Hello,

This email message is a notification to let you know that the following files have been updated in the Files area of the [email protected] group.

• /Tiny Spectrum Analyzer/Build instructions for the Tiny Spectrum Analyzer.pdf

Uploaded By: erik@...

Description:
A document describing how to build and use the tiny spectrum analyzer

Cheers,
The Groups.io Team

The drive capability of the GPIO outputs is rather limited. When loaded with 50 ohm the output is? between -21dBm and -22dBm. Excellent level for power calibration
--
HBTE Files section:?/g/HBTE/files
Erik, PD0EK

To get some confidence in what you are measuring with the tinySA it would be nice to have a well known signal that you can quickly use as input to check if everything works as expected.
It turns out the LO SI4432 can provide such a signal already for free.
After a reset to SI4432 sets GPIO2 output as a clock output with a frequency of 1MHz.
This frequency can be changes by setting the frequency control field of the LO SI4432 to one of these values:
0: 30 MHz
1: 15 MHz
2: 10 MHz
3: 4 MHz
4: 3 MHz
5: 2 MHz
6: 1 MHz

SA.exe controls the setting of this field with the "R counter" setting. Setting the "R counter" field to any of these values listed above will set the GPIO2 output of the LO SI4432 module to the indicated frequency.
Now I have to measure the output level so we can use this signal also as a power level calibration.

Update tinySA.ino(including display and rotary control) and SA.exe (version?1.0.7308.21301)? are available in the files area
--
HBTE Files section:?/g/HBTE/files
Erik, PD0EK

Hi all,
??? I think that my experience of using screening with the Arduino NANO might be of use:
Working with Tony Fishpool I recently built an attenuator using the parallel non latching version of the PE4302 module. This and an LCD display is controlled by a 5 volt version of the NANO.
??? I found that the attenuator output had 16 MHz? and all its harmonics up to about 800MHz on it! This seen on a Cushman CE-15 Spectrum "monitor", which for my purposes will do as a spectrum analyser.? These were at low level: mostly around -60 to -80dBm, and disappeared into the noise floor of the analyser? at 800MHz.
??? For some applications this would not matter, but for our current purposes it would be? a disaster!
??? The correct approach would have been to screen the Arduino and the display, but in a unit already built it was easier to completely screen the attenuator, which I did.
??? Ater that although the 32MHz Output was still there at about -80dBm, and there was a few outputs in the region of 100MHz at roughly -100dBm everything else was below the noise floor.
??? Of course I realise that other things can come into this - layout, wiring and decoupling for instance, and you may be lucky, do be aware that it could happen to you!

?? By the way;
???? I have not started on my analyser yet. I am still waiting for deliveries from China. I have just received the filters from Australia though, and intend to do the tests under discussion on another thread. If my results are different I will report them there.
Tony (G3PTD)

A few minor clarifications and addendums to my previous post:

>? So the bare B3550 SAW filter without a matching network
>? should have a resistive impedance of 31.59 ohms.
>? In series with that resistance is 65.87 ohms of capacitive reactance.

Looking into one of the two ports of a B3550 SAW filter element,
we would encounter an impedance of 31.59 - 65.87j
We might say that's the impedance from one of the B3550 pins to the center of the B3550.
And it is only true if the far side is properly terminated, perhaps by a 24.2 nH coil
(which is +65.87j ohms at 434mhz) and a 31.59 ohm resistor in series to ground.
If the far side is not properly terminated, all bets are off about the impedance we will see.

Eric's tinySA features a 250mhz LPF driven from a 50 ohm source, going into a mixer,
going into the B3550 SAW filters, going into the Si4432.
The only points where impedance is well defined are (hopefully) the signal source
and the Si4432 input port, which should both be close to 50 ohms.
If one of those two impedances is off (or the parts between aren't exactly right)
all the impedances from front to back are wrong, and the filters and mixers
don't work as expected.

Another issue is that the impedance of the 250mhz LC LPF and the SAW filters will vary
by frequency.? For example, a 500 mhz signal into the LPF will?see a very high impedance
instead of the expected 50 ohms, and the voltages of this undesired signal at the input
might be relatively high.? High enough to sneak through the filter, or perhaps capacitively
couple?into nearby circuits

For both of the above reasons, a good design often has a bunch of 6dB resistive pads
to keep things at 50 ohms.? And then some gain stages to make up for the loss.
But if the tinySA can get good performance while staying simple, I'm all for it.

When writing a complex number, python insists that the "j" follows the numeric digits
so it can easily figure out it is still a number:? ? 31.59 - 65.87j
More typical of electrical engineering texts is to put the "j" in front:? 31.59 - j65.87
The math majors use "i" instead of "j":? 31.59 - i65.87
since they don't have "i" reserved as a variable to represent current.

