|
I've searched through the message list looking for symptoms similar to mine and didn't find anything. I'm somewhat new to this and would appreciate any help you can offer.
I've been trying to get my NanoVNA-F (firmware 1.0.5) to show the results of a 40-meter low pass filter I've built. I know I'm doing something wrong here but can't figure it out. The VNA image is calibrated (including pass-thru), set for the range 6.5 to 29 MHz and logmag S21. Shouldn't this give a flat area leading up to the design frequency followed by a significant roll off through the remainder of the frequency range? My setup is very similar to W2AEW's YouTube demonstration of the NanoVNA with a LPF but gives entirely different results. The image is entirely different, with very large insertion loss and a steep area of suppression peaking at 11 MHz followed by a steep recovery.
I worried that my filter wasn't what I thought it was so I tested it with a 7 MHz square wave, looking at input and output with the FFT function of my 'scope. The before and after images show several things: (1) the square wave (yellow trace) is hardly a square wave. It wasn't much of one before attaching to the input of the filter but it's definitely loaded and distorted by the filter. (I can live with the distortion in the test, since the filter is supposed to reduce the high freq distortion anyway, right?) (2) The purple trace is the filter output--pretty nice sine wave. (3) The FFT shows a good 40+ db suppression of the harmonics, especially the odd numbers. So the filter seems to be doing its job.
There may be some things not right about my testing of the filter using signal generator and oscilloscope -- I'm new at this. For example, the shoulder on the filter output at about 10 MHz is troublesome, but still the logmag S21 output shouldn't be affected, should it?
Any help in understanding why my results look as they do would be greatly appreciated.
Best & 73,
Mitch NK3H
|
Please post a schematic of your filter. Assume it is designed to operate in a 50 ohm system.
|
Here's the filter I built. Laid out on copper plated board. Capacitors are all grounded on the copper ground plane. Thanks. Mitch NK3H
|
Great. Give a few moments to run a check on what we should be seeing.
|
Confirmed and looks ok... Mitch.
Assume that your layout is such that the inductors are ideally non coupled to each other, ideal each at right angles.
The connectors are panel mount? Soldered their body to ground and center conductor to the input pads. A photo
of you filter card would be useful. In any case, re calibrate and then check your calibration is ok by checking your
through cable fiving a S21 of nearly 0 dB and a S11 of at least 20 dB or better.
|
Also, I would do your setup and cal over a frequency range of 100 kHz to 30 MHz.
This way you should see the FLAT DC response of the filter, well below its cutoff and then
see its roll off and ultimate attenuation.
|
One last item, on your scope pulse gen measurement. Dial your pulse gen
back to 100 kHz, you should see the fidelity of the square wave as measured
at the output load termination R. As you increase frequency, the fidelity of the
square wave will suffer, of course, and for the most part you should see a sine wave
attenuated at the pulse input frequency.
|
1) Please post a schematic (already requested). 2) Did you measure the individual components before constructing the filter with them? 3) We assume the filter is designed for 50-ohms in and out? (already assumed) 4) Is the dielectric of the capacitors appropriate for the frequency range? 5) Instead of using the o'scope and FFT, please give us a plot of S21 using the NANOVNA. Please use SAVER (or equivalent) to accomplish your sweep with more points than are available on the native instrument. 6) If your signal generator can sweep, you can also measure the filter response with that. A flat output noise generator can also be used instead of the swept sig. gen. Dave - W?LEV On Wed, Mar 12, 2025 at 11:11?PM Mitch NK3H via groups.io <mitch= [email protected]> wrote: I've searched through the message list looking for symptoms similar to
mine and didn't find anything. I'm somewhat new to this and would
appreciate any help you can offer.
I've been trying to get my NanoVNA-F (firmware 1.0.5) to show the results
of a 40-meter low pass filter I've built. I know I'm doing something wrong
here but can't figure it out. The VNA image is calibrated (including
pass-thru), set for the range 6.5 to 29 MHz and logmag S21. Shouldn't this
give a flat area leading up to the design frequency followed by a
significant roll off through the remainder of the frequency range? My setup
is very similar to W2AEW's YouTube demonstration of the NanoVNA with a LPF
but gives entirely different results. The image is entirely different, with
very large insertion loss and a steep area of suppression peaking at 11 MHz
followed by a steep recovery.
