To get a fully (mostly) analog signal out of the Si5351 the signal needs to be filtered.
regulatory limits and calculations
I plan to use my transmitter mostly in the 50 MHz, 70 MHz and 144 MHz ham radio bands.
Radio regulations say (at least in Germany) that any signal other than the main signal (out-of-band emission) needs to be 60 dB lower or below 0,25 µW (-36 dBm), whatever is higher. I don't have a tool to measure the transmission power (which will be a few micro watts) I decided to go with the - 60 dB to be save.
In this blog post I will only analyze the 50 MHz signal. The calculation for the other bands are similar. The 50 MHz signal is the fundamental signal or first harmonic.
The Si5351 generates a square wave so the only out-of-bands signals are its harmonics. Harmonics are always multiples of the fundamental signal, so we only have to look at signals with a higher frequency. All even harmonics are 0. The first out-of-band signal is the third harmonic at 150 MHz.
With a square signal the harmonics have a lower amplitude than the fundamental signal and with higher frequency the amplitude gets even lower. So the critical signal is the third harmonic.
The third harmonic has only one third of the amplitude of the fundamental signal, so this signal is already 9,5 dB lower. The other 50 dB reduction have to be made by an external circuit to get a complete attenuation of 60 dB.
filter design
I decided to go with a seventh order chebyshev filter. There exist online calculators for this filter:
https://leleivre.com/rf_chebyshev_LPF.html
http://www.calculatoredge.com/electronics/ch%20pi%20low%20pass.htm
My parameters are:
cutoff frequency: 75 MHz
passband ripple: 1.0 dB
impedance: 50 Ohm
I chose the capacitors and inductors which are closest to the ones calculated.
As this is a seventh order filter the attenuation should be around 7 * 20 dB / decade = 140 dB / decade. So for 150 MHz this should be around 40 dB. (maybe not enough).
measurements
To verify the filter design I tried to measure it.
You need a network analyzer to test the filter. These devices are usually very expensive. But with the NanoVNA (https://nanorfe.com/nanovna-v2.html ) exists a cheap amateur device. I control the NanoVNA with a notebook running Ubuntu and the software NanoVNA-Saver: https://github.com/NanoVNA-Saver/nanovna-saver
The following plot was also made with this software.
I reworked my PCB and placed a SMA connector instead of the Si5351.
The NanoVNA will generate a signal with port 0 and measure it on port 1. The filter is connected to the NanoVNA like shown in the following picture.
The following plot shows the attenuation of the filter.
The first marker (red) is at 50 MHz, the second (green) at 75 MHz and the third (blue) at 150 MHz.
The plot shows that the filter somehow works as expected. There is a lot of ripple around the cutoff frequency which is typical for chebyshev filters. Unfortunately the fundamental signal at 50 MHz has also an attenuation of around -4 dB. The third harmonic at 150 MHz has an attenuation of around -50 dB which is a little bit more than expected.
conclusion
The filter for the 50 MHz band works like expected and the transmission should be within regulatory limits. But there is also room for improvement: The signal at 50 MHz could have less attenuation and the third harmonic at 150 MHz a little bit more as it is close to the limits.