EEVblog 1445 - How to Simulate an Oscilloscope Probe in LTSPICE
ฝัง
- เผยแพร่เมื่อ 20 ธ.ค. 2021
- How to simulate a x1 oscilloscope probe in LTSPICE.
And how to use the lossy transmission line model in practice with the spice directive.
Mysteries of x1 oscilloscope probes revealed: • EEVblog #453 - Mysteri...
Forum: www.eevblog.com/forum/blog/ee...
Subscribe on Odysee: odysee.com/@eevblog:7
EEVblog Web Site: www.eevblog.com
The 2nd EEVblog Channel: / eevblog2
EEVdiscover: / eevdiscover
Support the EEVblog through Patreon! / eevblog
AliExpress Affiliate: s.click.aliexpress.com/e/c2LRpe8g
Buy anything through that link and Dave gets a commission at no cost to you.
Donate With Bitcoin & Other Crypto Currencies!
www.eevblog.com/crypto-currency/
T-Shirts: teespring.com/stores/eevblog
#ElectronicsCreators #Simulation #LTSPICE - วิทยาศาสตร์และเทคโนโลยี
I think you forgot to divide the measured resistance by 1.2, as you did with the capacitance. Still, we *do* get the idea so it's all good.
Oh damn, so I did. Nice catch. I should have just set the unit length to 1.2m and then no dividing required.
@@EEVblog The directive is expecting values per unit length for resistance, capacitance (and inductance).
So, I guess you mean: Set unit length to 1.0 and then use absolute values for R, C and L!
Well, there are many other mistakes too. For example not realizing that the compensation is only for 1/10 divider. Or not realizing that the transition line model is not just "doesn't like floating input" but it actually lacks any conduction between its ports. Not a high resistance, but an inductor with series resistance must be inserted between input and output GNDs, or for an even more precise simulation an other transmittion line with grounded shield. Or the totally wrong inductance calculation. The characteristic impedance of a probe coax is definitely not 50 ohms, but much higher, in order to keep capacitance low. Typically the inside conductor is made of a very thin wire, 50um or less, and it is surrounded by air, not a high epsilon insulator, also to keep capacitance low. The small diameter affect characteristic impedance as well, pushing it well above 100 ohms, and inductance around 1 uH/m. If you want to calculate this TL inductance from capacitance, then you should start from delay! Since the insulator surrounding the wire is mostly air, the propagation delay is higher than 3.3 ns/m, but not by very much. Maybe 4 ns/m. And it equals to sqrt(LC). This gives you a much better estimation on L. But the C estimation of the coax is wrong already, because the specified value of datasheet includes everything else: input of the scope, and any other components, which are not known. In many cases there is an RC network at the connector, but for this particular probe you can not know without actually cut it open.
Inductance could be determined correctly by measuring its diameter and the inside diameter of the shield. Their ratio is typically ~40, what results an inductance of ln(40)×1,26E-6÷6,28
=0,740125 uH/m
Etc...
Very interesting stuff! I clearly have a lot to learn on the practical side here - I admittedly approached Derek's question as a physicist not an EE... On the validity of the signal I measured, my current (ha) confidence stems from the fact that when I spread the wires out into two big loops, the flat "pre-lightspeed" transient dropped to zero, so I doubt it was actually CAUSED by the probes, but the probes may have significantly messed up my measured amplitude on an absolute scale - I anxiously await the results of your simulation! Frankly if I got it right, I was lucky - I SHOULD have cared about the probes' impact more than I did. I actually had all four wired up and tuned in 10x mode in the morning, but there was so much noise on the line at one point in frustration I flipped everything back to 1x and never remembered to set it back... I'm glad there are good engineers out there with practical knowledge to keep the scientists honest =D
Actually I realized i should ask - I've got two questions that have been driving me nuts.
First of all, is there an accepted way to model cross-inductance on a transmission line? I know under "normal" transmission line use it doesn't matter, but in this weirdo case, I feel like the early current through the lightbulb could be a result of magnetic coupling in addition to electrostatic coupling, to phrase it in extremely physicsy terms...
