Just discovered this series of videos... loved the antenna series! Just FYI, we have the same SA, the HP8591E, FYI-2, there is a soft menu switch to flip the input impedance from 50 to 75 Ohms, go to Amplitude, then More (page 2) and there is an input Z option switch... 4th menu softkey. Hope it helps. Steve
Thanks. The 8591E is a great machine. The setting you described is great and even stays active after power cycling! (I think I have mine configured to recall 'last' configuration). I picked mine up on Ebay back around 2010 and it's gotten a lot of use. I found it for $800 listed as needing some repair, but when I got it, it worked fine. I was almost missing having the opportunity to try to fix it. It has developed a vertical sweep non-linearity issue (moderate) in the last few years. So maybe I'll get my chance after all ;-) But it settles out after about 15 minutes 🙂
Great Video! MSAI-Guy did a 75-50 conversion on his analyze: #866 HP 8711C Working at 50 ohms that didn't look too involved. Anyway appreciate all your videos and efforts to make them! 73...
I believe that you are also seeing the effects of the added cable in you experiments. Especially in your test with the crocodile clip cable you can see quite some rotation effect from the phase shift in that cable
I just discovered your channel. Great video. I never measured the input impedance of the 10:1 and the non terminated input of a scope. Very interesting to see.
Thanks. I first noticed this issue when probing an RF amp I had built. It's gain dropped when the probe was on the circuit (it should have only shifted frequency due to the 12pF added in shunt in the LC circuit). Checked and sure enough, it was shunting with C, but also something in the 1K region at 100 MHz. Fundamentally they kinda have to add a 50 Ohm in series with a capacitor inside the scope in order to deal with the inevitable 50 (or 75) Ohm cable that's part of the probe :-) Otherwise the scope's high-Z (1M) input would look like an open circuit at that end of the cable and the other end would be a short when the cable is lambda/4 long...
@MegawattKS Right, you would get reflections from the mismatch. I always terminated at the scope for high speed measurements. If I was worried about loading, I would use an active probe. I never tried to see the response. This was very interesting. You should write an article about it
@@twin1q Good scopes appear to install an AC connected 50 Ohm resistor to ground (or something like it) inside the scope - to prevent that. Sadly, some of the new low-cost portable scopes coming out don't have it. But all the premium ones I've tested appear to. I love active probes too! Found a couple on Ebay for a few hundred dollars that we have used in our lab when needed. Definitely important in any RF or high-speed digital lab as you said :-) FWIW, here's what I gave to students in our classes though. The PDF at the top of the page is a two-page handout that warns them about this issue :-) ecefiles.org/rf-circuits-course-section-9/
Hi MagawattKS, as you said the 75 ohm resistor that you show is a series resistor and therefore it is not necessarily providing a 75 ohm impedance termination for the probe (it would have to be parallel if it was acting as a "termination for 75 ohm cable"). If you do the math on a scope input that has a 1 Meg ohm in parallel with 15 pf specification as an example (which is what my Rigol scope has), then the input impedance is as follows (using complex math) : 1 MHz : Input Impedance = 112.56 - j 10609 ohms, 10 MHz : Input Impedance = 1.13 - j 1061 ohms, 100 MHz: Input Impedance = 0.01 - j 106.10 ohms. Therefore you can see that the parallel capacitance value is what drives the input impedance so low as frequency increases. P.S. you might consider changing your statement at the 0:46 mark (also at the 7:29 mark), because the impedance is indeed typically based on 10 Meg ohm in Parallel with 10 or 20 pf when using a 10 X probe and it's the parallel capacitance that drives the input impedance down so much with increasing frequency (do the math to better understand the impedance versus frequency). The capacitive reactance goes lower with an increase in frequency and this is what drives the input impedance lower with an increase in frequency for the parallel combination of resistance and capacitance. Just FYI, Don
Hi Don. Thanks for the comments! I figured this would be controversial and was hoping it would generate discussions :-) If I understand your 2nd comment, Zin should be 0.01 - j 106 at 100 MHz, but it's not in the measurements. It's closer to 70 - j 100, not 0 - j 100. So at RF it is much closer to test structure 7 on the demo board (C in series with R) than structure 9 (the pure cap - which is what 10M || 15 pF would tend toward at high frequency where Zc falls with frequency). The actual equivalent circuit for the 2215 scope is 75 Ohms in series with 1M || 30 pF. With the probe it is a little more tricky because of the coax involved. The bottom line is the measured Q is _way_ down from what it should be. I reason that this is because the amplifier provides C to ground in parallel with 10M, which tends to -j0 at high frequency, while the 75 resistive (R101) remains. And I reason this is to terminate the coax (probe cable). Please see the final slide at the end of the video with the yellow text inset on the chart to see the series to parallel impedance conversions from measurements of an Agilent unit with an Agilent probe. This is what messed up the gain and bandwidth of my RF amplifier at 100 MHz. The problem arises from the need for a 1 to 2meter cable between the probe tip and the scope. If this were terminated in 10M || 30pF inside the scope, then the 9M || 10 pF in the probe tip would drive into a highly uncontrolled impedance as frequency goes up and the cable does its transmission-line thing. The only fundamental solution appears to be the use of an active probe to create a low-impedance drive to the terminated coax (cable used in the probe). Thoughts ? 73s, Bill
@@MegawattKS Hi Bill, I will have to think more about my big picture response to you because the typical X1/X10 scope probe is even more complicated by a little known fact that the center wire in the scope probe coax does not have a low resistance (unlike normal coax), and you can check this by switching your scope probe to the X1 mode and then measure the resistance of its center conductor from end to end and you will see that it probably has a resistance of 100 to 1000 ohms and I believe this is to prevent ringing when used as a X1 probe based on some obscure documents I recall reading. I just measured my X1/X10 probe master 1 meter long probe and its center conductor measures 312 ohms. It's very possible this resistance and/or the 75 ohm resistance you see in the schematic causes the impedance measurement discrepancy you mention versus the values I said it should be based on just pure math (an added series resistance would definitely cause a discrepancy similar to what we are seeing). Bottom line is that the impedance of the scope probe loads down devices more at higher frequencies because the impedance drops due to the capacitive reactance dropping at higher frequencies which bypasses the pure resistance within the probe circuit because a majority of the resistance in the X10 probe circuit is in parallel with the capacitors in the probe circuit (this should not come as a surprise, but probably causes a lot of problems when you forget to consider its effect on the circuit being measured). Interesting stuff and more complicated than one would imagine.
