I’m 62 years old and just beginning to learn about electronics. This channel stands out because it provides not just detailed information, but the right level of detail to truly understand what’s going on and why things are done the way they are. It makes all the difference-thank you!
Please make a talk with Eric regarding chassis and connector connections. I see sometimes capacitors, sometimes resistors on chassis-ground. Sometimes you can see explanation regarding whether capacitor or resistor use, but there is no talk about values or other specifications.
It's kind of amazing how bad the Arduino and RPi ecosystems are in terms of signal integrity, given that they're such popular high-volume products, and how frequently it causes hard-to-debug problems for people who are new to electronics. Low layer counts, bad stackups, lots of long flying leads with no ground reference, SI issues bad enough to cause failures (e.g. the WiFi not working on some models of RPi if you set a particular display output resolution). I was joking yesterday that I really want someone to invent coaxial dupont wires to solve this exact problem with connectors in hobbyist breakouts.
fantastic to mention also that Arduino Uno was so popular that they never changed at the beginning of the era the board to align the two connectors in one side to match the 0.1" pitch header, causing all people to buy a shield to fit the pinout correctly.....I believe it is doen on purpose... my guess.
@@CallistoPili No, I don't believe that is on purpose. People make mistakes and I believe this mistake was made by focusing on something the designer was unsure about and after ticking all the boxes on the test sheet, no fault came up, because being compatible with uniform pitch boards, such as Veroboard/ sttripboard, simply was not an explicit requirement put on the test list for the release.
I really appreciate your questions / interview style and Professor Bogatin's explanations / lectures. I learn a lot from watching your videos. Thank you, to both of you!!!
Summary: 21:30 to 21:48 Noise on a static output when the other outputs change state together with 40mA load (50R and LED in series), using a single ground wire: 20:00 Excellent. Nicely done. Very useful. Thank you!
If you know people who are happy to share their knowledge in 18 seconds clips, please let me know. I am very interested in making videos with them. Thanks for your help.
@@RobertFeranec oh come on, the Prof does a great job with his books and these videos, but this video is really too long without any real reason to be so. It is not the length but it's the content/time ratio.
@@robegatt I get your point, and I wondered the same, but if someone doesn't have this information, they will likely lose a LOT more than 48 minutes! I respect the teacher here. I've sent this video to some university students to save them a lot of time. The teacher here doesn't miss out any understanding steps, and that's really important for a newcomer to the field, or anyone with a patchy or incorrect understanding. A problem with an 18 second version is that we can't tell whether it's correct. My reason for posting my original post was to give people a summary that they can substantiate by watching the video. All good!
@@JSB2500 I wrote my comment because this is a trend I've seen a lot recently... it's like they are "beating around the bush" to fill a certain time range rather than really just explain the steps to give context. It's probably a money thing.
Fun little experiment showcaseing exactly what I was looking into after our controller was resetting. Thanks for the great video! My case: Behold a 12V line pulling up to 10A from the supply and forwarding it to the second board. The controller resides on the second board and is supplied via 5V from a buck on the first board - all GND the "same" In between, a cable (and a big capacitance to PE, which is connected to the first boards GND and FE -> an EMC nightmare)... And a big impedance miss match, as not only is the GND pincount insufficient, it's also routed horribly on the outside of the board-connector, thus having limited access to the GND polygon. And yeah, switching that load on the 12V line just wreaks the voltage generally and the 5V of the micro controller shoots to 8V and down to 1.2V, ultimately causing a BOR. In fact, I tip my hat to atmel, that the chip even survives that. Other issues also came out along the way and what can I say. I thoroughly enjoy studying this atrocious example our "senior" made, as I am learning a lot along the way. For that specific case the things I plan to tackle: - Reroute the damn thing - directly supply the second board and generate new 5V - lower the load switching speed (dv/dt) - just a power output, but many. - Probably also have another look at the signal connector to avoid ground loops, but most are differential anyways. - Split GND off of FE and PE... Well, but first management needs to agree -.- Maybe I missed something to look out for?
Yes, I also learned this the hard way. Had a display hooked up over long parallel cable with single ground pin. I had lots of clocks false triggering causing a garbled display. Luckily for me, I got away with fixing it by just using two more pins, because the connector only had so many, haha!
Also note that steep edges increase Ldi/dt, so if you can't have sufficient current returns, do intentional sloping of your edges (and current limitation) with RC elements on outputs, and regeneration on receiving end with Schmitt trigger inputs.
Very good video on signal integrity, as usual on this channel. Eric knows the matter an also how to explain it in a very visual and clear way. Thanks for sharing!
I recognize the "interleaved ground" scheme on all flat cables in the PC (floppy, PATA etc...). Honestly I am surprised at the "sub optimal design" of such a well known commercial product as Arduino. This is really instructive.