I worried that my filter wasn't what I thought it was so I tested it with
a 7 MHz square wave, looking at input and output with the FFT function of
my 'scope. The before and after images show several things: (1) the square
wave (yellow trace) is hardly a square wave. It wasn't much of one before
attaching to the input of the filter but it's definitely loaded and
distorted by the filter. (I can live with the distortion in the test, since
the filter is supposed to reduce the high freq distortion anyway, right?)
(2) The purple trace is the filter output--pretty nice sine wave. (3) The
FFT shows a good 40+ db suppression of the harmonics, especially the odd
numbers. So the filter seems to be doing its job.
There may be some things not right about my testing of the filter using
signal generator and oscilloscope -- I'm new at this. For example, the
shoulder on the filter output at about 10 MHz is troublesome, but still the
logmag S21 output shouldn't be affected, should it?
Any help in understanding why my results look as they do would be greatly
appreciated.
Best & 73,
Mitch NK3H
-- *Dave - W?LEV* -- Dave - W?LEV
|
Thank you Alan. I've attached an image of the whole setup including the filter, the test rig and the VNA. Please note that I'm just using hookup wire to connect the filter to the testing rig--no SMA or BNC connectors. It made no visible difference when testing on the scope using the wires or direct to a leg of the input and output capacitors. The other image is with a wire connecting the input to the output of the test right. The result is 0db flat across the range for S21. The signal generator has 50 ohm output impedance, and the scope has a very high impedance input. Not sure about the VNA. I tried the setup with a 50 ohm resistor to ground on the filter output and found no difference in the VNA behavior. Thanks for your help. 73, Mitch NK3H
|
Please measure S21 instead of S11 using the NANOVNA! All S11 tells us is the match (which is pretty bad!). Dave - W?LEV On Thu, Mar 13, 2025 at 12:14?AM Mitch NK3H via groups.io <mitch= [email protected]> wrote: Thank you Alan. I've attached an image of the whole setup including the
filter, the test rig and the VNA. Please note that I'm just using hookup
wire to connect the filter to the testing rig--no SMA or BNC connectors. It
made no visible difference when testing on the scope using the wires or
direct to a leg of the input and output capacitors. The other image is with
a wire connecting the input to the output of the test right. The result is
0db flat across the range for S21. The signal generator has 50 ohm output
impedance, and the scope has a very high impedance input. Not sure about
the VNA. I tried the setup with a 50 ohm resistor to ground on the filter
output and found no difference in the VNA behavior. Thanks for your help.
73, Mitch NK3H
-- *Dave - W?LEV* -- Dave - W?LEV
|
Hi Dave,
S21 is the gray, I think (the not yellow one -- I'm color blind) print at the top right of the screen in the last two images I posted. It's line is flat at 0 dB across the frequency range with the test rig shorted and gives the unusual curve I showed when the filter is used. Connecting the Ports 0 & 1 of the VNA together (without the test rig) also show 0 dB flat across the frequency range. Is that what you meant?
Many thanks.
Mitch NK3H
|
Hi Mitch. Dave can assist and we have confirmed that you filter in indeed
set up for a 50 ohm source and load termination. The response is validation
from a simple AC simulation.
I looked at your photo's of your test measurement system and my first comment is: OUCH!
Neatness in construction of this RF stuff cannot be over emphasized.
Assume that your toroid L's are spot on or at least close to the desired values (you can measure them on your VNA)
and the C's all look reasonable... That said, try to build as though it were the final part. Get some RF
connectors, SMA or BNC or UHF and apply them to the in and output ports. And then using the appropriate
adapter, coax into the VNA test set. Dress those toroid L's up, short the leads and although it does not matter
as much with a toroid form, dress then at right angles to each other. Try to get the flow of your construction so that
it follows the flow of the schematic as posted.
|
Mitch,
It looks like the ground (copper plane) of your filter is not connected to the grounds (outer conductors) of the cables.
That would make a big difference.
--John Gord
|
One thing you could also try is running the filter in a program like Elsie (which is free, for Windows only, though) and look at the S11 plot from Elsie and see if it looks like what you're measuring for S11. A huge mismatch out of band isn't unusual, but it should be fairly decent below the filter cutoff.
