开云体育

On 5/20/22 11:20 AM, Jim Allyn - N7JA wrote:

On Fri, May 20, 2022 at 10:57 AM, Jim Lux wrote:

If you have Windows, download Elsie from Tonne Software.

Windows is not necessary. Elsie runs in Linux in WINE just fine! I think there is something similar for Mac, too, so it will probably run there, too.

I've run it with Parallels (but that's just running Windows in a VM on Mac). And with the new Macs based on a non x86 CPU, I don't know if one will be able to do that.

On Fri, May 20, 2022 at 10:57 AM, Jim Lux wrote:

If you have Windows, download Elsie from Tonne Software.

Windows is not necessary. Elsie runs in Linux in WINE just fine! I think there is something similar for Mac, too, so it will probably run there, too.

On 5/20/22 9:52 AM, Gerard wrote:

Hello

So i don't speeak English, i use reverso
Thank you for your answers.
I am at the beginning to break my head on the wall to understood how that work


I really have a hard time setting up and understanding how this filter works.
Before talking about 4 pole filter with a cascade of 2 2*poles filters,, I will simplify to a 2 pole filter all made (Type motorola 20j26 on Ebay or 45R15AZ on Alie Express (I have a lot of each))
If I test this filter alone with a generator and an oscilloscope we see that we have a maximum signal at for example 44.998 and 45.006 Mhz
If we move to 45mhz, the signal is much weaker.

That is exactly as expected - what you have is two filter sections, one tuned to 44.998, the other 45.006. The combination of the two is fairly broad, but there's a dip in between.

At one center frequency, there's no attenuation from one filter, and some from the other.

At the other center frequency, the same is true.

At the center, you're seeing both filters attenuate, so you get a dip.

If you bring the center frequencies together, there's a point where they just even out. If you bring them farther apart, you get two distinct peaks.

In all cases, the filter rejection "far out" is basically twice what a single filter can do.

If your design needs, for example, 100 dB rejection "out of band", and your filters can do 50 dB (far away from center), then you can stack 2 of them. But say your filters are 5 kHz 3dB bandwidth. If you directly cascade them at the same frequency, the -3dB point will be narrower - you'll be -6dB at the bandwidth (+/- 2.5 kHz).

So what you do is tune them a bit apart.? Then as you sweep, one filter is coming up to it's peak while the other is falling off.

If I place the filter in the assembly of my Ubitx homebrew, and I make a measurement I have the same behavior by putting the generator in input of the 1st amplifier (behind the mixer) and looking at the oscilloscope in output of the filter
(Clocks are disabled because plug-in module)
I wonder if the measure is right?
Actually, I’m gonna go back to zero to figure it out.
I need a test model diagram.
I am not an RF engineer and I really do not see how to have a "flat" part between for example 44993 and 45.007 which will make about 14 Khz of bandwidth.

Getting it "flat" is really tricky - and might not even be possible with 2 filters. That is, if you set the filters far enough apart that you get the bandwidth, you get a dip in the middle, because the peak is sharp, compared to the desired bandwidth.

This is the challenge with filter design in general - as the desired percentage bandwidth increases, while keeping low loss in band, and good rejection out of band, you need more and more sections.

If you have Windows, download Elsie from Tonne Software. The student version is free and you can experiment with the number of sections, and see what happens with Butterworth, Chebyshev, Cauer/Elliptical in a sort of interactive way.

Also, despite my module tested at the nanovna which had a nice curve, (see previus post) I also have the same behavior.
So we’re back to zero?
No simple for me.

cdt

Hello

So i don't speeak English, i use reverso
Thank you for your answers.
I am at the beginning to break my head on the wall to understood how that work


I really have a hard time setting up and understanding how this filter works.
Before talking about 4 pole filter with a cascade of 2 2*poles filters,, I will simplify to a 2 pole filter all made (Type motorola 20j26 on Ebay or 45R15AZ on Alie Express (I have a lot of each))
If I test this filter alone with a generator and an oscilloscope we see that we have a maximum signal at for example 44.998 and 45.006 Mhz
If we move to 45mhz, the signal is much weaker.
If I place the filter in the assembly of my Ubitx homebrew, and I make a measurement I have the same behavior by putting the generator in input of the 1st amplifier (behind the mixer) and looking at the oscilloscope in output of the filter
(Clocks are disabled because plug-in module)
I wonder if the measure is right?
Actually, I’m gonna go back to zero to figure it out.
I need a test model diagram.
I am not an RF engineer and I really do not see how to have a "flat" part between for example 44993 and 45.007 which will make about 14 Khz of bandwidth.

Also, despite my module tested at the nanovna which had a nice curve, (see previus post) I also have the same behavior.
So we’re back to zero?
No simple for me.

cdt

On 5/20/22 6:24 AM, alan victor wrote:

Motorola provided 4 pole filters from matched 2 poles. The dot mark on each xtal must face each other. At their junction a proper shunt C to gnd is added. This sets the coupling factor for the 4 pole. The xtals were delivered including the cap setting the coupling. Terminations for the filter pair were usually complex, RC, A COUPLE OF k shunt a few pF. You will need to add a LC network to provide from 50 ohm source and load.
Alan

If anyone wants to fool around with a sort of idealized "two resonators in series" and how you can stagger tune to get a "bandpass" see the attached spreadsheet. There's another sheet with 4 resonators..