Second, Electroboom's comment on my video (a few replies into our conversation) pointed out that the step height was weird considering that the "second" step, after one lightspeed delay, was approximately full voltage. In my mind, the intuitive explanation of "weak current followed by full current" didn't bother me very much, but if the early crosstalk current was occupying charge that otherwise would continue farther down the wire, then here would be a known ratio between these signals, and that's probably something EEs know really well, so I guess I'm wondering if the actual speed of light is poorly represented by inductors, or the more likely option, my probe capacitance screwed up the magnitude of the step I measured and it WAS approximately matched at 1k...
Unfortunately when I played with resistors, I also exchanged the resistors on the switch-side of the circuit, so I could be lazy and not rescale the traces on the scope, but if changing that near-side resistor was effectively changing my source impedance, the signal rise time, constituent frequencies, and the effective impedance of the capacitor between the wires, then I really didn't run a very good test when trying to match...
I'll see if I can package up all my trace screenshots with corresponding schematics and post them somewhere if people want to see
Given that you weren't really trying to nail down the first few ns, I doubt your probing mattered all that much. You were seeing steady results over a microsecond or so. Any capacitive loading at that time scale - insignificant. So you loaded the circuit down with a 1M resistor for all practical purposes, over that glacial (in scope terms) timescale.
@@AlphaPhoenixChannel Hi Alpha Phoenix: Thanks for your video. For Veritasium's ideal setup (ideal light bulb [turns on instantly]... ideal switch [no bounce]... etc. ) ... when the switch is closed, the light bulb will light, at about half brightness, in about 1/c... or about 3.3 nanoseconds. This is about how long (longer in a non-vacuum) it takes the displacement current to establish.
@@AlphaPhoenixChannel says, "First of all, is there an accepted way to model cross-inductance on a transmission line?" Well, there's inductance, self-inductance, mutual inductance, leakage inductance, kinetic inductance... but, what's 'cross-inductance?' Are you trying to calculate the inductance of your transmission line?
@@AlphaPhoenixChannel wrote " I really didn't run a very good test" Well if you had a fast pulse generator, and a sampling oscilloscope, with the proper probes, you would have better results. IMO, what you were doing is demonstrating a poor-man's TDR (time domain reflectometry), on a poorly characterised, and poorly matched transmission line. Apart from showing reflections on a long transmission line (something cable TV techs do every day)... To show Veritasium's claim, you want to see the pulse at the load, before any reflections... so, you need to have a faster scope.
Using LTSpice makes me feel like it's 1994 again. :-)
It is still an amazingly useful and relevant tool. Fun to see a probe considered and simulated.
One of the nicest things about LT-Spice (IMHO), is how you can just drop spice commands directly onto the schematic.
Last week I was shown a drive signal which ideally should be a rectangular pulse, but was oscillating. The engineer added a series resistor and got rid of the oscillation, but the rising edge was rounded. Thinking of a 'scope probe circuit, I suggested a variable capacitor across the resistor and simulated a simpler model. It consisted of a RC parallel circuit (R1 || C1) to model the probe, in series with a RC parallel circuit to model the 'scope input (R2 || C2). The source was a 0-to-1 V 1 kHz square wave. Simulations were done with compensation C too large, too small, and tuned for best pulse visually. Simulation was as-expected. ... Cheers!
Calling LTSpice GUI nice is kind of overstatement 😄
At least it's functional 😅 So many of the "new" styled interfaces are tedious and slow.
Practical, no frills. And yes, it gets some getting used to.
@@Thirsty_Fox I actually really don't understand why some new programs even bother. LTSpice is basically the de facto standard, otherwise MicroCap, which is also free these days.
LTspice... meh. It is a weird ass x86-only windows-only custom bodged jit architecture with virtually no interesting unique features, and proprietary to boot. People use it exclusively for its shitty bundled schematic capture tool (because the other free alternatives are even shittier).
@@Ormaaj That seems to be par for the course with electronics software. Heck, I've been using Cadence for IC design and layout for a bit and it reminds me of a road that's entirely made of patchwork. Pretty janky for an industry-standard. At least LT Spice is free!
Will never stop appreciating the analog knowledge dumps.