Other things to check - self resonance on the inductors? Or capacitive coupling around the filter. The S21 shows the rolloff, but then bounces back up. That's sort of suspicious. It almost looks like not all the sections are active (or that there's some other resonances, like in an Elliptic, you see a bounce back up)
toggle quoted message
Show quoted text
-----Original Message----- From: < [email protected]> Sent: Mar 12, 2025 4:11 PM To: < [email protected]> Subject: [nanovna-users] Help with LPF Measurement I've searched through the message list looking for symptoms similar to mine and didn't find anything. I'm somewhat new to this and would appreciate any help you can offer. I've been trying to get my NanoVNA-F (firmware 1.0.5) to show the results of a 40-meter low pass filter I've built. I know I'm doing something wrong here but can't figure it out. The VNA image is calibrated (including pass-thru), set for the range 6.5 to 29 MHz and logmag S21. Shouldn't this give a flat area leading up to the design frequency followed by a significant roll off through the remainder of the frequency range? My setup is very similar to W2AEW's YouTube demonstration of the NanoVNA with a LPF but gives entirely different results. The image is entirely different, with very large insertion loss and a steep area of suppression peaking at 11 MHz followed by a steep recovery. I worried that my filter wasn't what I thought it was so I tested it with a 7 MHz square wave, looking at input and output with the FFT function of my 'scope. The before and after images show several things: (1) the square wave (yellow trace) is hardly a square wave. It wasn't much of one before attaching to the input of the filter but it's definitely loaded and distorted by the filter. (I can live with the distortion in the test, since the filter is supposed to reduce the high freq distortion anyway, right?) (2) The purple trace is the filter output--pretty nice sine wave. (3) The FFT shows a good 40+ db suppression of the harmonics, especially the odd numbers. So the filter seems to be doing its job. There may be some things not right about my testing of the filter using signal generator and oscilloscope -- I'm new at this. For example, the shoulder on the filter output at about 10 MHz is troublesome, but still the logmag S21 output shouldn't be affected, should it? Any help in understanding why my results look as they do would be greatly appreciated. Best & 73, Mitch NK3H
|
Here's a plot of the expected response from Elsie
toggle quoted message
Show quoted text
-----Original Message----- From: < [email protected]> Sent: Mar 12, 2025 4:11 PM To: < [email protected]> Subject: [nanovna-users] Help with LPF Measurement I've searched through the message list looking for symptoms similar to mine and didn't find anything. I'm somewhat new to this and would appreciate any help you can offer. I've been trying to get my NanoVNA-F (firmware 1.0.5) to show the results of a 40-meter low pass filter I've built. I know I'm doing something wrong here but can't figure it out. The VNA image is calibrated (including pass-thru), set for the range 6.5 to 29 MHz and logmag S21. Shouldn't this give a flat area leading up to the design frequency followed by a significant roll off through the remainder of the frequency range? My setup is very similar to W2AEW's YouTube demonstration of the NanoVNA with a LPF but gives entirely different results. The image is entirely different, with very large insertion loss and a steep area of suppression peaking at 11 MHz followed by a steep recovery. I worried that my filter wasn't what I thought it was so I tested it with a 7 MHz square wave, looking at input and output with the FFT function of my 'scope. The before and after images show several things: (1) the square wave (yellow trace) is hardly a square wave. It wasn't much of one before attaching to the input of the filter but it's definitely loaded and distorted by the filter. (I can live with the distortion in the test, since the filter is supposed to reduce the high freq distortion anyway, right?) (2) The purple trace is the filter output--pretty nice sine wave. (3) The FFT shows a good 40+ db suppression of the harmonics, especially the odd numbers. So the filter seems to be doing its job. There may be some things not right about my testing of the filter using signal generator and oscilloscope -- I'm new at this. For example, the shoulder on the filter output at about 10 MHz is troublesome, but still the logmag S21 output shouldn't be affected, should it? Any help in understanding why my results look as they do would be greatly appreciated. Best & 73, Mitch NK3H
|
On Wed, Mar 12, 2025 at 05:54 PM, John Gord wrote:
It looks like the ground (copper plane) of your filter is not connected to the
grounds (outer conductors) of the cables.