Jerry, KE7ER
?

For those curious about how impedance matching works,
here's the algebra for what is going on when putting two SAW filters in series.
I believe this is correct, but would be interesting to verify it through the measurement of real parts.

Eric is using the B3550 SAW filter, which recommends using a 33nH coil at each end,
plus a 5.6pf cap from the far end of each coil to ground.?
This matching network converts the filter impedance to 50 ohms resistive.

Impedances in series and parallel work exactly the same as resistance, except we must use "complex arithmetic"
when doing the calculations:??
? ??

The freely available python interpreter knows about complex numbers, using "j" to represent the sqrt(-1):
For a specific frequency of operation, we can find the impedance of a coil or cap:??
? ??

Our frequency is 434 Mhz:? ?freq = 434e6? ? ? ? ? ? ? # Using scientific notation, the e6 means add 6 zeros
The impedance of our 33 nH coil is given by? ?Zcoil = 1j * 2 * 3.1416 * freq * 33e-9
And the impedance of our 5.6 pF cap is:? ? ? ? Zcap = 1/(1j * 2 * 3.1416 * freq * 5.6e-12)

Using the same rules for impedance as we do for resistance,
the series combination of the filter plus the coil is:? Zfilter + Zcoil
That impedance in parallel with Zcap is:? ? (Zfilter+Zcoil)*Zcap / ((Zfilter + Zcoil) + Zcap)? =? 50 ohms
(The B3550 datasheet says that a matching network of a 33 uH coil and 5.6 pF cap will?
make the B3550 SAW filter look like a pure resistance of 50 ohms.)

We can solve that equation for Zfilter using some very basic algebra:
? (Zfilter+Zcoil)*Zcap / ((Zfilter + Zcoil) + Zcap)? =? 50
? Zfilter*Zcap + Zcoil*Zcap = 50*Zfilter + 50*Zcoil + 50*Zcap
? Zfilter * (Zcap - 50) = 50*Zcoil + 50 *Zcap - Zcoil*Zcap
? Zfilter =? (50*Zcoil + 50 *Zcap - Zcoil*Zcap) / (Zcap - 50)

Putting it all together as a python script:

freq = 434e6? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? # Our operating frequency in Hz
Zcoil = 1j * 2 * 3.1416 * freq * 33e-9? ? ? ? ? ? ? ? ? ? # The complex impedance of the coil in ohms
Zcap = 1/(1j * 2 * 3.1416 * freq * 5.6e-12)? ? ? ? ? ?# The complex impedance of the cap in ohms
Zfilter =? (50*Zcoil + 50 *Zcap - Zcoil*Zcap) / (Zcap - 50)? ? ? ? ? ? ? ?# Compute the SAW filter's impedance
print(Zfilter)
??
That script gives me this result:

(31.59 - 65.87j)

So the bare B3550 SAW filter without a matching network
should have a resistive impedance of 31.59 ohms.
In series with that resistance is 65.87 ohms of capacitive reactance.

We could cancel out the 65.87 ohms of capacitve reactance by?
by adding 65.87 ohms of inductive reactance.
(Note that the capacitive reactance is -65.87j ohms, our inductance will be +65.97j ohms,
when added together the sum is zero ohms).

The impedance of a coil is? Zcoil = 1j * 2 * 3.1416 * freq * Henrys
so to get an impedance of +65.97j ohms:
65.97j = 1j * 2 * 3.1416 * 434e6 * Henrys
Henrys = 65.97 / (2 * 3.1416 * 434e6) = 24.2e-9

So if we add a 24.2 nH coil in series with one end of a B3550 SAW filter,
it should then present an impedance of around 31.59 ohms, purely resistive.


In the case of two B3550 SAW filters in series, we can have a matching network
of a single 48.4 nH coil in between the two SAW filters, and no capacitors.
Conceptually, this is two 24.2 nH coils in series, and in between those two coils
we see a pure resistance of 31.59 ohms when looking in either direction.??
The capacitive reactance of both SAW filters is canceled out by the added inductive reactance.
?
Jerry, KE7ER

Them mixers create a whole career's worth of complications to figure out.

My understanding is that a diode ring mixer has no specific impedance,
it simply passes the impedance it sees at the incoming port on to the output port.

I'd try adding a 6dB pad between the front end low pass filter and the mixer.
Also a 6dB pad between the local oscillator and the LO port of the mixer.
Once both of those are done, I'd expect the SAW filter to be getting driven
correctly from a clean 50 ohms resistive.