One minor comment: At that resolution, the red trace is actually pretty hard to see in the video. For future demonstrations, it would be better if you could use a different color (such as cyan) for secondary traces, etc. Also, I'm not sure if you didn't notice or just didn't mention: With the better ground, the rising edge of the switching lines also has much less rounding at the top. This shows that you were actually (technically) wrong about your explanation of why that happens. It is not actually happening because Vcc is changing. What you are actually seeing is that the sharp change in current running through that single ground return line is causing *ground to rise* on the other board (this is also why the quiet line goes "negative", because it is actually technically still at zero (relative to the Arduino) but on the remote board *the ground that it is referenced to* has (positive) switching noise on it that causes the Arduino and the other board to have different ground levels, and thus different "ideas" of what "zero voltage" actually is). It's very common to just think "ground is always zero", which is usually fine for DC, but when dealing with signals that change over time, ground is arguably a signal just like any other, and it can change over time too. This can make it _look_ like all the other signals are changing, when technically they aren't the ones that are being affected, it's just your reference point that you're comparing everything to that is changing due to the noise instead.
Sorry to say, but I have one question to that Arduino code. It has virtually nothing to do with signal integrity, but I am just curious how this happened. If you look at 8:25, there is on the line 7 command to turn all that outputs to Hi state, then it calls delayMicroseconds() function at line 8 with parameter 4, which according to Arduino's manual should wait for 4μs. Then it turns all that outputs to Lo state (line 9) and waits another 4μs, right? So the period of the output signal should be at least 8μs, in fact it would be a little bit longer, because all that operations and loops takes its time. But. If I look at 13:13, then what I see is a signal with a 6μs periodicity (which would be perfectly OK for a 2μs delays at lines 8 and 10). I know that that IDE sometimes does weird things. But when it comes to that "weird Arduino issues". I called that digitalWrite() stupid, because if you want to manipulate more than one bit, it takes so long. But now, after this, I realized, that that digitalWrite() thing might be a software solution to a hardware problem, because it does not let you to change state of more than one pin at a time. I would like to see what would happen if that code utilizes digitalWrite() instead of PORTB=. It would require much longer time to execute, but state of all that output pins should not change at the same moment, which should have some impact to crosstalk, even in that first poor ground scenario. Maybe I will test this somewhere in the future.
exactly, if not all the pins are controlled at the same time, that would improve the signals. Still, this may not be the perfect solution as the biggest problem on many not optimally designed systems is a random bad combination of events which happens only once a while and then the system may experience rare or occasional problems and you may never find out why because it is extremely difficult to replicate the problem. So ideally the problem should be prevented in hardware.
I think it's actually the other way around. The "hardware problem" they were working around is actually that you *can't* write to only one pin at a time, even though that's what most people want to do (it is very rare that anyone ever wants to change 8 IO lines all at the same time). Writing to the port directly _requires_ you to set all 8 of the lines at the same time, which means if there are some you don't want to change, you need to figure out what their values are (by reading the port first, etc), then change only the one or two bits you want to change, then write all 8 bits back to the port (which is basically what digitalWrite() does internally). digitalWrite() is just a convenience function to make life easier for the programmer in what is by far the most common situation. I really doubt anybody was thinking of these sorts of signal issue when they designed any of this, to be honest.
The thing about these delay instructions is they're not all that precise. The smallest unit of time this μc can use is about the length of one clock cycle, and increments are in multiples of that. Luckily for us, an 8MHz clock makes microseconds easy - 8 clock cycles = 1μs. But that doesn't guarantee precision. Not when you're working in Arduino C. Depending on how tightly that delay function is written, how much checking & housekeeping it does, how well it's calibrated to the particular μc, etc. it can be not only quite far out, but also somewhat inconsistent. The only real way to guarantee well-timed deterministic behaviour is to write all the time-critical bits in raw assembly (which the IDE handles fine). That ensures it's not doing anything you don't know about, and you can reliably tune it to deliver what you need (i.e. a delay of exactly 4μs).
@@foogod4237 It kind of makes sense that the I/O is banked, these are automotive microcontrollers, so their design brief revolves around speed & flexibility. Plus, I bet they implemented those ports as shift registers.