@@MegawattKS Hi again Bill, take a look at the 17:00 minute mark in this youtube video that discusses the scope probe cable in detail ruclips.net/video/OiAmER1OJh4/видео.html. This is related to my other comments about the scope probe center conductor having a relatively high resistance (I suspect this is why we have the measurement differences between what my math predicted versus what is observed on the VNA). Don
Thanks for the knowledge about the center conductor being high resistance wire ! I watched the EEVblog you linked. I also found some discussion of that in this document on page 15, after you pointed it out www.davmar.org/TE/TekConcepts/TekProbeCircuits.pdf . So in addition to messing with the Q inside the scope, they are doing this to degrade it in the cable as well. Wow. My poor little RF amp circuit never knew what hit it ! ;-)
@@MegawattKS Hi Bill, thanks for the link to that Tektronix document, and yes it does a very good job describing the high resistance center conductor in the cable.
Interesting facts about input impedance(s). I am confused, though, about the scoop. As are some meant to work at high and very high frequencies. I also know that good practice is to terminate the input of the scoop when testing for amplitudes, like bandwidth sweep. So definitely very useful piece of information here about the loading issues. As you wrap up, RF engineers use SA or VNA at higher frequencies, how do we measure then IN CIRCUIT, without loading much. For scope that could be differential probe (but I suppose very expensive for higher frequencies). But what about SA as those are all 50 or 75 Ohm input? I know there are special AVO meters for RF....
Excellent questions. There are two main methods used at RF. One is a passive, resistive voltage divider probe. For example, put a 450 Ohm resistor at the tip (in series) and then just use 50 Ohm coax going back to the 50 Ohm test equipment input. That gives a 500 Ohm load on the circuits, and a 10:1 attenuation (20dB). 500 Ohms is often OK at RF, but one could also use a 4950 (or 5K) resistor, making lighter loading, at the cost of 40 dB attenuation. Here's an example of a 500 Ohm one sold commercially (we had a couple of these in our lab). A little pricey, but you can also make one yourself. Of course, don't forget about the need for DC blocking (and generally don't use on high-power or tube equipment). This one includes a DC block internally, rated for 50VDC: www.auburntec.com/auburn_INFO%20P20B.html#SPECIFICATIONS The other option is to go with an "Active probe". They have to be powered, have limited DC voltage tolerance (maybe 5V), and are quite expensive. But they can present light loading, like 100K in parallel with 0.5pF. We also had those in our lab, but they are a little cumbersome to use. But not bad...
@MegawattKS Thanks for the explanations. It probably all depends on the Zout of the circuit part we want to measure. There is one old type of analog AVO meter that is suited for these measurements. Meratronik V640. Probably designed and built in Poland, but there was cooperation with Canadian designers. And Marconi company also brought one under their name, but probably not designed themselves. The meter uses a special RF probe. DIY project for broadband diff-probe is one on my list. For SA, maybe to use 1x follower that have a high Zin and low Zout?
Hi, sorry for the delay in replying. Have you seen any of the "Antenna Briefs" series on this channel? ruclips.net/p/PL9Ox3wpnB0kqNLnCdCtWcjGq7nRVHYxKA . While this doesn't directly show how to measure antenna gain, the series covers the definitions and utility of knowing the gain, and all the related things in a wireless system. Episodes 1, 2, and 6 are probably the closest to what you're asking about. Indeed, at about the 2 minute mark into Episode 1, there is a setup where the NanoVNA is used to look at bandwidth and path-loss (in addition to return loss) within a TX/RX system. It also covers issues of size vs frequency. In general, antenna measurements are rarely done at 30 MHz and below because of size and difficulty with setup, and the NanoVNA wouldn't be used for that. But at UHF and above, it can be used with a suitable "Anechoic chamber", or with a suitable "range" environment without reflections. Also, gain is related to the antenna pattern, which is arguably the more important issue in many cases. See Episode 6 for that. Hope that helps some.
I didn't hear what frequency the scope probe was specified for. Nothing has a fixed impedance at all frequencies. Everything will be way off ultimately if you go high enough in freq2quency. That's why the frequency ratings matter so much.
Good question. The probe shown in the video is a cheap one that came with a 100 MHz scope I have at home. BUT - I've also done these measurements with high-end Agilent probes rated for up to 300 MHz. In particular the N2863A, which is a 300 MHz 10:1 passive probe. Same basic results. At 100 MHz, it measured 49 - j160 Ohms (which is equivalent to 570 Ohms || 9 pF in parallel form). The VNA showed it moving along the 50 Ohm circle from the right side toward the center on the Smith chart, just as all passive probes will (some may move along a 70 Ohm circle). As you said, nothing is fixed impedance as frequency changes. Here the fundamental problem is the cable, which is a transmission line with distributed L and C. To deal with that, the probe has to be terminated in its characteristic impedance at high frequencies at the scope input jack. A modern Agilent/Keysight 100 MHz scope we tested measured 50 -jX almost perfectly when the calibrated VNA was attached directly to the scope's input jack, reaching 50 -j198 at 100 MHz. The probe necessarily has a coax cable between the scope tip and the scope input jack. Hence, the input impedance seen at the scope tip must morph from the proverbial 10M || 15 pF (or whatever) to something tending toward 50 Ohms resistive at high frequency, since there is no choice in a passive cabled structure if one is to avoid wild impedance and frequency response swings. See the comments/discussion below from WD8DSB and G0HZU for a deeper dive into some of the complexities the probe designers dealt with using resistive center conductors/etc. (The best solution for high frequency work is to use an active probe that has an amplifier at the tip that can present something like 100K || 0.5 pF, and then buffer that down to 50 Ohms to drive the cable run. We have a few of those in the lab for RF work and they're great - although a bit cumbersome due to the need for powering :-) )
The marker reads out either R + inductance(or capacitance), or R + jX. I think you have to go to the "Marker>Value" menu to select R+jX if its currently reading out the other.
@@davidc5027 Sorry - I am not familiar with that software. I would say try it, and see if it matches what you can visually read from the Smith Chart (with a marker set on a particular frequency and looking at what arc you're closest to). Also if it reads out as X+jY format, it's most likely showing reactance as the Y (though it could be admittance).