And once the SAW filter is behaving correctly, perhaps we can figure out how to
pick up the lost gain.? Take a look at W7ZOI's design for a spectrum analyzer:
? ?
He's got 6dB pads pretty much everywhere to avoid impedance mismatches.
He's got an MMIC amp immediately after the first mixer to pick up the gain again.
Which I don't quite understand, but am confident W7ZOI did.
I assume it comes down to managing the gain distribution for maximum dynamic range.

If the tinySA works well enough on $10 worth of parts, maybe that's good enough.
But I am curious if an extra dollar could make it better.
And a flat passband does strike me as better.

Jerry, KE7ER

After some more testing I discovered my nanoVNA was not correctly calibrated when measuring the SI4432 input impedance.
The SI4432 input impedance turned out to be fairly ok at 434MHz, about 45ohm.
The culprit was the mixer. After adding some isolation between the mixer (10dB attenuator or an amplifier) the pass band became flat again.
Do not put too much amplification between mixer and 434MHz filter as this will ruin your 1dB compression point and make it impossible to measure low levels of distortion as you lose your dynamic range
I tried 15dB and that was way too much.?
But not having the flat passband really did not make a difference in tinySA measurements so for now I will leave it as is.
--
HBTE Files section:?/g/HBTE/files
Erik, PD0EK

Joerg,

Thanks for this excellent input.
I also added a LCL (33nH, 12pF, 33nH) network between the two SAW filters


and the S11/S21 went from

to


Roughly 5dB S21 gain which is always nice!

Unfortunately, when measuring the S21 when inside the tinySA (you can do this by setting the sweep type to "First IF" i.s.o. "Regular") you get this:

The bad out of band rejection is direct leakage between the two SI4432 modules and can be ignored but as you can see the top of the passband is again not flat.
When measuring the input impedance of the RX SI4432 it measured far from 50ohm so there is another improvement opportunity.?

--
HBTE Files section:?/g/HBTE/files
Erik, PD0EK

Two capacitive ports in series makes for a really capacitive connection there.
These SAW filters are not designed for that, as post 666 has shown.
They need a conjugate match, which means adding inductance to each port
to cancel out the capacitance.?

A bit of physical separation between the two SAW filters might be a good thing,
keeps out of band signals from jumping across.?
Shielding one SAW filter from the other could also help.

Jerry Gaffke, KE7ER

开云体育

I've been wondering about noise from the Arduino as well. I'm working on a Kicad pcb layout and thought the Arduino pro mini 3.3V, 8MHz version would work best since it doesn't have a USB connector which can be a source of noise. I plan on using sockets for everything including the PE4302 attenuator module. The SAW filters I just ordered from Germany so I don't know when they'll arrive but I would like to make a module out of those too. I guess there will need to be some shielding around the Si4432 modules and Arduino and plenty of bypass capacitors on power lines. Since I've never built a SA it's all experimental for me. As for the SAW filters, the coupling between them is probably ok as the are in series so whatever impedance they are will be matched. Eric seemed to imply that the matching inductors and capacitors are not that important in his build article. However, if you take the modular construction approach, you and try out different variations to see which works best.

Jerry Ponko

AC9NM

Yes.
Good job!
That passband graph looks great!

To be clear, it looks like each of the individual SAW filters has?
the recommended impedance transformation network?
on each side of the SAW filter.
So a total of four 33uH inductors and four 6.8pF capacitors,
as recommended in the B3555 datasheet.
The datasheet for Eric's B3550 recommends the same 33uH inductors,
but changes the capacitors to be 5.6pF.

Some SAW filters require a much different impedance transformation network.

Having the two SAW filters and the inductors physically separated like that?
(or perhaps shielded from each other) will improve the stopband
by preventing unwanted signals from sneaking around your filter.

It would be ok for the two middle 33uH inductors to see each other,
in fact that could be one big coil.
But the inductors on the ends should not couple to any other inductors.

Jerry, KE7ER

I finally added a MAV-11 MMIC between the filters:


With this I reached an overall gain in the passband of 5.3 dB:


BTW: The peek at 437.8 MHz is pretty exactly, what is shown in the datasheet.
--
Joerg, ex-DB2OO

I noticed the same extreme ripple in the passband and the high loss, when I DIRECTLY coupled the SAW filters (EPCOS B3555 from the German "expensive" supplier on eBay, that I received 3 weeks ago):


S21 measured with NanoVNA:


Then I split the filters and added in between the 50 Ohm transformation network made up of the 33nH (4 turns, 3 mm diameter) and 6.8pF:


This resulted in a much lower passband loss (5.5 dB) and a smooth passband:



--
Joerg, ex-DB2OO

That's an extremely good price!
Especially if they work well enough.
Here's a datasheet, note the peak at 438mhz:
? ??
They look pretty much identical to the B3550's that Eric is using,
including termination impedance, passband, and the peak at 438mhz.

