ไม่สามารถเล่นวิดีโอนี้
ขออภัยในความไม่สะดวก
NanoVNA - Measuring S21 and S11 of a small-signal amplifier
ฝัง
- เผยแพร่เมื่อ 13 ส.ค. 2024
- Measuring a small-signal RF amplifier with a vector network analyzer (VNA). Highlights need for attenuators and shows how well the NanoVNA is able to calibrate through 20 dB of port 1 attenuation.
I have viewed several of your videos using the nanoVNA, and I find them very informative! I just bought my first VNA (the NanoVNA-F V2 from Sysjoint via Amazon). Thanks for these great videos - I look forward to trying out many similar tests! 🙂
That should be a good unit. I only have the NanoVNA-F. But I love it.
Nice presentation. Will use this information to get more familiar with my nanoVNA. Thank you for sharing.
Thanks! Glad it was helpful. You've got a great instrument now ! I still can't believe how well the NanoVNA works compared to the "big boy" $50,000 units in our lab (which admittedly have built-in attenuators and go higher in frequency :-)
That test amplifier board looks great for doing experiments!
I did it through "ExpressPCB" which has a nice free layout tool with a great tutorial help page, and fast 3-day'ish turnaround. Fixed size board of 2x3 inches allowed for a few different 1x1 inch layouts as seen in the first few seconds of this video as well as the one on RLC components :-)
What transistor did you use for this design? I'm trying to make a modified version for listening on the 2m ham band as an exercise
It was one from a collection of SMD parts and I don't recall the number - but I made sure it had a high enough fT value. In the university course, we used MMBR5179 transistors - which is an excellent choice for 2m IMO. The MMBR5179 is an SOT23 surface-mount packaged version of the classic 2N5179 (which I don't recommend due to the parasitics of it's leaded package). The fT is around 1 to 1.4 GHz for all of these devices, depending on collector bias current, so it has enough potential for gain at 146 MHz without undue danger of going unstable (as the 2N5179 variant is apt to do)...
@@MegawattKS great, thanks! I found some MMBTH10 in my collection that look similar enough to give it a go. SMD construction is indeed the plan, most of my component collection is SMD: the savings in part storage space alone already!
Thank you for this great series of videos. I am thoroughly enjoying them. I am not 100% sure about how you are measuring the gain through two attenuators. Is it that basically you calibrated the NanoVNA at the end of the extension, at those u.fl connector points? So that whatever NanoVNA shows is the actual gain of the amplifier? I am just trying to wrap my head around how the attenuators are accounted for in our reading.
Thanks for the comments. When using the attenuators, I generally just include them on the port 1 and port 2 connectors when doing a thru-calibration with a short SMA cable (which is often all I do if I'm only interested in S21). Then I switch to the cables with the u.fl connectors as needed (keeping the attenuators on the NanoVNA connectors). There is very little loss in cables and connectors so no adjustment is needed if S21 in dB is all we're interested in. Oh - and you'll have to change the Display>Scale>ReferencePosition menu item to move the plot down on the screen, since the NanoVNA likes to put 0 dB gain near the top. There are limits to how much attenuation you can calibrate through this way however, since the NanoVNA has some dynamic range limitations. 30 dB is the maximum total attenuation I feel comfortable calibrating through (e.g. 10dB on port 2 and 20dB on port1 for example). That works fine if the amplifier has gain of 30 dB or less peak. If one needs more than that (for testing a high-gain amp like 40+ dB), then I may add more attenuators when measuring the amp and then adjust for the amount added in my head. Hope that makes sense. (BTW - for doing S11 of an amp, one also needs like 20 dB on port 1 for many amps since the input circuits can go into saturation from the 0+ dBM level coming out of an unattenuated port 1 on the NanoVNA. Of course, in this case, I'll need to use the u.fl cables during the cal, and I use the short/open/load calibration sites on the RF Demo Kit board. I've been happily surprised that I can get a good S11 cal even with 20 dB attenuation on it. Some of our HP8753 VNAs in the lab often struggle with that - but this little NanoVNa is amazing :-)
@@MegawattKS After 2 years, I feel that I am understanding more about RF. I came back to this video again and still learning something. I was wondering if we could use NanoVNA to measure output impedance of the amplifier. (I guess it should be S22?? but I don't think nanoVNA can do that) Could one bias the amp and read S11 into the output port of the amp instead? Or any other ideas to actually measure the output impedance? (primarily in order to build the matching network effectively.)
@@pilobond Hi. That's good to hear. Yes - absolutely. As you said, it is "S22" and because of the symmetry of the definitions of S parameters, it is just the same as S11, except that you are measuring the other output port of the amplifier instead of the input port. Large VNA instruments often include a "test-set" that has internal switches to change the port for you. But that is not needed, and frankly I never used it much. It was much easier to just calibrate for S11 and S21, and if needed, swap the cables connecting to the amplifier myself (a full "two-port cal" is quite involved and only very marginally better in terms of accuracy). Just put a 50 Ohm termination on the amplifier input and with the VNA calibrated for a one-port measurement, look at the impedance at the output of the amp. This assumes you're using the same cable length for cal and measurement, and that you have an attenuator to protect the VNA (included during the cal) as talked about in the video.