I just love the content waterfall that Derek (veritasium) has caused 😂
Almost like the Cunningham's Law lol:
"The best way to get the right answer on the internet is not to ask a question; it's to post the wrong answer."
(But yes, Derek is not wrong really. More like a bit misleading)
@@mskiptr annoyingly, knowingly misleading.
And not for a second will I believe he did it with a lofty goal of getting everybody to learn more about the topic just to prove him wrong.
Damn Dirk. Triggered us all. The response videos were flying up within 5 minutes of upload.
Learned a lesson, thanks Dave.
Thumbs well deserved again. Great tutorial
Looking forward to the next video on the subject (;
You are the best!
At first I saw the video title as how to stimulate an oscilliscope probe and I was thinking...yeah I can tell you how to stimulate my probe.
cool..thanks for the tips Dave.. always gotta remember capacitance, inductance and resistance is everywhere.. nothing is "ideal"
Some OpAmp models that come with LTSPice has also similar problems with grounding. Those encrypted models sometimes include GND referenced connection inside. If put to floating operation it will result in error..
I seen the thumbnail video just yesterday!
Dave, your older video was very much better! Almost everything was correct. Haven't you watched it before making this one? Back then you actually disassembled a probe and you've seen and said there is no series resistor. You have actually measured bandwidth. You said many other important things that you've got correctly then, but forgot by this time. :-(
very interesting
So it's all in the passive probe! incredible! I'm totally amazed! Well you're so helpful !!As they say in English! Leave a place for Uncle Bob at dinner !
Is that Windows 7? No way !
Should R=318/1.2 as the lead is 1.2m long?
initial value was right
Yep, I forgot that.
when you changed the R value you should have recalculated the L with sqr root as you did the first time. (Also divide R by 1.2 as others have noticed)
Simulation Software Micro Cap: versions 10, 11, and 12 by Spectrum Software are now free and require no key, just to mention...
This reminds me of that I wanted to figure out a good way to build 100x probes myself with good performance and minimal equipment for calibration...
Of course, you could also chuck a big ole resistor on that second ground, and it would be more like real life (I assume your system in test is not in a separate dimension from your oscilloscope).
Do we get an update on solar roads soon?
19:21 The best measurements / analysis of my circuits I have done is with a 15 DM multimeter from the 80s.
I got a little confused with the compensation network how after the transmission line it was parallel to the transmission line but when you put it before the transmission line you wired it in series. Seems to me that in the first instance it's going to jigger the output impedance. If someone could explain or send a link that explains, I'd appreciate it.
I read it "probe stimulation" and thought Dave was making a corresponding face
Hi Dave, how come you included a 50Ohm series resistor on the "scope input". I was under the impression we tend to have just a 1M to ground with an optional Series capacitor for AC Coupling. But when a 50 Ohm "Terminator" is fitted or switched in, we would get 50 Ohm to deck, still not in series.
signal generators usually have loZ and hiZ outputs, lo being 50ohms. maybe he was working under that assumption?
@@WildEngineering he's drawn the source impedance on the left, but I'm talking about R3.
@@Taylor_26GE93 Exactly ... this is confusing the way Dave drew it. It muddies the waters between 50 ohm and 1M ohm input impedances. The simulation should still work and be valid with a wide variety of source impedances, as in real life. If this sim works the way it does because of the 50 ohm source impedance, well then that should be stated as well. I also dont understand
If there were a series 50 Ohm with 15 pF, and the useful signal were on the 15 pF, then the bandwidth had already been limited to 211 MHz with infinitely fast input frontend, so this must not be a correct model for a 200+ MHz scope. This is one of the mistakes of many. The input impedance of faster scopes must be more or less resistive at high freq, around 50 ohms, that is true, but the arrangement of R and C elements are not like this. R3 can not be a single lumped element in the front. It must be near the very end, and most likely distributed. Slower scopes can be pure capacitive at high freq, 50 Ohm is not neccessary. But none of these 50 ohms are really important for simulating a 1X probe. Nor R6, which is actually not even exist. And R4 should rather be ~1 uH+0.1ohms. L parameter is also seriously wrong, actually it should be about 0.5...1uH/m, because the center wire is very, very thin, Z is very much higher than 50 Ohm (this is one of the goals of the design of a 10x probe). Basically more than half of the model is wrong, but this can not be revealed, because only 1 parameter is observed the whole time.
