I think the formula "Z = R + jX" in the beginning is actually giving the wrong impression. The characteristic impedance of a trace is actually a real valued number and is based on the ratio between the current and voltage waves travelling down the line. The impedance seen by a VNA at the input has nothing to do with the couple of milli Ohms that you measured using the multimeter. PS Not that it matters to most, but that width over thickness trick of 2 for getting 50 Ohms impedance only works for materials with a relative dielectric constant of about 4, aka FR4 which is of course the most popular substrate material in PCBs, but not the only one.
"we have a little liquid nitrogen left so we're gonna try to make ice cream" This is giving me janky youtuber engineering (william osman and the like) vibes lmao
Electric engineers need to be taught that sending signals down a line is not too far from a drummer playing down an echoey hallway. You need just the right type of hallway for the right beat or it just turns to garbage.
As someone who still has a few liters of ferric chloride in storage, the ability to design and have multli-layer PCBs printed and stuffed with SMT components blows my mind! Keep up the great videos!
Yep, impedance. Also known as "RF black magic is afoot". That's why real high speed data communication needs awesome keysight oscilloscopes to discover these problems.
I think a crucial parts is missing at the start: Is that applicable to lower frequencies to near DC? Say a trace carrying Audio-signals, or Servo/fan-controls that are in the lower tens of kilohertz.
basic rule of thumb is you don't need to take transmission line effects into account until the line is longer than 10% of the wavelength of the highest frequency signal. For audio, wavelength at 22khz is 13.6km, so as long as your line is shorter than 1.36km, you don't really need to worry (at least about impedance matching - loss on a 1.36km cable might be something to worry about though)! You can do the math for other frequencies yourself (left as an exercise for the op)
@@gorak9000 I know that - that is why i remarked this is missing information in the video cause all the talk is about how 50 Ohm would result in the lowest reflections and even highest power transmission but at no point was it mentioned that this is for decently high frequencies only. I have seen people doing both - trying impedance-matching for current-driven circuitry and lowering DC resistance on a 5GHz differential transmission line.
@@ABaumstumpf I'd say the biggest problem with the video is it mentions traces should be 50 ohms, like that's applicable EVERYWHERE, when it is absolutely not. Traces should be 50 ohms if your source and load are matched to 50 ohms (aka usually in RF nets only - and even then, not every RF component has 50 ohm input or output impedance - that's why you check the datasheets and or put in matching networks). In digital circuits, TTL or CMOS inputs are going to be FAR from 50 ohms. I feel like the video gives the impression that every trace on every board should be 50 ohm, when that's far from correct.
Awesome video! I would love to see a deep dive comparing different style SMA connectors (vertical through hole, vertical surface mount, generic edge launch, and solderless compression) with all else being the same. The stub on through hole SMA connectors is something not well characterized in datasheets and seeing a comparison between all the different types would be really interesting.
I've seen some through-hole BNC connectors for 12G-SDI digital video that specify the required board thickness to obtain the specified performance from the connectors.
@@KeysightLabs while you're at it you could show the huge differences between cheap amazon / ebay / aliexpress sma connectors and good quality ones. I think there's more meaningful differences there than worrying about the little stub of the center conductor
The curve pattern shifts to either higher or lower frequencies as you cool or heat the board. Is that caused by the contraction and expansion of the board?
I would guess it is primarily caused by the change in the resistivity of the copper conductor. Copper has a positive thermal coefficient, so when cooled down, copper has a lower resistivity, and when heated up it has a higher resistivity.
We make a software for it that we used here called Electrical Performance Scan, which is part of the overall Pathwave family of simulation/emulation software. It's super powerful stuff: www.keysight.com/us/en/products/software/pathwave-design-software/pathwave-advanced-design-system.html
8:30 - i am pretty sure the trace is 13.4 "mil" and not Millimeter. And that would be its width, not thickness. a 13 millimeter thick trace would be a solid copper busbar and not just a trace :P
Hi, I would like to the exact PCB specs used in the demo and also I would like to be able reproduce the board traces, etc. Can you send me the relevant information for this such as the Gerber file, PCB manufacturer with data sheet, etc. etc.?