Also take an ohm meter and measure the resistance right at the input connector on your scope and you will measure very darn close to 1 Meg Ohm. If you measure your 10X scope probe with your ohm meter when the 10 X probe is connected to your scope you will indeed measure close to 10 Meg Ohm. Also since the parallel resistance is so high you can typically measure the parallel capacitance of the scope probe as well as the scope using a capacitance tester and get reasonably close to their published values. I used my smart tweezers to measure capacitance and measured the parallel input capacitance of my scope and it measured 17 pf which is very close to the published value of 15 pf as an example. I also measured the parallel capacitance of my 10 X probe when connected to my scope and it measured 14 pf. Therefore you can see that the typically stated values of parallel resistance and capacitance is indeed correct for oscilloscopes as well as their 10 X probes.
Thanks. I agree. At DC, audio and low frequencies. But it is not the proverbial 1M || 30pF (for scope), or 10M || 15 pF (for 10X probe case) at RF. I think what happened probably is that in the early days, there was no 75 Ohm termination for high frequencies since scopes stopped around 1 MHz or maybe 5 MHz, and cables/probe lengths were a small fraction of the wavelength. But as scopes went to 100 MHz and above and tried to maintain 1.2, 1.5, or 2 meter cable lengths, they had to add the 75 Ohm termination in series with the 1M || 30pF of the input amp. Please see reply to your second comment below also. Thanks for this clarification !
At frequencies above a few MHz a useful rule of thumb is that the loading effect of a typical x10 scope probe will look like 12pF in series with about 70 ohms resistance rather than 12pF in parallel with 10 Megohm. This is because there will typically be 12pF across the 9Meg tip resistor and this will appear to be in series with the input to a lossy transmission line (deliberately lossy coax used in the scope probe that will look like a 70 ohm resistor). Some x10 probes might load the circuit like 12pF in series with 50 ohms and some might be 12pF in series with 90 ohms. But many will fall within this range. In terms of the parallel equivalent this means the x10 probe will load the circuit with about 12pF in parallel with 25k ohm at 10MHz and it will load it with 12pF in parallel with only about 6500 ohms at 20MHz. By 50MHz it will be down to 1000 ohm in parallel with about 11pF. At these frequencies this is a long way away from 10Megohm in parallel with 12pF.
@@G0HZU Hi Jeremy, thanks for the "rule of thumb" and you will see down below that MegawattKS and I do discuss the lossy center conductor farther down in our discussion about scope probes. I did not include the lossy center conductor in my original scope probe impedance versus frequency examples that I provided MagawattKS down below, but then I brought up the fact that scope probes have lossy center conductors and that was the reason why my impedance estimates at various frequencies did not exactly match what the VNA measures. Also easy to measure the lossy center conductor resistance if your probe can be switched to X1 mode. MegawattKS also provided a link to an excellent paper discussing the lossy center conductor down below.
Sorry - I don't make that public. I prefer to do questions through the comments so they're available for others to see. Go ahead and ask and I'll try to keep the reply concise 🙂
Ok fair enough was hoping we could work together on a project I was starting but I understand thanks anyway I guess we cannot work together I wish your channel well thanks.
Hi Edi. The short answer is you need to calibrate the nanoVNA to the feedpoint of the antenna and then read out the impedance as discussed from 2 to 4 minutes into the video. To calibrate to the feedpoint, the short, open, and load used for the cal need to be placed at the end of the cable where the antenna is normally connected. However, if you don't need to know details of R and X, and just want to know if the antenna matches the cable impedance, you can do what is shown in this video: ruclips.net/video/hLPNOp0RXD8/видео.html
@@edisukriansyah5230 -20 dB is excellent. -15 dB is quite good. -10 dB is OK for most situations. -10 dB means that 10% of power is reflected. Although that may not sound great, the effect on transmission range is negligible. If S11 = -6dB, I start thinking about needing to fix things.
A scope does indeed load a circuit more at high frequencies because of the PARALLEL capacitance, but you are measuring the SERIES resistance and capacitance on the VNA. You should not be comparing series and parallel impedances and expect them to be similar. If the VNA will not display R in parallel with C, then you need to do some maths to convert between them. There are several online calculators that do this. Agilent, Tektronix etc are not lying about the input impedance. The reason you are not getting results similar to them is that you are not measuring the same thing! Do you realise that the bridge of a VNA only works well with a VSWR of 5:1 or less, and less well at 10:1? That means they work reasonably well for impedances 5-500 Ω. A VNA is not suitable for MΩ impedances.
Thanks for leaving a detailed comment. To help out anyone reading this, I will provide a detailed reply (because the situation is complicated). Yes - the VNA reads out in series form (either R+jX or R+C) and that should be converted to parallel form if one is assessing the 10M || 15 pF loading usually assumed. If the probe tip to ground impedance is measured for an Agilent scope such as a DSOX2002A (or similar) 200 MHz scope, using a Agilent 10074C 10:1 passive probe, the measured impedance at 100 MHz is 49 - j160 Ohms using a well-calibrated HP8753C VNA. Well inside the Gamma=1 circle. In parallel form, this becomes 570 Ohms in parallel with -j170 Ohms (to two significant digits which is about what the VNA's accuracy is going to provide in this region of the Smith Chart). This is 570 Ohms in parallel with a 9pF capacitance - not the 10M || 15 pF marked on the probe. I would agree that Agilent and Tek are not lying, but would offer that "All models are wrong, yet some are useful" and details can affect some situations. Historically there was no issue because scopes maxed out at 1 MHz or so and the input Z was as marked. Even today, at lower frequencies the actual impedance will converge to the stated 10M || 15pF, certainly at audio - but not at RF. As you point out, we can't read 10M with a VNA. However, 10M||15pF would be on the right side of the chart at audio, and then hug the lower half of the chart at MHz and above as a pure capacitance does - not spiral inward as we see in the demo. For audio and low MHz work, 10M||15pF is still an OK approximation. But for RF work it isn't, as it can cause errors in circuit operation/measured performance. While the 9pF can be absorbed into a tuned LC tank circuit of an RF amplifier the remaining parallel R needs to be very high to avoid degrading the tank Q - which 10M would do. Sadly the actual 570 Ohm parallel resistive part can significantly broaden the bandwidth and lower the gain, due to de-Qing the tank circuit (especially of the collector resistance is something like the 1.5K we often used in class project builds). This is in fact where we first ran into this issue. The gain was too low and the bandwidth too broad for what was expected with 10M||15pF loading the circuit. Fundamentally, the reason for all of this is that the probe uses a coax cable, which must be terminated in 50 Ohms (or whatever the Zo of the cable is) at high frequencies to avoid strong reflections at the scope input port. So the the input circuits have to transition from 10M || 15pF at audio, to 50 Ohms at UHF to achieve good damping factor and response flatness (another viewer pointed out that some probes use lossy coax to also help dampen the response). Of course the solution is to use an active probe if available and affordable. Or my personal favorite for RF work, a 10x (500 Ohm resistive) or 100x (5K resistive) low-C passive probe such as a Tektronix P6150, or the less expensive Auburn P-20a or P-20b units :-) (HP made a nice 5K one - but I can't recall the probe's part number).
