keep everything in balance between ground and the subject pin.
Fun Fact: Your equation remains wrong. Either it should be E on the right side leading to Ampere's law or rotE=-delB/delt (minus and without epsilon and µ) leading to Faraday's law ;)
Good subjects, but poor explanations. First, a lot of the ringing shown in the first waveform was probably due to the inductance of the scope probes ground lead, not reflections in the traces. Second, the reflections are not cause be a mismatch between the "load" and "source" impedances; the are caused by the load and/or source impedances not matching the trace impedance. So, pretty sloppy work here.
+Steve Robbins Any tutorials/video's or chapters of a book to cover your better explanations? cause I am more of a firmware guy and sort of killed boards probably some way or another and this seem to indictate the problems of my former way of connecting to things. (poor tutorials like "take this TTL usb adapter and just connect it" .. wich may work, but with logic analysers and scopes....)
As a firmware developer without a background in electronics, this video was very insightful to things I have seen before but did not understand. Most excellent!
I work in repairs and run into exactly same problem with replacement ATmega chips, as that company with Z80 boards. I kind of workaround it with some capacitors, but would never understand it if not for you (i don't have much of the scope, basically i just supposed to turn screws in and out). Now i try to calculate proper terminators for lines in question. Many thanks and definitely a thumb up!
Well done, Jack. I was just thinking you should mention that a poorly grounded probe can make it look like your board has a problem, when really it is nothing more than the way you are probing it. But it sounds like you are going to cover that in your next video.
What I would like to hear you talk about--and you may well have already covered it--is what sort of things you do when you have to push beyond the limits of your equipment. If you had to debug a DDR3 system running at a couple of GHz with your 500 MHz scope, you would have to do some smart things in your embedded firmware to debug the system.
In my industry 10 Gbps communications are the norm, but 16-28 Gbps single-channel communications are coming quickly and boy are the test instruments and fixtures expensive. Doing more with effectively less capable test equipment from the perspective of embedded firmware will always be a hot topic in my book.
Again, well done.
You can't please everybody! I can see that from many earlier comments.
I'd have to actually WORK to try and figure out who's right about what!
I came here from watching your interview with David Jones in Sydney Australia on one of his EEVblogs. I used to read your articles all the time, so it's nice to reconnect in a new way.
Oh yes, A. R. Jasso reminds me, I knew I could never do an EE job, because I was wayy too slow doing the "simple algebra" we got when solving a differential equation (not sure of the order 8-)) in controls class.
I was thinking of Fourier series / transform right from the start. That quick rise time was more like an impulse than the slower one.
Oddly enough you made me remember that I have actually encountered the same problem with a poorly laid out PCB and fast logic IC transitions you described. It must be quite a common one by the looks of it. But we did not redesign anything, instead we ended up using a somewhat dirty trick to overcome the issue.
When I undrstood what was going on the board I decided to just cut the traces right near the output pins of the ICs and inserted 470 ohm 0402 resistors in order to just confirm that it would fix the issue. And of course combined the capacity of the traces it was well enough to slow the signal down and pretty much eliminate any of the "ringing". The proper way would be to correctly redesign the board with modern components and technology in mind but it was a very old and obsolete soviet era digital device with very limited use so we wanted to get over it with as little time and effort as possible so we just documented my "solution", added the pads for the resistors and called it a day.
If you have an excessivly high performance logic output then it is not a problem to slow it down. Going the other way though is nowhere near that easy.
One thing I'd add to the question why software people should be aware of this topic: see, in many cases the pin's slew rate can be controlled from software. For example, the STM32 microcontrollers have an option in GPIO control registers, which allows to choose pin's speed grade. So programmer is really half-step away from physical effect. :)
I'm a biologist, not an electrical engineer, but I understand some of the basics here and I like being out of my field. Very interesting.
You're not alone. I studied biology and chemistry. My day job is in a medical lab but on my free time I fix electronics. Thinking about doing some RF to millimeter stuff myself. I don't like limiting myself to one thing in life.
There is a "tale" about a magic switch in some computer (it might be retold in the Hacker's Dictionary, I am not sure). One day a switch is found glued to the side of a minicomputer. It has to labels, Magic, and More Magic. It is set at More Magic. Only a single wire goes from the switch into the computer. The guy who found it flipped it - and the computer crashed. It was later discovered to be connected to the ground (of the metal frame).
What an educational video this was for me; keep up the great work Mr. Jack Ganssle.
I enjoyed your video. I'm not an EE, but I could follow this video some what. Could you explain the two resistors (220 Ohm and 270 Ohm ) at time 9:51. How does that work? What does it do to the impedance? Thank you.
Great you decided to do videos Jack! Will be another great hit like the Muse.
I was waiting for you to show up at youtube. thanks for the newsletter and the video Jack :)
This sounds like something useful to look at even with GPIO pins, like on a Pi or an AtMega setup. Especally on the Pi that is down in the CMOS range.....The books looks like an essential read too....
Very clear and interesting!
Thank you Mr. Ganssle.
Great video- thanks for sharing your knowledge.
Excellent video with excellent explanation. Keep up good work. Regards,
Great video - looking forward to more!
Thanks for posting this. It was a very interesting and educating video. Hope to see much more from you in the future.