This was really fun, but to be honest it felt a bit all over the place. First you're testing different trace widths, then suddenly you're drilling out vias, then you're playing with liquid nitrogen? I really kinda feel like those could have (and maybe should have) been three different videos, because they were all kinda unrelated to each other, and none of them really felt like they got the full treatment they deserved because you were just rushing from one thing to another. In particular, I would really have liked to see more investigation of the relationship between traces, trace widths, and surrounding materials (for example, using a ground plane vs using parallel ground traces, width of ground traces vs signal traces, etc), which it really looked like you were actually going to do based on some of the other traces on that board, but then you never did, which was kinda disappointing...
Agreed! I wish I could do more on TH-cam. I am putting together an hour+ version that explores all three boards including the coplanar vs. microstrip traces. We'll email that super cit out to folks who sign up for Live from the Lab, and the event also goes into a lot more detail
@@KeysightLabsI think the introductory presentation touched upon some key engineering points that may relate to electronics in deep space versus inside a vehicle dashboard, in the hot Florida sun. It’s a tremendous amount of information to convey is such a short time. Good job!
@@KeysightLabs too bad the world mostly lives on youtube, not locked away on the keysight website somewhere (where even to download manuals and stuff it annoyingly asks for email addresses continually). The good HP legacy is slowly fading...
Well I'm glad products like the nanoVNA exist, so we don't have to dream of owning GHz equipment.... I guess that's not the market keysight aims for but i guess i can follow along with it.....
@@KeysightLabs Yep, 1 zeptoparsec = 1.21483 mils. That's one hell of a fluke if you didn't plan it! 🤯 Time to start labelling everything in zpc, I think! 😹 At least then everyone will be equally confused ;)
For historical reasons, PCB manufacturing uses imperial units. Trace widths and clearances are specified in mils (thousandths of an inch), copper thickness is specified in ounces per square foot, common breadboard/perfboard/chip pin spacing is 0.1", surface mount passive packages are named based on the number of 10 mills in each dimension, and so on. Just to liven things up a bit, many other SMD parts will use millimeter units for their layouts, pin spacing, and clearances. So when doing PCB design you get pretty good at switching seamlessly between units.
@@funtechu Interesting info, thanks; I've always used metric up here in the old Europe, but the base is painfully clear when everything is a multiple/fraction of 2.54. The only field where imperial is generally used is plumbing, I guess for historical reasons too. Funny how a mil is both imperial AND metric, and not some "twelve sixteenths of a quarter inch" kind of nonsense .-)
David, today we learnt about You .. can't properly 1) solder 2) drill 3) handle lq. nitrogen 4) use a torch. I think that's enough for today 😖😖. But it's been a lot of fun🤡🤡
Please. Where does the magical 50 Ohm value even come from? Say I have an 100MHz bus with a source impedance of 100 Ohm and load impedance of over 100 kOhm, so where does 50 Ohm come in? Nowhere, it's not a transmission line. These sorts of buses already tend to be somewhat fragile when you design them yourself, there are easy mistakes to be made, but this sort of advice doesn't apply to them at all either.
Generally this is related to high speed digital signals, and it's very common to have a 50 ohm TX and RX. There are situations where the impedance isn't a 50 ohm target, but usually it's in that 30-70 ohm range. In those situations this advice still applies, but needs to be adjusted to match the target impedance.
@@KeysightLabs Oh, well, it's hard to make educational videos anyway, this felt like just a video about an experiment, not necessarily about learning something new. Don't bother on making 10min videos, that thing is on the past, just make the video good and easy to understand to the average-viewer (or at least the enthusiast ones[normie < enthusiast < developer]), it's not necessary to explain everything you can just leave links to other videos that explain the thing, i normally watch 20-40min videos if the video and topic are entertaining enough. haha
I try to make sure we don't lose too many newer folks in these things. I personally often benefit from fundamentals review, too, it often helps me think of things in new/different ways
@@KeysightLabs IMHO you are targeting the wrong audience but if that is your goal: expand on the topic, devote more time to the actual explanations of the fundamentals then ease into the advanced bits.
Fair enough! Definitely check out our Design Software channel, sounds like it might be up your alley. Some great deeper-dive videos there: youtube.com/@KeysightEEsofEDA
Come join me live on 14-March (or on demand) here: keysig.ht/HsK7PF
Hope to see you there!
This seems a good reason for end users not to buy from Keysight :-
th-cam.com/video/Od7ejIJPQwg/w-d-xo.html
Where to watch the whole past event video with Tim? The link is going nowhere.
We need to see more of these wave simulation animations!!! GALORE!!!!