@@MegawattKS I tried some experiments using the following parts. * HP 54600B 100 MHz oscilloscope. Marked 1 MΩ // 13 pF on front panel. * Agilent 4285A 75 kHz to 30 MHz precision LCR meter. (mains powered). The LCR meter 4 terminals for Kelvin connections. * HP 100 Ω resistor with 4 BNC terminals (P/N 04285-61001) * HP 16048A 1 m extension test leads for LCR meter (these are for Kelvin connections, so have 4 coaxial cables) * Kirkby Microwave 42091A short for LCR meter (equivalent of Agilent 42091A short) * Kirkby Microwave 42090A open for LCR meter (equivalent to Agilent 42090A open) * Home made fixture for the LCR meter having a BNC plug to connect to the oscilloscope. * 1 m long cable with banana plugs at each end. First the cable compensation of the LCR meter was performed. This is necessary for the highest accuracy. * Attach HP 100 Ω resistor on front of the LCR meter. * Attach 16048A 1 m extension test leads and put the 42091A short on the end * Replace short with 42090A open * Replace open with the 100 Ω resistor . Next the fixture I made was attached to the HP extension cables and calibrated with a BNC short, and leaving the connector open. The fixture I made was tested with passive components and found to work very well. Trying to measure the input resistance and capacitance of the oscilloscope was problematic. What become apparent was that the input impedance varied as both a function of frequency, and the following * Whether the oscilloscope was connected to the mains supply * Whether the oscilloscope was connected to the mains, but switched off at the power switch at the front. * Whether the oscilloscope was switched on or off, whilst connected to the mains. * Whether the 1 m long cable was used to connect the earth terminals of the LCR meter and oscilloscope, whilst instruments were powered on. I observed the following, where I have round results to 3 significant figures, despite the LCR meter has a 6 digit display. These are a few of the measurements I took. 30 MHz oscilloscope off, connected to the mains. 30.0 pF // 326 Ω. 30 MHz oscilloscope on, with ground cable 12.1 pF // 6.87 kΩ. 1 MHz oscilloscope on, no ground cable 12.4 pF // 490 kΩ 1 MHz oscilloscope on, a ground cable between scope and LCR meter 11.6 pF // 1.12 MΩ Both the oscilloscope and LCR meter are mains powered, from a 230 V AC supply with an earth conductor - standard in the UK. In summary, it was totally impossible to get any sensible results with these two mains powered instruments under the conditions I tested. I expect with mains isolation transformers and attention paid to stop ground loops it would be possible to make good measurements. But I'm not going to pursue this any more - making the fixture took several hours. Using a battery powered LCR meter would be better, as one does not need the accuracy of a high-end laboratory LCR meter. But since I don't actually need these measurements, and have better things to do with my time, I'm not going to pursue the matter any more. But if you have not noticed it, you are likely to get changes in the results due to grounding of the instruments.
Sorry for the delay in replying. For some reason, RUclips didn't notify me. Interesting data. Sounds like there could be "ground loop" or perhaps RFI pickup problems interfering with the 4285A measurement signals. Do you know the architecture of this instrument? It says it has 6-digit resolution, which leads me to believe (together with the f_max of 30 MHz) that it is not based on signal reflection like the VNA. FWIW, here are the measurements we did with the 8753C VNA. ecefiles.org/rf-circuits-course-section-9/ Also, if you have access to IEEE papers, the following might be of interest. I was involved in research on high-Q inductors and was annoyed with people reporting Q factors of, like, 150. As you said, VNAs are not good enough for that. So I wrote this paper after doing extensive assessments. As it turns out, the VNA is very accurate around the Gamma = -1 and Gamma = 1 points used in an SOLT cal. But for things like 0 + j100 Ohms, it's not very good at all and resonance techniques are needed (which themselves suffer from radiation losses :-( ) "Measuring and Reporting High Quality Factors of Inductors Using Vector Network Analyzers", IEEE Transactions on MTT, 2010. ieeexplore.ieee.org/document/5428814
@@MegawattKS It’s funny, but RUclips emailed me when you replied The 4285A uses 4 coaxial cables to connect to the DUT. Only the inner conductors are used for connectors to the DUT. It uses a 4-wire Kelvin connection with two cables (Hcur & Lcur) providing a current source and the other two cables (Lpot & Hpot) measuring the potential difference across the DUT. It uses an automatic balancing bridge. The range is 0.00001 Ω to 99.9999 MΩ. It measures from 0.00001 uF to 999.999 uF. Running the scope from a UPS would probably solve the problem too. Having spent several hours making the fixture, I should really investigate more, but I don’t have a lot of interest. I will take a look at the IEEE paper. Andrew Gregory at the National Physical Laboratory (NPL) has written quite an extensive paper on Q measurements (not just inductors). You will find it on the NPL website. eprintspublications.npl.co.uk/9304/3/MAT58.pdf
@@dr.davidkirkby1959 Thanks. I'll have a look at the report. Resonance is definitely useful for high-Q inductor assessment. Looks like a detailed treatment for that !