The comment I was going to reply to seems to have vanished, youtube at work. What I was going to say was I can see the point people make about the pacing of the video. This isn't intended as a criticism of you, or this video or anything like that, feedback at best, critique at worst. The content is great and I love how you always seem to ask the question I was about to ask. If this video were presented in a "tight" format 5 minutes or so would convey the information and examples briefly. 10 minutes would do so comfortably without significant loss of "educational content" though it may lose some of the anecdotes which are still good. I do understand that you are presenting this as a "conversation" more than a lesson/lecture etc but I must admit to not watching videos of yours that I was interested in due to the time commitment. The information is gold. But it feels more like something to put on while doing the laundry than something to sit and watch which isn't really doing the content justice. Especially as when there *is* something important happening you're nowhere near the screen to see the example. You mentioned you wouldn't mind working with people to produce shorter form content. It might be worth working with professional editors or agencies to try making some literal shorts (up to 3 minutes long now) as they are quite good for discovery and growing subscribers to your main content. The quality of the knowledge you present is top tier and sadly not reflected in your channels popularity or subscriber count. I don't think you'd need to add much new content to create those, mostly just an intro then the "highlights reel" of the video itself. Watching the noise jump up as each ground was pulled was the money shot. Then of course link to the longer form video. I know how long editing especially longer videos takes so outsourcing that may be worthwhile for you and isn't uncommon on channels of your size. Again, I am hoping this comes across as constructive. I know my tone of voice often comes across as hostile and that is the furthest thing from my intention. I really value what you do and give us here for free and in no way should you feel obliged to do anything at all to placate some random on the internet. Do what makes you happy.
I completely understand. I have done some short forms of long videos (tagged as #HighlightsRF). And I myself usually don't watch videos longer than 30minutes (and even then I use double speed). So, I also wish this kind of videos would exist. The biggest problem with making long videos short is that it takes a lot of time to make them. Many people assume, it is just question of cutting it down - but cutting down video breaks the flow, people start missing information and get confused what we are doing or talking about (unless they try editing videos by themselves they may not know what I am talking about). This means you need to record additional materials and somehow stitch everything together ... a lot of extra work. And instead of "only" 2 days working on the video it would take more. Also, I need to consider another factor - making these videos is about something for something, in this case, we can talk to clever people and they get exposure. If I cut down videos only to the main point, why my guests would be interested in helping me in making these videos? I can't present their ideas by myself, it needs to be them to explain it and also they are doing it to let people to know them better, to share with people their ideas / knowledge and thoughts, to build their image / brand / reputation and mention their company or other work. So I need to give them some time in the videos. Simply to say, if the videos are too short, they are good for viewers, but we would not have guests to talk to. Solution looks like making the long video first and then make short highlights and point back to the original video. I have tried that and I am still planning to do that when it is easy, I just don't have much time. Creating UA-cam videos is not my main job, I still have a lot of other work to do. I really understand what many people are saying, just many people only see this from one side - what is good for viewer. But there are more views to consider .... for example the guest point of view or my point of view. It would be nice to make super popular videos, but it is not my passion. It's not something I would keep thinking about all the time. I find video editing boring, I am not a script writer nor entertainer. I am an engineer. I enjoy working on electronic projects, creating tutorials and learning from clever people. And on my way of doing that I just share it.
Yep, starting with ATA-66, you had those 80-pins cables where every other pin was just a separator for the next pin. Not sure if those were actually connected to GND as the connectors on both ends were still 40 pins.
34:43 you can see that the live measurement time scale is on 50ns/div and the old capture is on 100ns/div. By mistake he thought his live capture is also on 100ns/div. Considering that, the results he got are even much better than what is shown in the scope here. For a second you might have thought its not a big change. But thats just different scaling.
Can you do a video with Mr. Bogatin about simulating reflection using ribbon cables and how to properly terminate long cables? I am using a GND wire for every signal return, but am still having EMC problems. My clock frequency is 3 MHz, but the cables are 2 meters long.
In this video Eric has created an antenna using one-side grounded wire (no big difference that he did it through a MOSFET inside of the MCU) and all the video he tried to shield it from RF signals using more independent ground wires... Strange experiments and poor explanations of physics, but conclusions are pretty good - use as much ground as possible. ) Greetings from Russia to all peaceful engineers. Lets develop useful things, not weapons! :)
I have a question about the layout, actually. If yoy have two rows of pins on your connector, is it better to put all the ground pins on one row and all the signal pins on the other row, or would it be better to alternate them so that the signal pins are diagonal to each other? Then you'd have diagonal distance between the signal pins rather than horizontal and grounds in horizontal spacing to the signals. Kind of like a shield? Maybe it doesn't matter...
That was fantastically instructive, thanks a lot! With a practical exemple in mind, what happens when the same connector is to power the daughter board? Will substituting a few GND wires for Vcc achieve the same (curative) effect? Intuitively I'd expect, Vcc an GND are both GND in AC model but I'd like to make sure there's no difference.
How much current is being switched here? A typical communication connection (spi, usb, etc) would pull much less current, correct? If so, would drawing much less current impact the noise with less grounds? Basically im wondering if the noise wouldn’t necessarily be as bad as we see here if the wires were carrying typical current for communications protocols like spi or usb or i2c Great video, thank you!
Rule #9: Never go anywhere without a knife. Yes, signal integrity is totally important. I did never care until I had problems with it. But it is easy to make it right on the PCB and with the wiring.