Wow this was a fantastic and very informative video! Thanks!!!
Hey know, some software/firmware guys still know about EM theory and FFT's ;). The new MDO's out there are king for helping with these termination mismatch problems. It'll never replace a true SA but those (*edit* MDO's /not/ MSOs -- you want real analog)Tek's and Agilents (or Keysight or whatever they're called now) are darn good in these situations
Excellent video, keep them coming!
Great stuff! Keep it coming!
Subscribed after watching your video with EEVBLog.
I knew it. There it is.... "transition time". There is no discontinuity, discrete increments anywhere. The line is solid being analogue, square, trapezoidal or whatever. Yes time is involved. You can't avoid moving through time axis no matter what. You can have horizontal line but you are not allowed straight verticle. If it was allowed, the line will not be there at all.
It's relatively easy to destroy a board or circuit you are probing if project's power supply, scope or signal are not mains insulated. especially when you connect your probe ground to other a reference point on a circuit other than its common ground.
I am just a hobbiest but this was an interesting video. I haven't thought about fourier series in a while.
Great explanation. Thanks.
Interesting. In my case I'm about to make an untuned wide band loop antenna with an LNA. Would love to find a neat way for all frequencies to "see" the same impedance through all the frequencies of interest. This would improve cross modulation?
I am not an embedded systems engineer although I am an electrical engineer. I still find these videos interesting.
I hope you used Kepner Tregoe Problem Resolution to narrow down that Z80 board problem. lol
The formula shown at 1:50 is wrong in many ways. Jack suggest replacing the B on the right-hand side with E. So far, so good. But that's only the curl of B in free space when there is no current around. What's missing is the current density over e-naught times c^2. To get the whole formula, find the Maxwell Equations in "The Feynman Lectures Volume 2", Chapter 18. It's publicly available at: www.feynmanlectures.caltech.edu/II_18.html
73 de DL5ULF
How does electromagnetic wave in electric energy form propagate in DC power supply since there are no oscillations. What does an oscilloscope show, one that is really sensitive? Can photons propagate without wave form?
Im 14 and im a hobbyist with an oscilloscope 4x my age how do you expect me to know what that equation is
+Jeremiah Lowe xD
***** yep
Great video!
very good tutorial
Good video!
well done
I love it! I think I have a new favorite show!
Oops, sounds like you let the smoke out. I've done it. I was working on a high end $$$ transistorized amp, and when I put my scope probe in the circuit, it killed it. No smoke, just no amp.
is it a common stereotype that engineers aren't familiar with the mathematics and physics side of things? i definitely harbor a love for the science, too.
very informative
great video!!!
Brilliant!
Im in 8th grade and planning to go to either ict(progranning) or engineer
The "Equality" sign (at 1:37)??!?
Cool. I'm glad I chose maths instead of EE. Going the other way round seems hard :P
Lol. Love your stats on that claim.
My dad is 50 and re-did his law degree in the UK and emigrated there.
How old do you think senior researchers are?
What is the average age of a publishing academic?
Please send us a link to your paper: "How The World Works"
Haha
Sci-Twi Believe what you want,
P.S It's never too late to learn stats.
Lol you are ignorant.
Wow, that is awesome! Congrats. NEY seems to have some age prejudices or has read some poorly conducted research :P
Yes. In fact, the older my professors, the better they are. There is a reason why older engineers are paid an order of magnitude more than graduates ...
I'll bet that Howard Johnson book costs a lot of clams.
Nice video. i love to lelarn, butt its hard. Jusst like you sead it in Australia, at dave`s plase : it takes to long time to whatch videos. butt im loving it, so thaks and god bye here from Norway. im a New Fan. Thanks
how /where do you connect the resistors?
Think of it as a variable resistor. You must first know your probes impedance. Use short lines and try to keep everything in balance between ground and the subject pin.
Uhmm.. subscribed.
Lmfao I have no idea what's going on in emag and this def makes me feel better about it
Faraday's Law includes mu-naught times initial amperage plus the remaining formula you showed.
I use to enjoy these videos... oh well
Fun Fact: Your equation remains wrong. Either it should be E on the right side leading to Ampere's law or rotE=-delB/delt (minus and without epsilon and µ) leading to Faraday's law ;)
(Remembers 3rd order differential equations) (shivers)
#Jack4President!
Good subjects, but poor explanations. First, a lot of the ringing shown in the first waveform was probably due to the inductance of the scope probes ground lead, not reflections in the traces. Second, the reflections are not cause be a mismatch between the "load" and "source" impedances; the are caused by the load and/or source impedances not matching the trace impedance. So, pretty sloppy work here.
+Steve Robbins Any tutorials/video's or chapters of a book to cover your better explanations? cause I am more of a firmware guy and sort of killed boards probably some way or another and this seem to indictate the problems of my former way of connecting to things. (poor tutorials like "take this TTL usb adapter and just connect it" .. wich may work, but with logic analysers and scopes....)
Yea, assuming its not because of a single pin probe.
Comb Filter resistors will make the signal smaller but the ratios will be preserved.
Comb Filter on resonant analogue circuits, agree. Often signals like this turn up as common mode noise.
too much talk....where are graphics.............
stop talking.....show us instead..............
"That 3ghz on your pentium" me: that 4ghz on my i7