But Dave for the compensation cap on the BNC shouldn't that also consist of a series R that has the parallel adjustable cap? After all you want to compensate the scopes input capacitance! With a transient analysis you could even trim it as you do on the real probe.
BTW four pole cuircit analysis is based on solving systems of linear equations. This sometimes poses an ill conditioned inverse problem in which case the solution matrix becomes singular. Way back when in around 1980 I had developed a spice software running on a ZX Spectrum in pure assembly. Not using it anymore but it still runs on the ZX emulator much faster. ... Back in these days I was part of a team developing read amps for Winchester drives where I designed ultra linear phase filters. But that was in a different life. ...
You don't need all this to figure out the bandwidth of the 1x probe.. the RC low-pass cutoff frequency of your 250 Ohm and 100 pF gives you about 6 MHz.
Now where it would get interesting is in modeling a 10x probe, trying to get the bandwidth up to 500 MHz... That's where the lossy transmission line, plus your 68 Ohm termination (not compensation!) resistor and 25 pF cap are really needed!
The point is setting up the simulation and accounting for various effects, not finding the answer, which is easily available in the datasheet.
@@jaro6985 If there is no question, than any model is good, even if it is completely unreal. (Like in this case. There are some serious mistakes, but since there is no defined purpose, there is no possibility to realize the mistakes.)
And yes, the compensation/termination elements are there for 1/10 mode, and the whole structure is optimized for this. A real, dedicated 1/1 probe can be made much better, a simple DIY one I made 25 years ago can go up to about 15 MHz.
I'm just guessing, but I believe the purpose of this would be appreciate (and potentially mitigate) the disturbance introduced by the probe into the DUT, Heisenberg type of shit. I was working on a resonant circuit that could be "detuned" (i.e. ZVS couldn't be mantained) by the probe capacitances parallel to the load. Under similar circumstances it would be reasonable to work with some sort of "worst case", meaning, a crappy probe.
@@Eng_Simoes The word "appreciate" has the opposite value, it means something favorable, while what you are talking about is a bad effect. And what could "this" mean? The whole video? The "compensation"? The lossy TL? Or what? It could refer to a dozen of thing, but I can't think anything that would be correct and relevant.
And no, this is absolutely not a Heisenberg type of thing. Here there is a perfectly predictible load effect (which is *not* investigated in the video), while Heisenberg uncertainity (not "shit") talks about the essentially
unpredictable nature of the quantities themself, even without any external effect.
@@Eng_Simoes Here there was no mitigation, and if you want to mitigate the loading, then you have to use not "worst case", or "crappy probe", but a better probe (lower capacitance: 1/10 divider, or active probe)!
Hi Dave, I have been wondering how Analog oscilloscope like Tek 2467 display annotations or text data on same screen where traces run? Could anyone in the know clarify how they work?
Can someone help explain to me why the compensation network is in parallel with o-scopes termination on one end, but in series with the source resistance on the probe's front end input? I'm struggling to understand this.
Thanks
Should you have adjusted the value of the series resistor to get 6MHz without the oscilloscope input capacitance in circuit (like you did in the last simulation), since the datasheet should just be measuring the bandwidth of the probe and not the bandwidth of the probe/scope combo. That could be different depending on what scope you are using.
Sure, but the spec is a minimum and often takes into account the input capacitance of a typical oscilloscope. eg my 1x 35MHz rated probe is 34MHz measured on the scope, so its "real" bandwidth would be 40MHz+.
The whole method is wrong many ways. Actually the bandwidth of my PP215 probe is 8 MHz measured on the scope. (Specified to 6 MHz.) The series resistance would have to be negative if the method was correct. But actually that resistor is nonexistant, only a wrong speculation. And every single parameter of the probe in the simulation are calculated wrong (except for G :-) ).
Does anyone knows why did the LT spice software was trying to delete all my files in the D drive which I installed LT spice in? Luckily I stopped it after sometime but still I lost some of my data from that drive.