What hardware is he using to capture the data to the PC directly? is it a "normal" scope that can also be controlled from the computer, or a scope that only has software control/display?
depends on for what: cables, board to board, FPC, ...and also depends on interfaces (gpio, display, ...). Ribbon cables are still fine for many applications. The important is to keep in mind to have enough grounds.
This video missed the mark. Most of the problem here is likely due to powering the LEDs over the same lines used for the high speed signaling. If the LEDs were powered by a separate power/ground with transistors used to turn the LEDs on/off based on the signals coming over the ribbon cable, the problems would be minimized. It's bad when signal pins don't have their own return lines. It's worse when higher current devices (LEDs in this case) don't have their own power/return lines, and it would be a catastrophe if you were using the signal lines to power inductive loads (e.g. motors or solenoids).
If you can't afford using high pin count connector with all the return gnd. I heard we need to increase the rise time of the signal, but how to do it without affecting the functionality? Is it just add a resistor in series? Thanks you.
When do you need a pull-down resistor on that extra line to keep it low? I mean it's floating already so why would you expect anything like anything in the world could make it go and pick up noise that's why you use pull down resistors right?
One request, since UA-cam is inserting so many ads inbetween... (every 5 - 10 minutes, some even mid sentence) Can you for your next videos increase the audio volume? The volume of your video is quite low, which makes the ads cause a heart attack as those are so loud compared to the audio level of your video.
I also noticed the volume going down even I use the same settings as I used to. I am not sure if it has something to do with the Davinci upgrade. I will need to have a look into this, I though it was only my feeling but looks like it's not only me who noticed. PS: did you try to use the "Stable volume" in the player settings? Not sure if that can help.
@@RobertFeranecfor spoken audio in a teaching environment like this take a look at "compressing" the audio, it'll bring the quiet bits up to the same volume as the loud bits so you can ride closer to maximum volume without clipping. Also UA-cam I believe do specify a desired audio level somewhere but it's in some weird units.
Great video with great content! My only (hopefully constructive) criticism is that it is a bit long, or rather not the length, but that it is very repetitive. It could have shortened to half, I guess, by saying the same things only once throughout the lecture, keeping a continuous advancement of the topic. It would still cover the same information, but would be shorter and more engaging. Perhaps summarizing only at the end one more time. There is an another inconvenience, for me that is, as I have a slight protan color deficiency. Red on black IS extremely DARK. I barely see the red trace on the scope, especially that I watched the video on a small screen (mobile). A cyan or blue trace would be better. For us, at least, then it will be worse for those with blue deficiency, I guess, so you can't please everyone, can you? 😄
This video needs a lot more editing. I like the idea of your videos but the execution could be better. Editing and planning it out a little more would make it much more watchable. I hope this is helpful.
Mr Robert, I am thinking to 3D print an Atari-520 ST case, witch is a computer and keyboard in one together. So i try to CAD design the case for modern motherboard, and the mechanical button switches. Could you advice please what I should do, to make it properly having no place for any fan ..
you should be able to find a non fan board to emulate atari on and then just build it around it (e.g. github.com/Jean-MarcHarvengt/MCUME/blob/master/README.md ). For the case, find a 3D model and you can use it as a starting point ( e.g. www.thingiverse.com/thing:2512286 )
We don't need to be rocket engineer scientist to use Arduino, RPi, etc simple boards. Our projects are simple and practical, for hobby and entertainment, so doesn't matter what's happening in 2 micro seconds 😂😅 ..
I’m 62 years old and just beginning to learn about electronics. This channel stands out because it provides not just detailed information, but the right level of detail to truly understand what’s going on and why things are done the way they are. It makes all the difference-thank you!
thank you very much for nice words
Please make a talk with Eric regarding chassis and connector connections. I see sometimes capacitors, sometimes resistors on chassis-ground. Sometimes you can see explanation regarding whether capacitor or resistor use, but there is no talk about values or other specifications.
It's kind of amazing how bad the Arduino and RPi ecosystems are in terms of signal integrity, given that they're such popular high-volume products, and how frequently it causes hard-to-debug problems for people who are new to electronics. Low layer counts, bad stackups, lots of long flying leads with no ground reference, SI issues bad enough to cause failures (e.g. the WiFi not working on some models of RPi if you set a particular display output resolution). I was joking yesterday that I really want someone to invent coaxial dupont wires to solve this exact problem with connectors in hobbyist breakouts.
look at the RP2350 documentation for power supply circuit design/layout as another example of what goes wrong hwre
fantastic to mention also that Arduino Uno was so popular that they never changed at the beginning of the era the board to align the two connectors in one side to match the 0.1" pitch header, causing all people to buy a shield to fit the pinout correctly.....I believe it is doen on purpose... my guess.
DSLogic fly wires are probably the closest I've seen to that. Thin coax, splitting to a signal and gnd dupont on the end.
