VK3XDK PLL – fitting an Si4133D

One thing that became apparent when using a single XDK PLL board in a dual-band 2.4/3.4 GHz transverter system, is that the PLL’s limitation on producing even MHz only, necessitates using different IFs.

For 2403, (1970 LO), a 433 MHz IF is required.
For 3400, (2968 LO), a 432 MHz is required.

In the heat of the moment I can see myself forgetting to change the FT-817’s operating frequency to accommodate this requirement! 😕
To try and make it idiot proof, I decided on replacing the Si4133G PLL chip, with an Si4133D chip.

What’s the difference?

The “D” variant has a programmable Reference divider, whereas the “G’s” R divider is fixed at “65”.
So the “D” is a lot more versatile.


In it’s real application, with a 13 MHz reference, the “G” would produce frequencies in 200 kHz steps.
However, with a 10 MHz reference, it produces odd steps of 153.846 kHz (10000/65).
Now 13 x 153.846 kHz is 2 MHz… which is why we are limited to using “even MHz”, (which translates to “N” counter multiples of 13).

The Si4133D (also known as Si4133-GT, just to confuse us) has a programmable R divider, and becomes more versatile.

Above: Setting R divider to 10, allows 1 MHz steps

Whilst it won’t matter in a solo operator environment, in a multi-operator environment, having a 432 MHz IF is not desirable at all.
We would get IF break though from the 70cm station who is most likely operating on 432.100 MHz… right on our IF for 3400.100, 5760.100 and 10368.100 MHz. 🙁

With the Si4133D (Si4133-GT) we can easily produce an LO of 2967 MHz, so that our 3400 IF is 433 MHz.
In this case, we get the PLL to generate 1483.500 MHz, instead of 1484.000 MHz

Right, time to change the PLL chip!


The Si4133D PLL chips were purchased from X-On Electronics, based in Perth.
If this is the first time you are ordering from X-On, you will need to register an account on their web site.
This is quick and easy.
Searching for the Si4133, we are presented with a few options.
Not clear what the story was here, I telephoned their sales dept, and was told that some have an MOQ of 50pcs (now it becomes obvious!), and the difference with the other two listings comes down to supplier.

Given the $14.70 and $15.29 parts are the same thing, you might as well order the $14.70 part.
With orders under $200, X-On charge a $15 handling and delivery fee, so you might as well purchase several of them at once (or for a group buy), to do all your PLL boards.
They took 8 days to arrive.

Above: Si4133-GT (Si4133D) has just been fitted to the PCB, awaiting the solder short across inductor pins 7 & 8

These SSOP24 packages are small, so unless you have a capable SMD rework station at your disposal, and know how to use it, I would enlist the help of somebody who does!

I’d like to thank Graeme, VK4FI, for fitting the new chips to my PLL boards.
It’s amazing what favors can be done for a bottle of decent Merlot. 😉

Now with an Si4133D we are not limited to even MHz steps anymore.

Above: PLL on 1483.500 MHz, (PD = 500kHz), and after x2 multiplier board, becomes our 2967 MHz LO for 3400

For a 433 MHz IF for 5760, we generate 1775.6667 MHz (PD = 333.333 kHz), instead of 1776 MHz for a 432 IF.
For a 433 MHz IF for 10368 we generate 1655.8333 MHz (PD = 833.3 kHz), instead of 1656 MHz for a 432 IF.

Phase Noise

One other benefit with a “D”, and not being stuck with relatively small 153.846 kHz steps, is a potential improvement with Phase Noise.

Above: Comparison generating 1970 MHz [color=#0000FF]with PD=153.846 kHz (as with a “G”)[/color], and [color=#FF00BF]with a PD=1 MHz (as with a “D”)[/color]

We immediately see a close-in improvement with Phase Noise by around 10dB.
(I cannot explain why I saw a relative rise around 100 kHz, and maybe due to a quirk of this particular spectrum analyzer?)
This plot was generated by the [url=http://www.thegleam.com/ke5fx/gpib/readme.htm]KE5FX GPIB Toolkit[/url], a useful (and free!) collection of windows-based spectrum analyzer applications for GPIB equipped instruments.

For amusement, I also generated plots, again at 1970 MHz, but with PD steps of 100 kHz, and 10 kHz.


As expected, the Phase Noise starts to get rather poor with a small PD frequency of 10 kHz.

The version of PICAXE code I am using for the Si4133D is attached below.
This is an adaptation of the v0.3 code from Dave, VK3HZ.

By pulling IN4 (physical leg 3) high, we enable the IF synthesizer.
The IF synthesizer frequency is in the first EEPROM slot.

The DIP switch select the RF synth’s operating frequency, as “normal”.

I found that the PICAXE 14M does not have enough memory, so three of the EEPROM frequency listings had to go from the table.
So there is only room for 13 frequencies, not 16.
(Like this, there are only 3 spare bytes left in the 14M!)
Realistically, this should be ample for even a 5-band transverter system sharing the same PLL!

The 14M2 chip has much more memory, and would accommodate the full table of 16 frequencies.
(For an M2 part, the I/O pin coding would need changing, as the M2 parts use the PORT.BIT format when referring to pins.)

Seeing as the highest addressable DIP switch position could vary depending on whether you use a 14M or a 14M2, I have used memory address #0 for the programmable IF frequency.
ie; the first slot.

This can remain constant, even if you do upgrade to a 14M2 part, and add the three extra EEPROM frequencies to the end of the table, otherwise you will need to change the table pointer to the IF frequency in line 259.

A selection of useful frequencies for 2.4, 3.4, 5.7 and 10G include LOs for both 433 and 432 MHz, as well as the “dot zero” reference frequencies.

icon_picaxeVK4GHZ Si4133D v5.bas