Just discovered this series of videos... loved the antenna series! Just FYI, we have the same SA, the HP8591E, FYI-2, there is a soft menu switch to flip the input impedance from 50 to 75 Ohms, go to Amplitude, then More (page 2) and there is an input Z option switch... 4th menu softkey. Hope it helps. Steve
Thanks. The 8591E is a great machine. The setting you described is great and even stays active after power cycling! (I think I have mine configured to recall 'last' configuration). I picked mine up on Ebay back around 2010 and it's gotten a lot of use. I found it for $800 listed as needing some repair, but when I got it, it worked fine. I was almost missing having the opportunity to try to fix it. It has developed a vertical sweep non-linearity issue (moderate) in the last few years. So maybe I'll get my chance after all ;-) But it settles out after about 15 minutes 🙂
Great Video!
MSAI-Guy did a 75-50 conversion on his analyze: #866 HP 8711C Working at 50 ohms that didn't look too involved.
Anyway appreciate all your videos and efforts to make them!
73...
Thanks !
I believe that you are also seeing the effects of the added cable in you experiments. Especially in your test with the crocodile clip cable you can see quite some rotation effect from the phase shift in that cable
This is great info to have. Thank you!
Glad it was helpful!
I just discovered your channel. Great video. I never measured the input impedance of the 10:1 and the non terminated input of a scope. Very interesting to see.
Thanks. I first noticed this issue when probing an RF amp I had built. It's gain dropped when the probe was on the circuit (it should have only shifted frequency due to the 12pF added in shunt in the LC circuit). Checked and sure enough, it was shunting with C, but also something in the 1K region at 100 MHz. Fundamentally they kinda have to add a 50 Ohm in series with a capacitor inside the scope in order to deal with the inevitable 50 (or 75) Ohm cable that's part of the probe :-) Otherwise the scope's high-Z (1M) input would look like an open circuit at that end of the cable and the other end would be a short when the cable is lambda/4 long...
@MegawattKS Right, you would get reflections from the mismatch. I always terminated at the scope for high speed measurements. If I was worried about loading, I would use an active probe. I never tried to see the response. This was very interesting. You should write an article about it
@@twin1q Good scopes appear to install an AC connected 50 Ohm resistor to ground (or something like it) inside the scope - to prevent that. Sadly, some of the new low-cost portable scopes coming out don't have it. But all the premium ones I've tested appear to. I love active probes too! Found a couple on Ebay for a few hundred dollars that we have used in our lab when needed. Definitely important in any RF or high-speed digital lab as you said :-) FWIW, here's what I gave to students in our classes though. The PDF at the top of the page is a two-page handout that warns them about this issue :-) ecefiles.org/rf-circuits-course-section-9/
Hi MagawattKS, as you said the 75 ohm resistor that you show is a series resistor and therefore it is not necessarily providing a 75 ohm impedance termination for the probe (it would have to be parallel if it was acting as a "termination for 75 ohm cable"). If you do the math on a scope input that has a 1 Meg ohm in parallel with 15 pf specification as an example (which is what my Rigol scope has), then the input impedance is as follows (using complex math) : 1 MHz : Input Impedance = 112.56 - j 10609 ohms, 10 MHz : Input Impedance = 1.13 - j 1061 ohms, 100 MHz: Input Impedance = 0.01 - j 106.10 ohms. Therefore you can see that the parallel capacitance value is what drives the input impedance so low as frequency increases.
P.S. you might consider changing your statement at the 0:46 mark (also at the 7:29 mark), because the impedance is indeed typically based on 10 Meg ohm in Parallel with 10 or 20 pf when using a 10 X probe and it's the parallel capacitance that drives the input impedance down so much with increasing frequency (do the math to better understand the impedance versus frequency). The capacitive reactance goes lower with an increase in frequency and this is what drives the input impedance lower with an increase in frequency for the parallel combination of resistance and capacitance. Just FYI, Don
Hi Don. Thanks for the comments! I figured this would be controversial and was hoping it would generate discussions :-) If I understand your 2nd comment, Zin should be 0.01 - j 106 at 100 MHz, but it's not in the measurements. It's closer to 70 - j 100, not 0 - j 100. So at RF it is much closer to test structure 7 on the demo board (C in series with R) than structure 9 (the pure cap - which is what 10M || 15 pF would tend toward at high frequency where Zc falls with frequency). The actual equivalent circuit for the 2215 scope is 75 Ohms in series with 1M || 30 pF. With the probe it is a little more tricky because of the coax involved. The bottom line is the measured Q is _way_ down from what it should be. I reason that this is because the amplifier provides C to ground in parallel with 10M, which tends to -j0 at high frequency, while the 75 resistive (R101) remains. And I reason this is to terminate the coax (probe cable). Please see the final slide at the end of the video with the yellow text inset on the chart to see the series to parallel impedance conversions from measurements of an Agilent unit with an Agilent probe. This is what messed up the gain and bandwidth of my RF amplifier at 100 MHz. The problem arises from the need for a 1 to 2meter cable between the probe tip and the scope. If this were terminated in 10M || 30pF inside the scope, then the 9M || 10 pF in the probe tip would drive into a highly uncontrolled impedance as frequency goes up and the cable does its transmission-line thing. The only fundamental solution appears to be the use of an active probe to create a low-impedance drive to the terminated coax (cable used in the probe). Thoughts ? 73s, Bill
@@MegawattKS Hi Bill, I will have to think more about my big picture response to you because the typical X1/X10 scope probe is even more complicated by a little known fact that the center wire in the scope probe coax does not have a low resistance (unlike normal coax), and you can check this by switching your scope probe to the X1 mode and then measure the resistance of its center conductor from end to end and you will see that it probably has a resistance of 100 to 1000 ohms and I believe this is to prevent ringing when used as a X1 probe based on some obscure documents I recall reading. I just measured my X1/X10 probe master 1 meter long probe and its center conductor measures 312 ohms. It's very possible this resistance and/or the 75 ohm resistance you see in the schematic causes the impedance measurement discrepancy you mention versus the values I said it should be based on just pure math (an added series resistance would definitely cause a discrepancy similar to what we are seeing). Bottom line is that the impedance of the scope probe loads down devices more at higher frequencies because the impedance drops due to the capacitive reactance dropping at higher frequencies which bypasses the pure resistance within the probe circuit because a majority of the resistance in the X10 probe circuit is in parallel with the capacitors in the probe circuit (this should not come as a surprise, but probably causes a lot of problems when you forget to consider its effect on the circuit being measured). Interesting stuff and more complicated than one would imagine.
