Man that scope is huge. Looks like it has a DVD drive...can it play movies? Lol. Seriously though, this was a great explanation, And I believe I understand these concepts much better now. Thanks so much for sharing your discoveries!
At first, I was wondering at the back of my mind how good resolution that scope would have to have to see such fine details in a 3 ns pulse, but it wasn't until I was halfway into the video that I started realizing that that would havta be a really expensive scope, and then I saw it had four channels as well. I then Googled the retail price to, indeed, be $21,900. Pretty nice hobby equipment you have sitting around on that bench!
Hi Robert I have a question. let's say I have both source imp and characteristic imp 50 ohm. should we have 50 ohm as well on the load impedance to prevent a reflection?, we know that mostly our load is high impedance, should I parallel terminate it with a resistor close to the load? load impedance e.g another CMOS buffer of MCU, or maybe a transistor.
Awesome video! It's nice to see this in application to PCB traces, especially since most sources (and university courses) cover it using coaxial cables. Would you be willing to do a comparison of other termination techniques such as an end termination and the voltage divider end termination? (I have seen all three used for different DRAM types / signals) It would also be interesting to see how the termination technique changes the behavior if the line is loaded (i.e. if you have multiple tri-state drivers with series termination for example - with only one of them driving at a time). As for the propagation speed, that needs to be calculated from the distributed capacitance and inductance of the transmission line, which will depend on many factors (trace width, PCB thickness assuming a ground plane, clearance to other traces, solder-mask thickness and material, near field metallic objects, etc. - A PCB trace is an antenna). Most undergraduate physics students do this in their electronics lab using a coaxial cable and a fast pulse generator (you can also measure the transmission line impedance by using an end termination resistor of known resistance). Note that the impedance and propagation speed will also change with frequency (and therefore rise time and the Fourier components of the square wave signal), which is most likely why the return pulse appears to be less sharp.
I will probably be doing this. One issue I realized just as I finished the video upload is that I didn't cover one important case to me, which is a driver with a series termination in the middle of the line. My original thought was that it would act the same, but then I realized that the two transmission lines would present an effective impedance of 50 ohms (2 lines of 100 ohms in parallel), so the series termination would have to be 50 ohms. I want to test that, so round 2 of order-PCB-from-China is on!
That's what I meant by "loading the line", though your description is more accurate. I think you are correct in terms of the impedance changing, which would mean that the impedance would change with each new device on the line. In addition, since they are at different points, they should reflect at different times (though most likely not an issue for the frequency you were looking at). The only complexity would then be adjusting the resistor values to match, assuming you have a known number of drivers. Perhaps that's why most back-plane based systems use end termination instead.
Great video thanks! As I understand, the "properly" terminated board does have source impedance of 100 Ω and estimated to have 100 Ω characteristic impedance of the PCB. However, it is not terminated in the end right? If you would have terminated it in the end, you would get zero reflections but then of course, you would only get halve the voltage through out the whole transmissions line...
Are the probe wires to your oscilloscope a transmission line as well? Or are the probe wires too short to form reflections? Could that be a concern when doing these kind of measurements?
Yes! The probe is also a transmission line, but it doesn't affect the circuit too much since its series resistance is 9Mohm, terminated by 1Mohm. It affects the line more because of its capacitance. My probe is a 500MHz probe, any higher and you need an expensive active probe.
Here I hate to be that guy that asks for the baseline data but what did the signal generator signal really look like before you connected it to the "reflection" stick? Merely curious how square was this input square wave or was there some high freq ringing at the edges or .. what?
Exactly the scope I would like to buy. The DPO series.
Great Video!
Awesome how much Detail you can get with that Scope
Also cant wait for the next Part of the Risc-V Project
Man that scope is huge. Looks like it has a DVD drive...can it play movies? Lol.
Seriously though, this was a great explanation, And I believe I understand these concepts much better now. Thanks so much for sharing your discoveries!
Great video! I've never seen this so clearly explained before, thanks!
Awesome
At first, I was wondering at the back of my mind how good resolution that scope would have to have to see such fine details in a 3 ns pulse, but it wasn't until I was halfway into the video that I started realizing that that would havta be a really expensive scope, and then I saw it had four channels as well. I then Googled the retail price to, indeed, be $21,900. Pretty nice hobby equipment you have sitting around on that bench!
Ha ha, no, I got this one surplus from equipnet.com for $2100. Nobody wants one with a bashed-in case no matter how well it works.
so good, thank you very much.
woah a DPO7104, that must have cost a bomb. I love your videos. please keep them coming!
Cj Chin Surplus :) Case was bashed in, but it's built like a brick outhouse, so it works fine.
amazing. thanks.
Thanks ;)
Hi Robert I have a question. let's say I have both source imp and characteristic imp 50 ohm. should we have 50 ohm as well on the load impedance to prevent a reflection?, we know that mostly our load is high impedance, should I parallel terminate it with a resistor close to the load? load impedance e.g another CMOS buffer of MCU, or maybe a transistor.
Awesome video! It's nice to see this in application to PCB traces, especially since most sources (and university courses) cover it using coaxial cables. Would you be willing to do a comparison of other termination techniques such as an end termination and the voltage divider end termination? (I have seen all three used for different DRAM types / signals) It would also be interesting to see how the termination technique changes the behavior if the line is loaded (i.e. if you have multiple tri-state drivers with series termination for example - with only one of them driving at a time).
As for the propagation speed, that needs to be calculated from the distributed capacitance and inductance of the transmission line, which will depend on many factors (trace width, PCB thickness assuming a ground plane, clearance to other traces, solder-mask thickness and material, near field metallic objects, etc. - A PCB trace is an antenna). Most undergraduate physics students do this in their electronics lab using a coaxial cable and a fast pulse generator (you can also measure the transmission line impedance by using an end termination resistor of known resistance). Note that the impedance and propagation speed will also change with frequency (and therefore rise time and the Fourier components of the square wave signal), which is most likely why the return pulse appears to be less sharp.
I will probably be doing this. One issue I realized just as I finished the video upload is that I didn't cover one important case to me, which is a driver with a series termination in the middle of the line. My original thought was that it would act the same, but then I realized that the two transmission lines would present an effective impedance of 50 ohms (2 lines of 100 ohms in parallel), so the series termination would have to be 50 ohms. I want to test that, so round 2 of order-PCB-from-China is on!
That's what I meant by "loading the line", though your description is more accurate. I think you are correct in terms of the impedance changing, which would mean that the impedance would change with each new device on the line. In addition, since they are at different points, they should reflect at different times (though most likely not an issue for the frequency you were looking at). The only complexity would then be adjusting the resistor values to match, assuming you have a known number of drivers. Perhaps that's why most back-plane based systems use end termination instead.
Great video thanks! As I understand, the "properly" terminated board does have source impedance of 100 Ω and estimated to have 100 Ω characteristic impedance of the PCB. However, it is not terminated in the end right? If you would have terminated it in the end, you would get zero reflections but then of course, you would only get halve the voltage through out the whole transmissions line...
Are the probe wires to your oscilloscope a transmission line as well? Or are the probe wires too short to form reflections? Could that be a concern when doing these kind of measurements?
Yes! The probe is also a transmission line, but it doesn't affect the circuit too much since its series resistance is 9Mohm, terminated by 1Mohm. It affects the line more because of its capacitance. My probe is a 500MHz probe, any higher and you need an expensive active probe.
What is the value you use and how for making it properly terminated?
Here I hate to be that guy that asks for the baseline data but what did the signal generator signal really look like before you connected it to the "reflection" stick? Merely curious how square was this input square wave or was there some high freq ringing at the edges or .. what?