Dave, how long before you figure out Veratasium is out over his skiis?
He don't have snow for skiing. :-) On the other hand Dave calculated every single parameters of the probe simulation wrong. Sorry!
I thought that the compensation network was not in circuit in 1x mode. It does absolutely nothing to a calibration signal on the scope, when adjusted.
Is the compensation cap in the circuit for 1X? I thought it was only across the 9M resistor when switched into 10X mode. Perhaps I'm misunderstanding something...
Unfortunately you're correct in that it is for 10X range, but it is not clear which capacitance of the divider is trimmed. The lower part is more convenient and more safe because of the lower voltage.
@@PafiTheOne very good point. I’ve got a few probes with adj at probe end and some at bnc end. I’ll hook them up to my LCR meter with the bnc unconnected and set to 1x. If the adjustment affects the capacitance reading, then it is in the 1M “portion”. I’ll post back what I find.
@@bobwhite137 You're right, this tells where the compensation cap is connected to. For my PP215 Cin changes very much, so it must be parallel to the coax, not the upper part of the divider. And Cinmin is somewhere around 60 pF, so the coax capacitance must be a little lower.
It would be interesting to show the effect of the probe on a simulated circuit. Say a simple square wave oscillator, overlay the signal from the components that LTspice shows, and then through the probe so we can see the effect the probe has on the displayed signal. Bonus points for showing the result on a live circuit.
SimIlarly, showing how a probe can upset a circuit. Probe off, it works, probe on it fizzles.
Thank for an interesting series of oscilloscope and measurement topics.
We had the opposite situation on a high speed circuit. No probe, it was intermittent. With probe, it worked every time.
@@lawrencemiller3829 heheh now you have. Circuit yiu can substitute for the probe.
build your own oscilloscope with a needle ,a cloth ,3 matches and a compass
How about completing this video with the probe in 10X mode?
Somehow I found in LTspice, we can't set G nonzero in lossy Transmission Line.
Did you model the LTRA symbol by drawing it yourself?
Please address fidelity when th e simulation is excited by a step function. You should be ableto see thefidelity to aperfect square wave. This, I think, is important when thinkingabout probes. BTW I usually build my probe into a pcb.
What do you mean by build into a pcb?
@@Eng_Simoes If I design a circuit I think about points in the network that I would like to see on my oscilloscope. At that point I place a resistor and an appropriate connector for a coaxial cable. I plan to use a coaxial cable (50 ohms Z0) from my PCB to the 'scope. The 'scope I set at 50 ohms. The resistor between the test point (at the coaxial) and the waveform of interest may be 450 ohms, hence 10:1 ratio on the 'scope screen.
I did once design a very small PCB oscilloscope 'probe' made with two small pins made to match two small plugs designed into another PCB I designed. A 'home-made' oscilloscope probe.
Of course these all lacked a compensation network but because leads and the layout was very small and carefully laid out there was no need for compensation.
Hope this helps.
Ok, but what is the graph showing? I though AC source was set to 10kHz, and the AC simulation is running up to 20MegHez right? so what is it showing ? Hz on the X but what on the Y, I know its gain so It should be Vout/Vin, Vin being the AC source, but Vout where is it being probed?
The AC simulation overrides the manual source frequency.
@@EEVblog and the probe is places at the "oscilloscope" input right?
Vout is the top of R2/C1
9:56 No Keanu? :(
Huh, so soldering in a BNC socket to a test jig could give me different results compared to clipping my probe to it?
Yes exactly. For slow varying signals it might not make much of a difference, besides being a 1x divider, but for faster edges you would see oscillations and reflections.
7 days old video and no follow up to fix all mistakes?
Ps, 1/c sec is 3.3nS I'm not interested in anything after 3.7nS. that's a generous +/- 15%
1/c is 3.3 ns/m. :-) On the other hand 3.3 nanosiemens is quite a low conductance, causes only 0.33 % error in parallel with 1 Mohm. ;-)
There are numerous ways to deal with floating parts. The big resistor bodge of course being an easy one. Technically it only really needs an initial condition, though total lack of dc ground path can at times be fragile anyway. (Just a note for total noobs. You'll figure that one out pretty quickly.)