@@MegawattKS Hi again Bill, take a look at the 17:00 minute mark in this youtube video that discusses the scope probe cable in detail ruclips.net/video/OiAmER1OJh4/видео.html. This is related to my other comments about the scope probe center conductor having a relatively high resistance (I suspect this is why we have the measurement differences between what my math predicted versus what is observed on the VNA). Don
Thanks for the knowledge about the center conductor being high resistance wire ! I watched the EEVblog you linked. I also found some discussion of that in this document on page 15, after you pointed it out www.davmar.org/TE/TekConcepts/TekProbeCircuits.pdf . So in addition to messing with the Q inside the scope, they are doing this to degrade it in the cable as well. Wow. My poor little RF amp circuit never knew what hit it ! ;-)
@@MegawattKS Hi Bill, thanks for the link to that Tektronix document, and yes it does a very good job describing the high resistance center conductor in the cable.
Interesting facts about input impedance(s). I am confused, though, about the scoop. As are some meant to work at high and very high frequencies. I also know that good practice is to terminate the input of the scoop when testing for amplitudes, like bandwidth sweep. So definitely very useful piece of information here about the loading issues. As you wrap up, RF engineers use SA or VNA at higher frequencies, how do we measure then IN CIRCUIT, without loading much. For scope that could be differential probe (but I suppose very expensive for higher frequencies). But what about SA as those are all 50 or 75 Ohm input? I know there are special AVO meters for RF....
Excellent questions. There are two main methods used at RF. One is a passive, resistive voltage divider probe. For example, put a 450 Ohm resistor at the tip (in series) and then just use 50 Ohm coax going back to the 50 Ohm test equipment input. That gives a 500 Ohm load on the circuits, and a 10:1 attenuation (20dB). 500 Ohms is often OK at RF, but one could also use a 4950 (or 5K) resistor, making lighter loading, at the cost of 40 dB attenuation. Here's an example of a 500 Ohm one sold commercially (we had a couple of these in our lab). A little pricey, but you can also make one yourself. Of course, don't forget about the need for DC blocking (and generally don't use on high-power or tube equipment). This one includes a DC block internally, rated for 50VDC: www.auburntec.com/auburn_INFO%20P20B.html#SPECIFICATIONS The other option is to go with an "Active probe". They have to be powered, have limited DC voltage tolerance (maybe 5V), and are quite expensive. But they can present light loading, like 100K in parallel with 0.5pF. We also had those in our lab, but they are a little cumbersome to use. But not bad...
@MegawattKS Thanks for the explanations. It probably all depends on the Zout of the circuit part we want to measure. There is one old type of analog AVO meter that is suited for these measurements. Meratronik V640. Probably designed and built in Poland, but there was cooperation with Canadian designers. And Marconi company also brought one under their name, but probably not designed themselves. The meter uses a special RF probe. DIY project for broadband diff-probe is one on my list. For SA, maybe to use 1x follower that have a high Zin and low Zout?
...I am looking into OPA858, maybe to make an probe with it (TIA - Trans Impedance Amp).
Can you make tutorial how to measure gain antenna using nano vna?
Hi, sorry for the delay in replying. Have you seen any of the "Antenna Briefs" series on this channel? ruclips.net/p/PL9Ox3wpnB0kqNLnCdCtWcjGq7nRVHYxKA . While this doesn't directly show how to measure antenna gain, the series covers the definitions and utility of knowing the gain, and all the related things in a wireless system. Episodes 1, 2, and 6 are probably the closest to what you're asking about. Indeed, at about the 2 minute mark into Episode 1, there is a setup where the NanoVNA is used to look at bandwidth and path-loss (in addition to return loss) within a TX/RX system. It also covers issues of size vs frequency. In general, antenna measurements are rarely done at 30 MHz and below because of size and difficulty with setup, and the NanoVNA wouldn't be used for that. But at UHF and above, it can be used with a suitable "Anechoic chamber", or with a suitable "range" environment without reflections. Also, gain is related to the antenna pattern, which is arguably the more important issue in many cases. See Episode 6 for that. Hope that helps some.
@@MegawattKS thanks you so much sir.
I didn't hear what frequency the scope probe was specified for. Nothing has a fixed impedance at all frequencies. Everything will be way off ultimately if you go high enough in freq2quency. That's why the frequency ratings matter so much.
Good question. The probe shown in the video is a cheap one that came with a 100 MHz scope I have at home. BUT - I've also done these measurements with high-end Agilent probes rated for up to 300 MHz. In particular the N2863A, which is a 300 MHz 10:1 passive probe. Same basic results. At 100 MHz, it measured 49 - j160 Ohms (which is equivalent to 570 Ohms || 9 pF in parallel form). The VNA showed it moving along the 50 Ohm circle from the right side toward the center on the Smith chart, just as all passive probes will (some may move along a 70 Ohm circle). As you said, nothing is fixed impedance as frequency changes. Here the fundamental problem is the cable, which is a transmission line with distributed L and C. To deal with that, the probe has to be terminated in its characteristic impedance at high frequencies at the scope input jack. A modern Agilent/Keysight 100 MHz scope we tested measured 50 -jX almost perfectly when the calibrated VNA was attached directly to the scope's input jack, reaching 50 -j198 at 100 MHz. The probe necessarily has a coax cable between the scope tip and the scope input jack. Hence, the input impedance seen at the scope tip must morph from the proverbial 10M || 15 pF (or whatever) to something tending toward 50 Ohms resistive at high frequency, since there is no choice in a passive cabled structure if one is to avoid wild impedance and frequency response swings. See the comments/discussion below from WD8DSB and G0HZU for a deeper dive into some of the complexities the probe designers dealt with using resistive center conductors/etc. (The best solution for high frequency work is to use an active probe that has an amplifier at the tip that can present something like 100K || 0.5 pF, and then buffer that down to 50 Ohms to drive the cable run. We have a few of those in the lab for RF work and they're great - although a bit cumbersome due to the need for powering :-) )
What setting measures reactance?
The marker reads out either R + inductance(or capacitance), or R + jX. I think you have to go to the "Marker>Value" menu to select R+jX if its currently reading out the other.
@@MegawattKS I'm using the VNA_QT_Windows program, and didn't know what drop down I should chose. I'm thinking it is Z-re
@@davidc5027 Sorry - I am not familiar with that software. I would say try it, and see if it matches what you can visually read from the Smith Chart (with a marker set on a particular frequency and looking at what arc you're closest to). Also if it reads out as X+jY format, it's most likely showing reactance as the Y (though it could be admittance).