Wow that visual graphic of the wave travelling on the board was really neat. Great find on that, very cool.
It's very very cool, shoutout to @samerps on Twitter for helping out!
I love the 3D printed via and traces! If only manufacturers would let you go so big.
I BUILT THE WORLD'S LARGEST CIRCUIT BOARD
I think the formula "Z = R + jX" in the beginning is actually giving the wrong impression. The characteristic impedance of a trace is actually a real valued number and is based on the ratio between the current and voltage waves travelling down the line. The impedance seen by a VNA at the input has nothing to do with the couple of milli Ohms that you measured using the multimeter. PS Not that it matters to most, but that width over thickness trick of 2 for getting 50 Ohms impedance only works for materials with a relative dielectric constant of about 4, aka FR4 which is of course the most popular substrate material in PCBs, but not the only one.
Excellent points!
my guess was right, I have done aerial design way back in the 80's. Long forgotten but still remembered.
"we have a little liquid nitrogen left so we're gonna try to make ice cream"
This is giving me janky youtuber engineering (william osman and the like) vibes lmao
"Three twists of salt, approximately" is so very PhD humor! Accuracy vs precision. ❤❤❤
Electric engineers need to be taught that sending signals down a line is not too far from a drummer playing down an echoey hallway. You need just the right type of hallway for the right beat or it just turns to garbage.
As engineers we need to stop calling reactive components imaginary. Makes us sound like we're making everything up haha
Daniel Bogdanoff, "Clearly a man of science". Such a great video!
Haha, thank you!
As someone who still has a few liters of ferric chloride in storage, the ability to design and have multli-layer PCBs printed and stuffed with SMT components blows my mind! Keep up the great videos!
Best visual representation of wave travel impairments since a video from the 40s I saw on YT!
Came here from the stream, lucky it is uploaded here because I did not want to miss this!
Yep, impedance. Also known as "RF black magic is afoot".
That's why real high speed data communication needs awesome keysight oscilloscopes to discover these problems.
I think a crucial parts is missing at the start:
Is that applicable to lower frequencies to near DC? Say a trace carrying Audio-signals, or Servo/fan-controls that are in the lower tens of kilohertz.
basic rule of thumb is you don't need to take transmission line effects into account until the line is longer than 10% of the wavelength of the highest frequency signal. For audio, wavelength at 22khz is 13.6km, so as long as your line is shorter than 1.36km, you don't really need to worry (at least about impedance matching - loss on a 1.36km cable might be something to worry about though)! You can do the math for other frequencies yourself (left as an exercise for the op)
@@gorak9000 I know that - that is why i remarked this is missing information in the video cause all the talk is about how 50 Ohm would result in the lowest reflections and even highest power transmission but at no point was it mentioned that this is for decently high frequencies only.
I have seen people doing both - trying impedance-matching for current-driven circuitry and lowering DC resistance on a 5GHz differential transmission line.
@@ABaumstumpf I'd say the biggest problem with the video is it mentions traces should be 50 ohms, like that's applicable EVERYWHERE, when it is absolutely not. Traces should be 50 ohms if your source and load are matched to 50 ohms (aka usually in RF nets only - and even then, not every RF component has 50 ohm input or output impedance - that's why you check the datasheets and or put in matching networks). In digital circuits, TTL or CMOS inputs are going to be FAR from 50 ohms. I feel like the video gives the impression that every trace on every board should be 50 ohm, when that's far from correct.
This is awesome. I've just been learning about this stuff at uni so seeing some real world application is awesome
Awesome video! I would love to see a deep dive comparing different style SMA connectors (vertical through hole, vertical surface mount, generic edge launch, and solderless compression) with all else being the same. The stub on through hole SMA connectors is something not well characterized in datasheets and seeing a comparison between all the different types would be really interesting.
That would be super cool! Great idea
I've seen some through-hole BNC connectors for 12G-SDI digital video that specify the required board thickness to obtain the specified performance from the connectors.
@@KeysightLabs while you're at it you could show the huge differences between cheap amazon / ebay / aliexpress sma connectors and good quality ones. I think there's more meaningful differences there than worrying about the little stub of the center conductor
OMG you got the best job in the world!!!! Driving to the welding store and then to your garage. LOL.
The ice cream was just too much LOL.
Awesome video! I love watching you guys messing about, and that rule of thumb is nice!