@@CallistoPili No, I don't believe that is on purpose. People make mistakes and I believe this mistake was made by focusing on something the designer was unsure about and after ticking all the boxes on the test sheet, no fault came up, because being compatible with uniform pitch boards, such as Veroboard/ sttripboard, simply was not an explicit requirement put on the test list for the release.
Arduino boards and high level code is so bad, I always recommend people going away from it as soon as possible, after they learn the basics.
I really appreciate your questions / interview style and Professor Bogatin's explanations / lectures. I learn a lot from watching your videos. Thank you, to both of you!!!
Summary: 21:30 to 21:48
Noise on a static output when the other outputs change state together with 40mA load (50R and LED in series), using a single ground wire: 20:00
Excellent. Nicely done. Very useful.
Thank you!
Lol. 18 seconds of info smeared on 48 minutes... I think this a guinness world record.
If you know people who are happy to share their knowledge in 18 seconds clips, please let me know. I am very interested in making videos with them. Thanks for your help.
@@RobertFeranec oh come on, the Prof does a great job with his books and these videos, but this video is really too long without any real reason to be so. It is not the length but it's the content/time ratio.
@@robegatt I get your point, and I wondered the same, but if someone doesn't have this information, they will likely lose a LOT more than 48 minutes!
I respect the teacher here. I've sent this video to some university students to save them a lot of time. The teacher here doesn't miss out any understanding steps, and that's really important for a newcomer to the field, or anyone with a patchy or incorrect understanding.
A problem with an 18 second version is that we can't tell whether it's correct.
My reason for posting my original post was to give people a summary that they can substantiate by watching the video.
All good!
@@JSB2500 I wrote my comment because this is a trend I've seen a lot recently... it's like they are "beating around the bush" to fill a certain time range rather than really just explain the steps to give context. It's probably a money thing.
What an amazing video with two amazing people. Love the videos, keep them coming.
thank you very much
Great video and great presentation from Dr. Bogatin!
Robert, thank you. I ran into some crosstalk issues in a board recently. This really helped.
Fun little experiment showcaseing exactly what I was looking into after our controller was resetting. Thanks for the great video!
My case:
Behold a 12V line pulling up to 10A from the supply and forwarding it to the second board. The controller resides on the second board and is supplied via 5V from a buck on the first board - all GND the "same"
In between, a cable (and a big capacitance to PE, which is connected to the first boards GND and FE -> an EMC nightmare)... And a big impedance miss match, as not only is the GND pincount insufficient, it's also routed horribly on the outside of the board-connector, thus having limited access to the GND polygon.
And yeah, switching that load on the 12V line just wreaks the voltage generally and the 5V of the micro controller shoots to 8V and down to 1.2V, ultimately causing a BOR. In fact, I tip my hat to atmel, that the chip even survives that.
Other issues also came out along the way and what can I say. I thoroughly enjoy studying this atrocious example our "senior" made, as I am learning a lot along the way.
For that specific case the things I plan to tackle:
- Reroute the damn thing
- directly supply the second board and generate new 5V
- lower the load switching speed (dv/dt) - just a power output, but many.
- Probably also have another look at the signal connector to avoid ground loops, but most are differential anyways.
- Split GND off of FE and PE...
Well, but first management needs to agree -.-
Maybe I missed something to look out for?
Yes, I also learned this the hard way. Had a display hooked up over long parallel cable with single ground pin. I had lots of clocks false triggering causing a garbled display. Luckily for me, I got away with fixing it by just using two more pins, because the connector only had so many, haha!
Also note that steep edges increase Ldi/dt, so if you can't have sufficient current returns, do intentional sloping of your edges (and current limitation) with RC elements on outputs, and regeneration on receiving end with Schmitt trigger inputs.
Thank you so much for this video. I have seen this problem on multiple products that use internal ribbon cables.
Very good video on signal integrity, as usual on this channel. Eric knows the matter an also how to explain it in a very visual and clear way. Thanks for sharing!
Very useful information. Thank you so much for all these videos and the effort you make to share knowledge !
I recognize the "interleaved ground" scheme on all flat cables in the PC (floppy, PATA etc...).
Honestly I am surprised at the "sub optimal design" of such a well known commercial product as Arduino.
This is really instructive.
One minor comment: At that resolution, the red trace is actually pretty hard to see in the video. For future demonstrations, it would be better if you could use a different color (such as cyan) for secondary traces, etc.
Also, I'm not sure if you didn't notice or just didn't mention: With the better ground, the rising edge of the switching lines also has much less rounding at the top. This shows that you were actually (technically) wrong about your explanation of why that happens. It is not actually happening because Vcc is changing. What you are actually seeing is that the sharp change in current running through that single ground return line is causing *ground to rise* on the other board (this is also why the quiet line goes "negative", because it is actually technically still at zero (relative to the Arduino) but on the remote board *the ground that it is referenced to* has (positive) switching noise on it that causes the Arduino and the other board to have different ground levels, and thus different "ideas" of what "zero voltage" actually is).