Also take an ohm meter and measure the resistance right at the input connector on your scope and you will measure very darn close to 1 Meg Ohm. If you measure your 10X scope probe with your ohm meter when the 10 X probe is connected to your scope you will indeed measure close to 10 Meg Ohm. Also since the parallel resistance is so high you can typically measure the parallel capacitance of the scope probe as well as the scope using a capacitance tester and get reasonably close to their published values. I used my smart tweezers to measure capacitance and measured the parallel input capacitance of my scope and it measured 17 pf which is very close to the published value of 15 pf as an example. I also measured the parallel capacitance of my 10 X probe when connected to my scope and it measured 14 pf. Therefore you can see that the typically stated values of parallel resistance and capacitance is indeed correct for oscilloscopes as well as their 10 X probes.
Thanks. I agree. At DC, audio and low frequencies. But it is not the proverbial 1M || 30pF (for scope), or 10M || 15 pF (for 10X probe case) at RF. I think what happened probably is that in the early days, there was no 75 Ohm termination for high frequencies since scopes stopped around 1 MHz or maybe 5 MHz, and cables/probe lengths were a small fraction of the wavelength. But as scopes went to 100 MHz and above and tried to maintain 1.2, 1.5, or 2 meter cable lengths, they had to add the 75 Ohm termination in series with the 1M || 30pF of the input amp. Please see reply to your second comment below also. Thanks for this clarification !
At frequencies above a few MHz a useful rule of thumb is that the loading effect of a typical x10 scope probe will look like 12pF in series with about 70 ohms resistance rather than 12pF in parallel with 10 Megohm. This is because there will typically be 12pF across the 9Meg tip resistor and this will appear to be in series with the input to a lossy transmission line (deliberately lossy coax used in the scope probe that will look like a 70 ohm resistor). Some x10 probes might load the circuit like 12pF in series with 50 ohms and some might be 12pF in series with 90 ohms. But many will fall within this range. In terms of the parallel equivalent this means the x10 probe will load the circuit with about 12pF in parallel with 25k ohm at 10MHz and it will load it with 12pF in parallel with only about 6500 ohms at 20MHz. By 50MHz it will be down to 1000 ohm in parallel with about 11pF. At these frequencies this is a long way away from 10Megohm in parallel with 12pF.
@@G0HZU Hi Jeremy, thanks for the "rule of thumb" and you will see down below that MegawattKS and I do discuss the lossy center conductor farther down in our discussion about scope probes. I did not include the lossy center conductor in my original scope probe impedance versus frequency examples that I provided MagawattKS down below, but then I brought up the fact that scope probes have lossy center conductors and that was the reason why my impedance estimates at various frequencies did not exactly match what the VNA measures. Also easy to measure the lossy center conductor resistance if your probe can be switched to X1 mode. MegawattKS also provided a link to an excellent paper discussing the lossy center conductor down below.
Do you have an email address I have a question and really dont want to make a comment out of it as it may be a little long winded
Sorry - I don't make that public. I prefer to do questions through the comments so they're available for others to see. Go ahead and ask and I'll try to keep the reply concise 🙂
Ok fair enough was hoping we could work together on a project I was starting but I understand thanks anyway I guess we cannot work together I wish your channel well thanks.
how to measure impedance antenna using nanoVNA ?
Hi Edi. The short answer is you need to calibrate the nanoVNA to the feedpoint of the antenna and then read out the impedance as discussed from 2 to 4 minutes into the video. To calibrate to the feedpoint, the short, open, and load used for the cal need to be placed at the end of the cable where the antenna is normally connected. However, if you don't need to know details of R and X, and just want to know if the antenna matches the cable impedance, you can do what is shown in this video: ruclips.net/video/hLPNOp0RXD8/видео.html
@@MegawattKS thanks sir. Ideally, value of S11 = .... ?
@@edisukriansyah5230 -20 dB is excellent. -15 dB is quite good. -10 dB is OK for most situations. -10 dB means that 10% of power is reflected. Although that may not sound great, the effect on transmission range is negligible. If S11 = -6dB, I start thinking about needing to fix things.
@@MegawattKS thanks you for taking the time to help me
A scope does indeed load a circuit more at high frequencies because of the PARALLEL capacitance, but you are measuring the SERIES resistance and capacitance on the VNA. You should not be comparing series and parallel impedances and expect them to be similar. If the VNA will not display R in parallel with C, then you need to do some maths to convert between them. There are several online calculators that do this.
Agilent, Tektronix etc are not lying about the input impedance. The reason you are not getting results similar to them is that you are not measuring the same thing!
Do you realise that the bridge of a VNA only works well with a VSWR of 5:1 or less, and less well at 10:1? That means they work reasonably well for impedances 5-500 Ω. A VNA is not suitable for MΩ impedances.
Thanks for leaving a detailed comment. To help out anyone reading this, I will provide a detailed reply (because the situation is complicated). Yes - the VNA reads out in series form (either R+jX or R+C) and that should be converted to parallel form if one is assessing the 10M || 15 pF loading usually assumed. If the probe tip to ground impedance is measured for an Agilent scope such as a DSOX2002A (or similar) 200 MHz scope, using a Agilent 10074C 10:1 passive probe, the measured impedance at 100 MHz is 49 - j160 Ohms using a well-calibrated HP8753C VNA. Well inside the Gamma=1 circle. In parallel form, this becomes 570 Ohms in parallel with -j170 Ohms (to two significant digits which is about what the VNA's accuracy is going to provide in this region of the Smith Chart). This is 570 Ohms in parallel with a 9pF capacitance - not the 10M || 15 pF marked on the probe. I would agree that Agilent and Tek are not lying, but would offer that "All models are wrong, yet some are useful" and details can affect some situations. Historically there was no issue because scopes maxed out at 1 MHz or so and the input Z was as marked. Even today, at lower frequencies the actual impedance will converge to the stated 10M || 15pF, certainly at audio - but not at RF. As you point out, we can't read 10M with a VNA. However, 10M||15pF would be on the right side of the chart at audio, and then hug the lower half of the chart at MHz and above as a pure capacitance does - not spiral inward as we see in the demo. For audio and low MHz work, 10M||15pF is still an OK approximation. But for RF work it isn't, as it can cause errors in circuit operation/measured performance. While the 9pF can be absorbed into a tuned LC tank circuit of an RF amplifier the remaining parallel R needs to be very high to avoid degrading the tank Q - which 10M would do. Sadly the actual 570 Ohm parallel resistive part can significantly broaden the bandwidth and lower the gain, due to de-Qing the tank circuit (especially of the collector resistance is something like the 1.5K we often used in class project builds). This is in fact where we first ran into this issue. The gain was too low and the bandwidth too broad for what was expected with 10M||15pF loading the circuit. Fundamentally, the reason for all of this is that the probe uses a coax cable, which must be terminated in 50 Ohms (or whatever the Zo of the cable is) at high frequencies to avoid strong reflections at the scope input port. So the the input circuits have to transition from 10M || 15pF at audio, to 50 Ohms at UHF to achieve good damping factor and response flatness (another viewer pointed out that some probes use lossy coax to also help dampen the response). Of course the solution is to use an active probe if available and affordable. Or my personal favorite for RF work, a 10x (500 Ohm resistive) or 100x (5K resistive) low-C passive probe such as a Tektronix P6150, or the less expensive Auburn P-20a or P-20b units :-) (HP made a nice 5K one - but I can't recall the probe's part number).