You can change resonant frequency and Q (the equation's not perfect, but it's close)

Nor does it take into account things like coupling, etc.

But you can see how you can cascade resonators to get a fairly flat bandpass characteristic, and you can interactively "tune" the stages (oh, just like tuning a cavity filter with a VNA)

Motorola provided 4 pole filters from matched 2 poles. The dot mark on each xtal must face each other. At their junction a proper shunt C to gnd is added. This sets the coupling factor for the 4 pole. The xtals were delivered including the cap setting the coupling. Terminations for the filter pair were usually complex, RC, A COUPLE OF k shunt a few pF. You will need to add a LC network to provide from 50 ohm source and load.
Alan

Saver reads the values from the nano as they are. If the nano is
calibrated, Saver gets calibrated values. If you reset or disable the
calibration on the nano, Saver will get uncalibrated data. Saver has no
way of knowing which it is.

Any calibration done in Saver is done 'on top of' the data from the nano.
So at first glance it would seem that turning off calibration in the nano
would be the right thing to do when calibrating within the Saver app. And
indeed this does work very well, at least for some of us. The catch is
that Saver's software wants to see well-behaved data points, within the
expected data range of -1 to 1 (if I remember correctly). But some nanos,
due to variations in component tolerances in the nano hardware, will
sometimes emit data values a bit outside that range. Save does not deal
with this, and will crash or otherwise misbehave.

So the Saver authors recommend calibrating the nano (usually over a broad
freq range that encompasses your needs) to avoid this issue. You can then
do additional layers of calibration in Saver, including the very useful
feature of multiple segments to multiply the number of points.

But if you try it on your nano and it works correctly for your scenario,
using Saver with the nano's calibration turned off is a good (better?) way
to go. It is then calibrating raw data, and there is no concern about
compatibility of the frequency range used to calibrate the nano vs that
used in Saver.

Jim,

filters are 45R15AZ from Aliexpress, but no datasheet found. Sold as equivalent to 45M15A?
Nothing specified on couplage value.
cdt

On 5/19/22 7:51 AM, Gerard wrote:

re,

I’m back on track. I wonder, should we not focus too much on the gain curve, but rather improve the loss curve?
see attachement
cdt

The S11 (return loss)?

Most filters have a terrible return loss outside their passband.

What you need to ask is "what's driving this" - will the reflected power out of band cause a problem? (i.e. make an amplifier unstable) - generally not.

And, because S22 is likely similar, will the next stage care about the input having a poor match out of band (i.e. become unstable).

If you're running into/out of a mixer, the usual thing is to put some resistive attenuation (or a circulator if you're at microwaves) - a 3 dB pad means that no matter how bad the filter is, the mixer only sees -6dB at worst.

The fact that the match varies "in band" is already taken into account in the S21 measurement - it's normalized to a ideal 50 ohm source, so the fact that the filter reflects 10% of the power (-10dB) (transmitting 90% into the filter) at some frequencies and 1% at others (-20dB, transmitting 99%) is accounted for in the S21 measurement.

10 dB reflected corresponds to -0.5 dB transmitted.


A device that reflected -10dB at some points and -20 dB at other points, when looked at in the transmitted sense (leaving aside any absorption internally) would look like varying between -0.5 and -0.004 dB gain.? Real filters have some loss inside the filter, so the gain is typically a bit flatter than you'd expect just from the S11 and S22.

On 5/19/22 7:33 AM, Gerard wrote:

Hello,

I come back to you because I need information on the interpretation of the nanovna curves.
So I wanted to make a filter 4 poles from 2 filters 2 poles.
By simulating on a model the 45MHZ filter of the Ubitx I managed to have a relatively flat gain curve.
On the other hand, when I test with a generator, whether on the model or filter inserted in the Ubitx, I have a fluctuation of the signal.
To sum up I have a stronger signal at 44.992 or 45.005 than at 45.000Mhz.

That's not unusual - at 5dB/div it looks pretty flat. How flat do you need it? 0.1 dB is challenging.

Recall that a multi pole narrow band bandpass filter is made by staggering the peaks of the resonators making up the filter. Imagine cascading a bunch of classic (1-(w/w0)^2) sections. There's no way to get perfectly flat by combining lumps, you can get it flatter by adding lumps, but the lumps are still there.

Compare this to wideband bandpass filters, which are a low pass and a high pass that are cascaded. And you can get a very flat passband by using a multisection Butterworth characteristic.

I have the impression, in fact, that we follow the "loss" curve more than the gain curve during the measurement.
Therefore, I wonder if the measure is well done.???

Looks pretty good.? Reasonably flat on S21, and the S11 shows a decent match within the passband and a terrible match outside the passband, which is typical for run of the mill filters.