It's very common to just think "ground is always zero", which is usually fine for DC, but when dealing with signals that change over time, ground is arguably a signal just like any other, and it can change over time too. This can make it _look_ like all the other signals are changing, when technically they aren't the ones that are being affected, it's just your reference point that you're comparing everything to that is changing due to the noise instead.
Sorry to say, but I have one question to that Arduino code. It has virtually nothing to do with signal integrity, but I am just curious how this happened. If you look at 8:25, there is on the line 7 command to turn all that outputs to Hi state, then it calls delayMicroseconds() function at line 8 with parameter 4, which according to Arduino's manual should wait for 4μs. Then it turns all that outputs to Lo state (line 9) and waits another 4μs, right? So the period of the output signal should be at least 8μs, in fact it would be a little bit longer, because all that operations and loops takes its time. But. If I look at 13:13, then what I see is a signal with a 6μs periodicity (which would be perfectly OK for a 2μs delays at lines 8 and 10). I know that that IDE sometimes does weird things.
But when it comes to that "weird Arduino issues". I called that digitalWrite() stupid, because if you want to manipulate more than one bit, it takes so long. But now, after this, I realized, that that digitalWrite() thing might be a software solution to a hardware problem, because it does not let you to change state of more than one pin at a time. I would like to see what would happen if that code utilizes digitalWrite() instead of PORTB=. It would require much longer time to execute, but state of all that output pins should not change at the same moment, which should have some impact to crosstalk, even in that first poor ground scenario. Maybe I will test this somewhere in the future.
exactly, if not all the pins are controlled at the same time, that would improve the signals.
Still, this may not be the perfect solution as the biggest problem on many not optimally designed systems is a random bad combination of events which happens only once a while and then the system may experience rare or occasional problems and you may never find out why because it is extremely difficult to replicate the problem. So ideally the problem should be prevented in hardware.
I think it's actually the other way around. The "hardware problem" they were working around is actually that you *can't* write to only one pin at a time, even though that's what most people want to do (it is very rare that anyone ever wants to change 8 IO lines all at the same time). Writing to the port directly _requires_ you to set all 8 of the lines at the same time, which means if there are some you don't want to change, you need to figure out what their values are (by reading the port first, etc), then change only the one or two bits you want to change, then write all 8 bits back to the port (which is basically what digitalWrite() does internally). digitalWrite() is just a convenience function to make life easier for the programmer in what is by far the most common situation.
I really doubt anybody was thinking of these sorts of signal issue when they designed any of this, to be honest.
The thing about these delay instructions is they're not all that precise. The smallest unit of time this μc can use is about the length of one clock cycle, and increments are in multiples of that. Luckily for us, an 8MHz clock makes microseconds easy - 8 clock cycles = 1μs. But that doesn't guarantee precision. Not when you're working in Arduino C. Depending on how tightly that delay function is written, how much checking & housekeeping it does, how well it's calibrated to the particular μc, etc. it can be not only quite far out, but also somewhat inconsistent.
The only real way to guarantee well-timed deterministic behaviour is to write all the time-critical bits in raw assembly (which the IDE handles fine). That ensures it's not doing anything you don't know about, and you can reliably tune it to deliver what you need (i.e. a delay of exactly 4μs).
@@foogod4237 It kind of makes sense that the I/O is banked, these are automotive microcontrollers, so their design brief revolves around speed & flexibility. Plus, I bet they implemented those ports as shift registers.
God damn it Eric is a great guy with great knowledge. I love every video he is in.
The comment I was going to reply to seems to have vanished, youtube at work. What I was going to say was I can see the point people make about the pacing of the video. This isn't intended as a criticism of you, or this video or anything like that, feedback at best, critique at worst. The content is great and I love how you always seem to ask the question I was about to ask.
If this video were presented in a "tight" format 5 minutes or so would convey the information and examples briefly. 10 minutes would do so comfortably without significant loss of "educational content" though it may lose some of the anecdotes which are still good. I do understand that you are presenting this as a "conversation" more than a lesson/lecture etc but I must admit to not watching videos of yours that I was interested in due to the time commitment. The information is gold. But it feels more like something to put on while doing the laundry than something to sit and watch which isn't really doing the content justice. Especially as when there *is* something important happening you're nowhere near the screen to see the example.
You mentioned you wouldn't mind working with people to produce shorter form content. It might be worth working with professional editors or agencies to try making some literal shorts (up to 3 minutes long now) as they are quite good for discovery and growing subscribers to your main content. The quality of the knowledge you present is top tier and sadly not reflected in your channels popularity or subscriber count. I don't think you'd need to add much new content to create those, mostly just an intro then the "highlights reel" of the video itself. Watching the noise jump up as each ground was pulled was the money shot. Then of course link to the longer form video. I know how long editing especially longer videos takes so outsourcing that may be worthwhile for you and isn't uncommon on channels of your size.