Those glasses look similar to that of the "Stealy eyed missile man" from the Apollo missions of NASA. No doubt you are a man of science
Activate windows still makes me laugh all the time! Do an antenna next time!
The trace "Thiccness" matters a lot in high power applications.
Fantastic!!!!! Very instructive!!!! Keep going!
Can we access the streams recorded? The timing may not be suitable for some.
Yes! Will be available on-demand
What is the rule of thumb for an INTERNAL trace ... still 2X ??
The curve pattern shifts to either higher or lower frequencies as you cool or heat the board. Is that caused by the contraction and expansion of the board?
I would guess it is primarily caused by the change in the resistivity of the copper conductor. Copper has a positive thermal coefficient, so when cooled down, copper has a lower resistivity, and when heated up it has a higher resistivity.
What program you guys used for those animation/simulation? I need to learn this.
I didn't know trace simulation exists. What software do motherboard manufacturers use?
We make a software for it that we used here called Electrical Performance Scan, which is part of the overall Pathwave family of simulation/emulation software. It's super powerful stuff: www.keysight.com/us/en/products/software/pathwave-design-software/pathwave-advanced-design-system.html
@@KeysightLabs yeah, and you need a super powerful wallet to buy it!
Thjs vjdeo should be shown jn every electronjcs lecture
you are good and fun, thank you!
Thanks Daniel
8:30 - i am pretty sure the trace is 13.4 "mil" and not Millimeter. And that would be its width, not thickness. a 13 millimeter thick trace would be a solid copper busbar and not just a trace :P
Yeah, definitely misspoke there!
Hi, I would like to the exact PCB specs used in the demo and also I would like to be able reproduce the board traces, etc. Can you send me the relevant information for this such as the Gerber file, PCB manufacturer with data sheet, etc. etc.?
Yes, I'll put the board files up for folks
EDIT: Here are the board files www.dropbox.com/sh/36ln3j2scbpgi6o/AAC7mdYIlX9uEi8XvKn5YRFHa?dl=0
That 3d printed via 😲
Sure is!
This was really fun, but to be honest it felt a bit all over the place. First you're testing different trace widths, then suddenly you're drilling out vias, then you're playing with liquid nitrogen? I really kinda feel like those could have (and maybe should have) been three different videos, because they were all kinda unrelated to each other, and none of them really felt like they got the full treatment they deserved because you were just rushing from one thing to another.
In particular, I would really have liked to see more investigation of the relationship between traces, trace widths, and surrounding materials (for example, using a ground plane vs using parallel ground traces, width of ground traces vs signal traces, etc), which it really looked like you were actually going to do based on some of the other traces on that board, but then you never did, which was kinda disappointing...
Agreed! I wish I could do more on TH-cam. I am putting together an hour+ version that explores all three boards including the coplanar vs. microstrip traces. We'll email that super cit out to folks who sign up for Live from the Lab, and the event also goes into a lot more detail
@@KeysightLabsI think the introductory presentation touched upon some key engineering points that may relate to electronics in deep space versus inside a vehicle dashboard, in the hot Florida sun. It’s a tremendous amount of information to convey is such a short time. Good job!
@@KeysightLabs too bad the world mostly lives on youtube, not locked away on the keysight website somewhere (where even to download manuals and stuff it annoyingly asks for email addresses continually). The good HP legacy is slowly fading...
Bro is so chill loved it
my return value on the PCB is classic ...
Who is today's winner?
So let's see some white-water footage Danial! Maybe test the IP rating of a DMM on the kayak?
At some point I want to do a four-helmet day with the DMM - snow bike climb paddle
Well I'm glad products like the nanoVNA exist, so we don't have to dream of owning GHz equipment.... I guess that's not the market keysight aims for but i guess i can follow along with it.....
Isn't that Wong from Dr Strange?
why wasn't my college physics this entertaining? i might have not flunked it all those times...
Great video! But consider using regular units, not those funky mils, mits, miliburgers, etc.
Next time everything will be in zeptoparsecs
Edit: holy moly this is actual nearly 1:1 if my math is correct
@@KeysightLabs Yep, 1 zeptoparsec = 1.21483 mils. That's one hell of a fluke if you didn't plan it! 🤯
Time to start labelling everything in zpc, I think! 😹 At least then everyone will be equally confused ;)
I definitely did not plan it, bit I definitely will start using it!
that quarter wavelength discussion sounded interesting 🤔 but you cut it out 🤨
Yeah that's a whole deep dive... Vid is already 17 minutes!