Yes, but the best way is inserting the actual impedance of the shield, which is roughly 1uH+100mOhm. In parallel with the GND side of the TL.
5.5 MHz vs 6 MHz bandwith. That's got to be the marketing department thinking, let's round it up in there!
It is spot on 6 MHz
@@diavalus Maybe I got confused which probe he measured and when he fiddled the numbers. He got 5.5 MHz at some point.
@@Valenorious okay, could be. Hope it all made sense in the end :)
I measured, and it is 8 MHz for PP215. Probably the same for PP510. Should've been measured instead of guessing and trusting a specification without tolerance. Every single parameters of the probe in the simulation are calculated wrong, except for G.
Use everycircuit for android. If you really need to fine tune a design just build it.
Never used LTspice (I'm more of an ngspice man), but wouldn't the R and G parameter be mutually exclusive?
No, R is in series with the wire, while G is in parallel with the insulator. They represent 2 physically different part of the cable.
@@PafiTheOne So G is a kind of shunt conductance? It does makes more sense to express it in terms of Siemens rather than Ohms
@@sniperwolf50 Exactly.
Starting model name with a number? I see that you too like to live dangerously :)
Does anyone know of a company that makes a BNC adapter for Tektronix oscilloscopes to fake out the probe detection (appropriate resistors inside the adapter) so a generic oscilloscope probe can be used?
Of course you mean probe *stimulation
Or does he ;)
Shouldn't "unit length" be the length of the CRO probe itself rather than some multiple/divisor of a meter? The datasheet doesn't have any of that "capacitance/meter" rubbish.
Oh. I thought you meant an alien probe.
0:24 how is the young guy in the background?
17:44 Sorry, but it's not cool, because the mere existence of that resistor is just a baseless speculation without any evidence, and the value calculation is done with many theoretical and practical errors. Neither the bandwidth, nor the capacitance specification has enough precision to allow calculating such tiny resistance with meaningful reliability when connected in series with such a huge coax resistance. Scope manufacturers almost always specify their products with some headroom in frequency range as you know. Actually I have measured the bandwidth (-3dB point) of a PP215 probe, and it is about 8 MHz. With this bandwidth, according to your method the value of R6 should be strongly negative, which is obviously impossible, so your method can not be correct.
Nevermind that his speech has gone a little flat in tone overall, he is working up to that 800VDC rail in the lab, and that's how we like to see circuits proven empirically. Also missed the explainer when he is reading off 175 "puff" tbh...on and on, pμF unit makes no sense...
What Bob is Your Uncle ?? what does that mean?
I can help you with that but i have no idea about anything else he said, he kept talking about puffs and he didn't even have a joint 🤷♂️.. bobs your uncle is just an old saying that means it works
It's a saying about some politician that went to service just because his uncle was already a powerful one.
@@markshort9098 lol, he fries and smokes capacitors
It means you're in like family, not too much an outsider.
I believe the reason the simulated bandwidth is showing lower than it should is because LT spices uses 20*log(V) instead of 10*log(V) for calculated dBs. So the cut-off frequency is actually at -6dB. I remember getting burned by this in the past
Nooo, the cut off is at -3dB and that's the half power point, aka where the voltage is at 1/sqrt(2) (approx. 0.707) of the initial value, because power is proportional to voltage squared, that's why that 20 is there.
I have about 5% of my portfolio in uranium bitcoin any advice on any other crypto currency that I can grow my $300k capital to a million dollars?.
$16,000just in two weeks Mr Williams Charles you are so amazing.
I have also been trading with him, The profits are secured and over a 100% return on investment directly sent to your wallet. I made up to $27,000 in 1month trading with him
@@grahamjose3052 I have really heard a lot about him,how can I contact Williams Charles?
〽️＋𝟭𝟰𝟬𝟲𝟳𝟰𝟴𝟲𝟯𝟰𝟰〽️
Good morning sir,iam your other subscriber here in the philippines.sir I want to have a multimeter like fluke... very long year, wishing that item.so ,why I mgs you thru your videos,I want you to help me to have it.pls..