Again, I am hoping this comes across as constructive. I know my tone of voice often comes across as hostile and that is the furthest thing from my intention. I really value what you do and give us here for free and in no way should you feel obliged to do anything at all to placate some random on the internet. Do what makes you happy.
I completely understand. I have done some short forms of long videos (tagged as #HighlightsRF). And I myself usually don't watch videos longer than 30minutes (and even then I use double speed). So, I also wish this kind of videos would exist.
The biggest problem with making long videos short is that it takes a lot of time to make them. Many people assume, it is just question of cutting it down - but cutting down video breaks the flow, people start missing information and get confused what we are doing or talking about (unless they try editing videos by themselves they may not know what I am talking about). This means you need to record additional materials and somehow stitch everything together ... a lot of extra work. And instead of "only" 2 days working on the video it would take more.
Also, I need to consider another factor - making these videos is about something for something, in this case, we can talk to clever people and they get exposure. If I cut down videos only to the main point, why my guests would be interested in helping me in making these videos? I can't present their ideas by myself, it needs to be them to explain it and also they are doing it to let people to know them better, to share with people their ideas / knowledge and thoughts, to build their image / brand / reputation and mention their company or other work. So I need to give them some time in the videos.
Simply to say, if the videos are too short, they are good for viewers, but we would not have guests to talk to.
Solution looks like making the long video first and then make short highlights and point back to the original video. I have tried that and I am still planning to do that when it is easy, I just don't have much time. Creating UA-cam videos is not my main job, I still have a lot of other work to do.
I really understand what many people are saying, just many people only see this from one side - what is good for viewer. But there are more views to consider .... for example the guest point of view or my point of view.
It would be nice to make super popular videos, but it is not my passion. It's not something I would keep thinking about all the time. I find video editing boring, I am not a script writer nor entertainer. I am an engineer. I enjoy working on electronic projects, creating tutorials and learning from clever people. And on my way of doing that I just share it.
@@RobertFeranec I like long video - I can get a deep dive in topic
Great explanation and demonstration! What equipment and software do you use to show the scope and video camera?
19:00 maybe I missed it, but does this IO line set with an internal pulldown or pullup, and if not, would that solve the hardware issue in software?
It was very interesting, thank you so much for valuable video. I am from Myanmar.
Nice presentation 🤞🙏
Pi Pico has a nice pinout, one ground pin every 4 signal pins. I hope it becomes a trend!
Hi, improvement advice: make a summary at the end that presents the result, in a graphical way it is best) so we can fast compare the results.
Thank you
the old fashion IDE/PATA connectors specifically are designed to have the ground pins exactly as shown in the video by Bogatin!!!
Yep, starting with ATA-66, you had those 80-pins cables where every other pin was just a separator for the next pin.
Not sure if those were actually connected to GND as the connectors on both ends were still 40 pins.
34:43 you can see that the live measurement time scale is on 50ns/div and the old capture is on 100ns/div. By mistake he thought his live capture is also on 100ns/div.
Considering that, the results he got are even much better than what is shown in the scope here. For a second you might have thought its not a big change. But thats just different scaling.
Is there a place we can find the board that Mr Bogatin has? Maybe order it from JLPCB or PCBWay.
I suggest to use for each signal pin a dedicated GND pin, if possible.
Can you do a video with Mr. Bogatin about simulating reflection using ribbon cables and how to properly terminate long cables? I am using a GND wire for every signal return, but am still having EMC problems. My clock frequency is 3 MHz, but the cables are 2 meters long.
In this video Eric has created an antenna using one-side grounded wire (no big difference that he did it through a MOSFET inside of the MCU) and all the video he tried to shield it from RF signals using more independent ground wires... Strange experiments and poor explanations of physics, but conclusions are pretty good - use as much ground as possible. )
Greetings from Russia to all peaceful engineers. Lets develop useful things, not weapons! :)
I have a question about the layout, actually. If yoy have two rows of pins on your connector, is it better to put all the ground pins on one row and all the signal pins on the other row, or would it be better to alternate them so that the signal pins are diagonal to each other? Then you'd have diagonal distance between the signal pins rather than horizontal and grounds in horizontal spacing to the signals. Kind of like a shield?
Maybe it doesn't matter...
That was fantastically instructive, thanks a lot! With a practical exemple in mind, what happens when the same connector is to power the daughter board? Will substituting a few GND wires for Vcc achieve the same (curative) effect? Intuitively I'd expect, Vcc an GND are both GND in AC model but I'd like to make sure there's no difference.
How much current is being switched here? A typical communication connection (spi, usb, etc) would pull much less current, correct? If so, would drawing much less current impact the noise with less grounds? Basically im wondering if the noise wouldn’t necessarily be as bad as we see here if the wires were carrying typical current for communications protocols like spi or usb or i2c
Great video, thank you!