@pilobond One question and a related note: What kind of amplifier are you looking at? High power amps are probably not something you want to do this with. They require high power attenuators (in case the amplifier starts oscillating) and too much attenuation to be able to get accurate return-loss measurements of the device under test. In addition, they are often not matched the same way as things like a low-power low noise amp. High power amps often have a very bad S22, but still work fine. But for a low-power LNA like in the video, just do the steps above. They often are better when matched, since the next stage in the receiver may expect to see a good 50 Ohm source...
@@MegawattKS I am a ham and like making my own gears. I have made a few transmitters with IRF510 final that can put out 6-25 W depending on the power supply voltage. That is on 7MHz. That was based on other people's design I copied from the internet. They often use a 1:4 choke/autotransformer for matching on a ferrite core. I wanted to understand better about matching and try other transistors. But was not sure how to obtain the output impedance.
Always learning something new when watching your videos, thanks. Could I use the NanoVNA to measure the input/output impedance of power regulators (DC/DC) powered on or off ?
Thanks for the feedback and question. Unfortunately I would have to say, generally speaking, no. The NanoVNA (and virtually all VNAs) only work on "linear" circuits. Power circuits are generally not linear. And the frequency range of 50 kHz and above for the NanoVNA means it probably won't cover the range you'd be interested in anyway - I would guess. However... if it's a so-called "linear regulator", and you are super careful to use a DC blocking capacitor of sufficient rating, and the IC is not outputting anything except DC (e.g. not oscillating), then it could give some output impedance info from 50 kHz and up, I think. Hope that helps.
MegawattKS, can you make more videos on measuring RF circuits using the PHASE measurements and electrical delay. I'm not sure why RF specs need to know the electrical delay and phase of a circuit for RF signals, any reasons why the phase and electrical delay will affect the RF signal?
Thanks for the suggestion. I might do a video on the design of the amplifier and if so, will try to remember to see if I can fit that in. Until then however, here's one answer:
There is a thing called "group delay". It's the delay over a band of frequencies such as the bandwidth of a filter. But that delay varies with frequency, even in the bandpass filter's passband. This is because the phase does not change linearly with frequency.
The upshot is that this can cause distortion in the demodulation of something like classical FM radio signals (not much, but measurable). We were able to measure it in the FM receivers we designed in our class, but honestly it wasn't perceptable since it was below about 1% for the filters we were using.
In digital signaling, I think maybe non-linear phase (non-constant group-delay) may cause changes in the waveshape and mess up the sensitivity some by degrading the so-called eye pattern in the recovered serial bits. Would have to try to remember and think through that more though...
Hope that makes some sense and helps some.
@@MegawattKS yes thanks, When making your new video on the RF amplifier can you show the compression points, gain, RF output wattage. Because the compression points will compress the FET and the RF output wattage will get lowered. 1dB compression Point and so on.
Hi
I need to know where can obtain projects that used in this video?
Sorry - the actual project assignments aren't available. The videos are just there to support the class lecture material. But I'll look into what could be done in the future. E.g. documenting the parts kit where some preselected transistor types and ICs are specified to constrain the design options some, and have things on-hand so we didn't run out of time for logistics reasons...
How did you calibrate NanoVNA? After the attenuator or before attenuator?
The attenuators were in-place (attached to ports 1 and 2) during the cal. As noted around 4:40 in the video, I was very happy with the NanoVNA's ability to "cal-through" the attenuators. Even the 8753 VNAs in our university lab can struggle with this much attenuation - but the NanoVNA took it in stride ! :-) th-cam.com/video/GAycxgbncHU/w-d-xo.html
@@MegawattKS Thanks.
I am going to buy two Baluns 1:1 50 ohm and cal after Baluns. I am trying to learn how to use NanoVNA with differential circuits.
The fact that you have calibrate after attenuator and tested your circuit, makes me more confident that I can calibrate after Balun.
Thanks for producing this clip.
I made a build,
0. RF Simulator: works great! numbers match calculations.
1. SMD(coplanar with ground plane, but pcb design wasn't any good for impendence controlled traces, home etched with no VIAs). Failed terrible. Input impedance was a total mess, SWR=8, almost negligible amplification :(
2. Leaded(components suspended in ugly construction method on a PCB plane): 9dB gain. But with ton of parasitics, had to stretch the Inductor in output matching network to much beyond its design to compensate for the parasitic effects.
The PCB used in your project, was it designed with impedence controlled trace widths/lengths for 100MHz? I don't have access to those Chinese $5 PCB manufacturers, to do a decent 2 layer PCB with via's and close ground traces :(
i did use lame 2N3904 though. have to figure out if any of transistors in sample book has decent fT.
2SC3356 did get it a little better to 9dB, but way off from 16dB gain.
ok, 12V supply gets it almost there. But wonder if for a 1mA current amp, it really needs 12v.
Stretching inductors is OK. You should see some of the ones we ended up with in class projects ! (but of course those need to be replaced with new ones with less turns) On the CPW, yes, I don't think they work well without good ground-via stitching. It always amazes me that simulator/calculator tools for CPW show them with no stitching. But that's the world of simulation. Too idealized and not faithful to the actual builds. On the PCB used in the video, it was just done through ExpressPCB since I like their layout tool and quick turnaround (and lack of soldermask for prototyping). At 100 MHz, wavelength is about 3m in free space and 1.5m on FR4 boards, so I didn't worry about controlled impedance traces. Everything much shorter than lambda/4. But inductive parasitics can come into play for some parts of the circuit, so the layout needs to be tight...
Maybe just keep the lower Vcc value and change the bias (e.g. Remitter) to get the current up.