What also surprises me is that in principle, the 2 filters 2 poles being connected together, the central leg (output 1 to input 2)must have a capacitor coupling to the ground.

Is that in the mfr data sheet?

If I put in a variable capacitor the gain curve is horrible regardless of the position, I actually put an impedance in parralelle sweat the capacitor and there only I have a correct gain curve, but the capacitor must be at 0.

so is it actually necessary with the best loss curve at the detrimant of the gain curve or other
cdt

re,

I’m back on track. I wonder, should we not focus too much on the gain curve, but rather improve the loss curve?
see attachement
cdt

Hello,

I come back to you because I need information on the interpretation of the nanovna curves.
So I wanted to make a filter 4 poles from 2 filters 2 poles.
By simulating on a model the 45MHZ filter of the Ubitx I managed to have a relatively flat gain curve.
On the other hand, when I test with a generator, whether on the model or filter inserted in the Ubitx, I have a fluctuation of the signal.
To sum up I have a stronger signal at 44.992 or 45.005 than at 45.000Mhz.
I have the impression, in fact, that we follow the "loss" curve more than the gain curve during the measurement.
Therefore, I wonder if the measure is well done.???
What also surprises me is that in principle, the 2 filters 2 poles being connected together, the central leg (output 1 to input 2)must have a capacitor coupling to the ground.
If I put in a variable capacitor the gain curve is horrible regardless of the position, I actually put an impedance in parralelle sweat the capacitor and there only I have a correct gain curve, but the capacitor must be at 0.

so is it actually necessary with the best loss curve at the detrimant of the gain curve or other
cdt

I've had the same question for some time, but never bothered to ask. I've always just calibrated directly on the VNA because of what NanoSaver says.

Dave, can you please elaborate? Or anyone else? Why does the software have the waring/suggestion that it's calibrations be done directly on the NanoVNA?

yes, once you have calibrated in SAVER, you can store it on the laptop or
PC for future use.

Dave - W0LEV

--
Dave - W?LEV

Hello,

I've been using my Nano VNA as stand alone for a while and have recently started using the Nano VNA saver software for windows. I really like the software and find it very useful.

I've been using it with calibration made on the actual VNA itself, but have noticed that I can do it via the software also, which I thought might be worth doing. I didn't do it as I note the software suggests that it's actually better not to and rather calibrate via the VNA itself.

Does anyone have any input on this? I had hoped to create several calibration sweeps and save them via the software to make things easier to recall, plus being able to recall many scenarios.

Thanks,

Thanks Jim,

Yeah, I was wondering about calibration. I've seen some conflicting info online. Here's what I did.

Removed UUT.
For Isolation, I connected 50 ohm standards to both CH0 and CH1.
For Through, I connected CH0 directly to CH1.

Is that the right way?

73,

Phil

On 5/17/22 11:17 PM, Phil via groups.io wrote:

It's looking to me like my NanoVNA-H has an 87 ohm input impedance when I try to do a through measurement. Shouldn't it be 50?

Here's how I came to that conclusion:

I connected a 4.99k ohm resistor from CH0 to CH1,
Then I configured the NanoVNA as follows:
Display Format LOGMAG
Channel CH1 THROUGH
Then I did a calibration.
The display shows -35.2 dB.
The attenuation would be 10^(35.2/20) = 0.0173
Which means the NanoVNA's input impedance is 88.33(1-0.0173) = 87.5 ohms.

As a confirmation, I replaced the resistor with a 0.01 uF capacitor, and it was 3 dB down at 187 kHz, which is what I would expect for an input impedance of 87.5 ohms.

Am I doing something wrong here?

You should cal using the cal standards (i.e. a zero ohm thru), then make the measurement.

If you're doing the cal with the UUT in the loop, who knows what you'll get.


If you want to measure the Ch0 Z, you'd need to look at the actual cal coefficients (not the displayed values) calculated with a standard cal - I don't know that those are accessible. Or, reset the cal to none, measure some standards, and from that you can calculate the port's apparent impedance.


Or, hook up another NanoVNA, and directly measure it.

1/have you a copy of your soft ware 5.1.0
you need to look at bottom for plastic stuch under cove
this caught me out

It's looking to me like my NanoVNA-H has an 87 ohm input impedance when I try to do a through measurement. Shouldn't it be 50?

Here's how I came to that conclusion:

I connected a 4.99k ohm resistor from CH0 to CH1,
Then I configured the NanoVNA as follows:
Display Format LOGMAG
Channel CH1 THROUGH
Then I did a calibration.
The display shows -35.2 dB.
The attenuation would be 10^(35.2/20) = 0.0173
Which means the NanoVNA's input impedance is 88.33(1-0.0173) = 87.5 ohms.

As a confirmation, I replaced the resistor with a 0.01 uF capacitor, and it was 3 dB down at 187 kHz, which is what I would expect for an input impedance of 87.5 ohms.

Am I doing something wrong here?

73,
Phil

I had a similar issue with an H4. A firmware upgrade fixed it, and improved
the menu access, as well as the softkeys for setting frequencies.