Is that ATMEGA capable of driving 200mA simultaneously in the first place? It does not look like you are solving much even with 7 GND wires.
oh no wonder the pi pico has some grounds on its board
Rule #9: Never go anywhere without a knife.
Yes, signal integrity is totally important. I did never care until I had problems with it. But it is easy to make it right on the PCB and with the wiring.
does this also apply to ic packages where QFN package is much better than QFP in signal integrity?
What hardware is he using to capture the data to the PC directly? is it a "normal" scope that can also be controlled from the computer, or a scope that only has software control/display?
Which connectors you recommend using?
depends on for what: cables, board to board, FPC, ...and also depends on interfaces (gpio, display, ...). Ribbon cables are still fine for many applications. The important is to keep in mind to have enough grounds.
4:47 - 4:49 what is this video? I want to watch about bad ground fill on top and bottom of arduino
We have not make a direct video about Arduino ground fill, but this video can help: ua-cam.com/video/52fxuRGifLU/v-deo.html
This video missed the mark.
Most of the problem here is likely due to powering the LEDs over the same lines used for the high speed signaling. If the LEDs were powered by a separate power/ground with transistors used to turn the LEDs on/off based on the signals coming over the ribbon cable, the problems would be minimized. It's bad when signal pins don't have their own return lines. It's worse when higher current devices (LEDs in this case) don't have their own power/return lines, and it would be a catastrophe if you were using the signal lines to power inductive loads (e.g. motors or solenoids).
If you can't afford using high pin count connector with all the return gnd. I heard we need to increase the rise time of the signal, but how to do it without affecting the functionality? Is it just add a resistor in series?
Thanks you.
Yes, resistors in series for driving logic.
or sometimes the MCU/CPU has register settings to slow down the edges, that can also help
When do you need a pull-down resistor on that extra line to keep it low? I mean it's floating already so why would you expect anything like anything in the world could make it go and pick up noise that's why you use pull down resistors right?
the pin is not floating. it is set as an output with value 0
One request, since UA-cam is inserting so many ads inbetween... (every 5 - 10 minutes, some even mid sentence)
Can you for your next videos increase the audio volume? The volume of your video is quite low, which makes the ads cause a heart attack as those are so loud compared to the audio level of your video.
I also noticed the volume going down even I use the same settings as I used to. I am not sure if it has something to do with the Davinci upgrade. I will need to have a look into this, I though it was only my feeling but looks like it's not only me who noticed. PS: did you try to use the "Stable volume" in the player settings? Not sure if that can help.
@@RobertFeranecfor spoken audio in a teaching environment like this take a look at "compressing" the audio, it'll bring the quiet bits up to the same volume as the loud bits so you can ride closer to maximum volume without clipping.
Also UA-cam I believe do specify a desired audio level somewhere but it's in some weird units.
Hurrah!!!
Thanks'! :)
Great video with great content!
My only (hopefully constructive) criticism is that it is a bit long, or rather not the length, but that it is very repetitive. It could have shortened to half, I guess, by saying the same things only once throughout the lecture, keeping a continuous advancement of the topic. It would still cover the same information, but would be shorter and more engaging. Perhaps summarizing only at the end one more time.
There is an another inconvenience, for me that is, as I have a slight protan color deficiency. Red on black IS extremely DARK. I barely see the red trace on the scope, especially that I watched the video on a small screen (mobile). A cyan or blue trace would be better. For us, at least, then it will be worse for those with blue deficiency, I guess, so you can't please everyone, can you? 😄
This video needs a lot more editing. I like the idea of your videos but the execution could be better. Editing and planning it out a little more would make it much more watchable. I hope this is helpful.
Sometimes
Could have been said and shown in 5 minutes..
Mr Robert, I am thinking to 3D print an Atari-520 ST case, witch is a computer and keyboard in one together.
So i try to CAD design the case for modern motherboard, and the mechanical button switches.
Could you advice please what I should do, to make it properly having no place for any fan ..
you should be able to find a non fan board to emulate atari on and then just build it around it (e.g. github.com/Jean-MarcHarvengt/MCUME/blob/master/README.md ). For the case, find a 3D model and you can use it as a starting point ( e.g. www.thingiverse.com/thing:2512286 )
Mais qu’est ce qu’il raconte ?
dude, I am 17 minutes in and have observed about 60seconds worth of information..
Yeah, it's a lot to take in, just watch it again a couple of times, and you will understand the rest.
Yes, could have edited out the changing of wires and such and made a much more compact video with zero information loss.
We don't need to be rocket engineer scientist to use Arduino, RPi, etc simple boards. Our projects are simple and practical, for hobby and entertainment, so doesn't matter what's happening in 2 micro seconds 😂😅 ..
They are talking about the end result on commercial circuit boards.
SPI? i2c? i2s? No? Nothing?
problem is, that sometimes projects built based on these boards will behave randomly or weird