Jerry

Hi Jerry,
These are the filters I ordered....



Erik. Thanks for confirming those results matched with yours.

Kind regards
Tony

Tony,

Are these the same B3550 filters that Eric was using?
Here's a datasheet, the B3550 is in the first section:
? ??
Page 7 of that datasheet shows a graph of the passband,?
with a major artifact near 438mhz.

Ripple and attenuation could be explained by improper timination.
Page 6 of that datasheet shows that a B3550 should have a series 33nH inductor
on each end, and then a 5.6pF cap to ground beyond that.
This will convert whatever's going on inside the B3550 to an impedance of 50 ohms resistive.
To couple two B3550's together, I believe we should have a total of four such inductors and caps,
so that the node where they join is also 50 ohms.? ?(Or one could use a bit of algebra to figure
out the capacitive impedance looking into a B3550 port, and then get by with a single inductor between?
the two SAW filters.)

If you don't have a datasheet for the SAW filters you got so cheaply (or if they are factory seconds),
it may take some work to determine how to properly terminate the filter.
I suggest you stick with the two networks you have, but add inductance between the two SAW filters.
Start with 33+33 = 66nH, and then try other values above and below, see if that cures the
attenuation and ripple.

After having looked hard at the B3550 datasheet, compare it with this one:
? ??

Be aware that the center pin of the SMA connectors may want to turn as you tighten
down the plug at the end of a cable to it.? This will break any SMD components
soldered directly to that pin.? ?Hard to tell from the photo, but you may have a large
enough copper island for the SMA pin to prevent this.? If it becomes a problem,
perhaps use a wire from SMA pin to nearby parts, the wire can twist.

Jerry, KE7ER

Tony,

Good to see these results.
My SAW filters have exactly the same ripple and loss. It's a pity but my tinySA does not seem to bother much
Out of curiosity you could try to measure a single saw filter and see if (times 2) the results are the same. If not then the direct connection between the two filter is an issue and a series inductor (or even an LCL network) could be considered
As you seem to be able to mill away the copper rather accurate you may consider milling the rather simple footprint of the saw filters and solder them as they should be (not dead bug).
Or you can consider a copper foil at both sides of the SAW filter to create a massive ground plan
W.r.t the? measurement with the nanoVNA for the out of band rejection:
Can I suggest you calibrate the nanoVNA S21 with an amplifier (say 20dB) and an attenuator(say -20dB) between port 1 and port 2?.
When you measure the filter S21 you than remove the attenuator but keep the amplifier between the output of the SAW filter and port 2 and you have 20dB more dynamic range
This is possible because he SA612 mixers have their signal levels set for the fundamental but you measure a 10dB lower 3rd harmonic through a filter with 10dB loss (as you measure above 300MHz)?

--
HBTE Files section:?/g/HBTE/files
Erik, PD0EK

Niels,

That may work as is.
Or it may need a few iterations of the layout to work well at these frequencies.

A few things I'd be scratching my head about:

How solid is the ground plane???
Can the receive path from SMA to Si4432 be optimized for a short and straight path?
from one end to the other to minimize coupling between elements
(perhaps using coax for the LO from the second Si4432 to the mixer)?
Should the longer high frequency traces be fatter to better approach the 50 ohm system impedance?
Will noise from the Arduino overwhelm what you are trying to see?
Can the circuit elements be broken up to allow for shields to be added?
Should there be a series inductor added between the two SAW filters?

Jerry

I'm still waiting for some of the other modules to arrive. However the filters arrived from the Polish EBay source (that has been previously mentioned) so I set about building that module first. It would be of interest to compare test results with others in the group. The attenuation is more than I would have expected - plus there is something happening over on the HF side at 437.5 and 438.

Some photos are attached showing construction and tests using a VNWA 3 and a NanoVNA. As you can see, the NanoVNA didn't do a half bad job.

I'm seeing some ripple and the impedance isn't constant (or even 50 ohms) across the section of interest. It's not massively way off and I wonder what results you've seen?

I suppose most of the cost is postage. The Polish chap sells them in 50's and I probably won't live long enough to find a use for the 48 left.

This offer is only going to be usful in the UK but I'm happy to supply (until they are gone) two filters for free. All you need to do is mail me an addressed envelope with a UK standard 2nd class postage stamp. If you send an A5 sized envelope I can include a free back issue of the GQRP Sprat magazine (if required).
The filters and a Sprat will glide in under 100 grams.

EU or DX postage is much more expensive and you may as well order them from the suppliers direct.

There will not be a paypal option and I would appreciate a direct email (and not via the group) so that I can let you know if there are any left before you send postage.

Kind Regards
Tony G4WIF