@@MegawattKS I tried some experiments using the following parts.
* HP 54600B 100 MHz oscilloscope. Marked 1 MΩ // 13 pF on front panel.
* Agilent 4285A 75 kHz to 30 MHz precision LCR meter. (mains powered). The LCR meter 4 terminals for Kelvin connections.
* HP 100 Ω resistor with 4 BNC terminals (P/N 04285-61001)
* HP 16048A 1 m extension test leads for LCR meter (these are for Kelvin connections, so have 4 coaxial cables)
* Kirkby Microwave 42091A short for LCR meter (equivalent of Agilent 42091A short)
* Kirkby Microwave 42090A open for LCR meter (equivalent to Agilent 42090A open)
* Home made fixture for the LCR meter having a BNC plug to connect to the oscilloscope.
* 1 m long cable with banana plugs at each end.
First the cable compensation of the LCR meter was performed. This is necessary for the highest accuracy.
* Attach HP 100 Ω resistor on front of the LCR meter.
* Attach 16048A 1 m extension test leads and put the 42091A short on the end
* Replace short with 42090A open
* Replace open with the 100 Ω resistor .
Next the fixture I made was attached to the HP extension cables and calibrated with a BNC short, and leaving the connector open. The fixture I made was tested with passive components and found to work very well.
Trying to measure the input resistance and capacitance of the oscilloscope was problematic. What become apparent was that the input impedance varied as both a function of frequency, and the following
* Whether the oscilloscope was connected to the mains supply
* Whether the oscilloscope was connected to the mains, but switched off at the power switch at the front.
* Whether the oscilloscope was switched on or off, whilst connected to the mains.
* Whether the 1 m long cable was used to connect the earth terminals of the LCR meter and oscilloscope, whilst instruments were powered on.
I observed the following, where I have round results to 3 significant figures, despite the LCR meter has a 6 digit display. These are a few of the measurements I took.
30 MHz oscilloscope off, connected to the mains. 30.0 pF // 326 Ω.
30 MHz oscilloscope on, with ground cable 12.1 pF // 6.87 kΩ.
1 MHz oscilloscope on, no ground cable 12.4 pF // 490 kΩ
1 MHz oscilloscope on, a ground cable between scope and LCR meter 11.6 pF // 1.12 MΩ
Both the oscilloscope and LCR meter are mains powered, from a 230 V AC supply with an earth conductor - standard in the UK.
In summary, it was totally impossible to get any sensible results with these two mains powered instruments under the conditions I tested. I expect with mains isolation transformers and attention paid to stop ground loops it would be possible to make good measurements. But I'm not going to pursue this any more - making the fixture took several hours.
Using a battery powered LCR meter would be better, as one does not need the accuracy of a high-end laboratory LCR meter. But since I don't actually need these measurements, and have better things to do with my time, I'm not going to pursue the matter any more. But if you have not noticed it, you are likely to get changes in the results due to grounding of the instruments.
Sorry for the delay in replying. For some reason, RUclips didn't notify me.
Interesting data. Sounds like there could be "ground loop" or perhaps RFI pickup problems interfering with the 4285A measurement signals. Do you know the architecture of this instrument?
It says it has 6-digit resolution, which leads me to believe (together with the f_max of 30 MHz) that it is not based on signal reflection like the VNA.
FWIW, here are the measurements we did with the 8753C VNA.
ecefiles.org/rf-circuits-course-section-9/
Also, if you have access to IEEE papers, the following might be of interest. I was involved in research on high-Q inductors and was annoyed with people reporting Q factors of, like, 150. As you said, VNAs are not good enough for that. So I wrote this paper after doing extensive assessments. As it turns out, the VNA is very accurate around the Gamma = -1 and Gamma = 1 points used in an SOLT cal. But for things like 0 + j100 Ohms, it's not very good at all and resonance techniques are needed (which themselves suffer from radiation losses :-( )
"Measuring and Reporting High Quality Factors of Inductors Using Vector Network Analyzers", IEEE Transactions on MTT, 2010.
ieeexplore.ieee.org/document/5428814
@@MegawattKS It’s funny, but RUclips emailed me when you replied
The 4285A uses 4 coaxial cables to connect to the DUT. Only the inner conductors are used for connectors to the DUT. It uses a 4-wire Kelvin connection with two cables (Hcur & Lcur) providing a current source and the other two cables (Lpot & Hpot) measuring the potential difference across the DUT. It uses an automatic balancing bridge. The range is 0.00001 Ω to 99.9999 MΩ. It measures from 0.00001 uF to 999.999 uF.
Running the scope from a UPS would probably solve the problem too. Having spent several hours making the fixture, I should really investigate more, but I don’t have a lot of interest.
I will take a look at the IEEE paper. Andrew Gregory at the National Physical Laboratory (NPL) has written quite an extensive paper on Q measurements (not just inductors). You will find it on the NPL website. eprintspublications.npl.co.uk/9304/3/MAT58.pdf
@@dr.davidkirkby1959 Thanks. I'll have a look at the report. Resonance is definitely useful for high-Q inductor assessment. Looks like a detailed treatment for that !