Geek level ten thousand
your skill will raise you up Sir like some of my classmates or my classmates from other electronic engineering major
Weakest point of this video is at 7:39 - measuring a trace with a DMM and confusing DC resistance with transmission line impedance.
Valid, but I'm kinda trying to make the point overall that you can't just think DC
you have quite a bit of radiation loss in that trace which will not go away by cooling it down.
LOVE THIS!!!
Thank you!
"Mil"? Measuring science with body parts :D
For historical reasons, PCB manufacturing uses imperial units. Trace widths and clearances are specified in mils (thousandths of an inch), copper thickness is specified in ounces per square foot, common breadboard/perfboard/chip pin spacing is 0.1", surface mount passive packages are named based on the number of 10 mills in each dimension, and so on. Just to liven things up a bit, many other SMD parts will use millimeter units for their layouts, pin spacing, and clearances. So when doing PCB design you get pretty good at switching seamlessly between units.
@@funtechu Interesting info, thanks; I've always used metric up here in the old Europe, but the base is painfully clear when everything is a multiple/fraction of 2.54. The only field where imperial is generally used is plumbing, I guess for historical reasons too.
Funny how a mil is both imperial AND metric, and not some "twelve sixteenths of a quarter inch" kind of nonsense .-)
Love the content ....
Good board designers do this all day everyday
Yep! They call this type of board a "coupon board" and it's pretty common to get one & test parameters, albeit not as extremely varying as this one
Yes, of course we do....... yawn......
Coffee time!
@@Theineluctable_SOME_CANT I have been working on 112G serdes transition lately, less yawn then normal layout.
Super great video and really fun! Good job guys thanks for the video also cleared up some S11 versus S21 measurements!
David, today we learnt about You .. can't properly 1) solder 2) drill 3) handle lq. nitrogen 4) use a torch. I think that's enough for today 😖😖. But it's been a lot of fun🤡🤡
🙃
I'm here because the stream keeps laggin
Liquid nitrogen is so cool!!
I see what you did there
4:36 they sell the most expensive instruments in the world but don't have the money for a Windows license. Congratulations, good impression
Yeah, I didn't see that until later 🤦♂️. Was just an IT error we got fixed
Maybe they don't have the NEED for a Windows license.
@@IlBiggo if you’re using it, you need it.
@@IlBiggo I don’t think works the same with keysight licenses…
Impedance value? What happened there, lol.
haha
Please. Where does the magical 50 Ohm value even come from?
Say I have an 100MHz bus with a source impedance of 100 Ohm and load impedance of over 100 kOhm, so where does 50 Ohm come in? Nowhere, it's not a transmission line. These sorts of buses already tend to be somewhat fragile when you design them yourself, there are easy mistakes to be made, but this sort of advice doesn't apply to them at all either.
Generally this is related to high speed digital signals, and it's very common to have a 50 ohm TX and RX. There are situations where the impedance isn't a 50 ohm target, but usually it's in that 30-70 ohm range. In those situations this advice still applies, but needs to be adjusted to match the target impedance.
As always giveaways never reach me in serbia
Here from today's stream
Welcome!
You never explained the purpose or applications of what you were doing before starting to trow data at the viewer. 0/10
Sorry to disappoint! It's a delicate balance between too much and too little.
@@KeysightLabs
Oh, well, it's hard to make educational videos anyway, this felt like just a video about an experiment, not necessarily about learning something new.
Don't bother on making 10min videos, that thing is on the past, just make the video good and easy to understand to the average-viewer (or at least the enthusiast ones[normie < enthusiast < developer]), it's not necessary to explain everything you can just leave links to other videos that explain the thing, i normally watch 20-40min videos if the video and topic are entertaining enough. haha
Judging by the explanations you are trying to introduce quite an advanced topic to a beginner level audience. Kind of pointless.
I try to make sure we don't lose too many newer folks in these things. I personally often benefit from fundamentals review, too, it often helps me think of things in new/different ways
@@KeysightLabs IMHO you are targeting the wrong audience but if that is your goal: expand on the topic, devote more time to the actual explanations of the fundamentals then ease into the advanced bits.
Fair enough! Definitely check out our Design Software channel, sounds like it might be up your alley. Some great deeper-dive videos there: youtube.com/@KeysightEEsofEDA
@@KeysightLabs Interesting, I somehow missed that one, thanks.
First 😮