That would make a big difference.
EXACTLY... As I was walking away from the screen shot and getting ready to pull the plug! Again... NEATNESS COUNTS in this process.
|
I see two yellow wires for input and output but NO ground lead from the copper board to each of the coax connectors
|
I don't see a connection of the ground to the copper clad.
But more than that, the leads are pretty long. A rule of thumb is that 1 cm is 10 nH, so it looks like you've got 40-50 nH in series with those capacitors.
So what you really have is a 40 nH inductor in series with a 270 pF capacitor.
At 20 MHz, the impedance of the capacitor is about 0.25 ohms (i.e. it shorts the input to ground), but it's in series with a 5 ohm impedance of the inductor, so it's like the capacitor isn't even there.
You need the leads *really* short - even if you got them, say, 5 mm long, instead of 2cm each, you've still got 10 nH in series, which is 1.25 ohms.
This is why you see chip caps and if you're using disks, they're on a pcb and really trimmed short.
Also, why do you have TWO capacitors on the ends, 270 pF is a standard value. And having two in parallel makes it even more tricky (the inductances would be less, maybe) but...
toggle quoted message
Show quoted text
-----Original Message----- From: < [email protected]> Sent: Mar 12, 2025 5:39 PM To: < [email protected]> Subject: Re: [nanovna-users] Help with LPF Measurement Hi Dave, S21 is the gray, I think (the not yellow one -- I'm color blind) print at the top right of the screen in the last two images I posted. It's line is flat at 0 dB across the frequency range with the test rig shorted and gives the unusual curve I showed when the filter is used. Connecting the Ports 0 & 1 of the VNA together (without the test rig) also show 0 dB flat across the frequency range. Is that what you meant? Many thanks. Mitch NK3H
|
Well, there's a long history of dead bug construction, and 3D sculptures. But it's the short leads thing that's biting you here. You can solder the capacitor lead to the copperclad really close and use the capacitor as a sort of standoff sticking up. Then solder the lead to the inductor to the other cap lead really close. Extra length in the inductor isn't so bad. And, be careful about the inductor windings being close to the copper clad - the capacitance from the winding to the copper clad is essentially shunting. (The overall effect would be to lower the cutoff frequency of the filter).
This is actually pretty useful for you - the NanoVNA said "the filter's not working right", and as you change things, you'll see it improve. Which is a whole lot better than a sort of handwavey - lead lengths are important. You have a chance to quantitatively see how long a lead makes a difference.
BTW, people designing and building these filters often take the original design, build it up, find out it doesn't quite work, and then adjust the values of the components until it does. Maybe they change the 270 pF to 330pF (to account for the inductance in the leads). This is a lot of the "art" of building RF circuits with real, non-ideal components.
I think we've all been burned by something like this (resistors that are inductors is a notorious one)
toggle quoted message
Show quoted text
-----Original Message----- From: < [email protected]> Sent: Mar 12, 2025 6:10 PM To: < [email protected]> Subject: Re: [nanovna-users] Help with LPF Measurement On Wed, Mar 12, 2025 at 05:54 PM, John Gord wrote:
It looks like the ground (copper plane) of your filter is not connected to the
grounds (outer conductors) of the cables.
That would make a big difference.
EXACTLY... As I was walking away from the screen shot and getting ready to pull the plug! Again... NEATNESS COUNTS in this process.
|
The only measurement noted at the top of the screen was S11. Was S21 also there? I didn't see it. Dave <> Virus-free.www.avg.com <> <#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2> On Thu, Mar 13, 2025 at 12:39?AM Mitch NK3H via groups.io <mitch= [email protected]> wrote: Hi Dave,
S21 is the gray, I think (the not yellow one -- I'm color blind) print at
the top right of the screen in the last two images I posted. It's line is
flat at 0 dB across the frequency range with the test rig shorted and gives
the unusual curve I showed when the filter is used. Connecting the Ports 0
& 1 of the VNA together (without the test rig) also show 0 dB flat across
the frequency range. Is that what you meant?
Many thanks.
Mitch NK3H
-- *Dave - W?LEV* -- Dave - W?LEV
|
