Softrock Lite 6.2
Adventures in Electronics and Radio
Elecraft K2 and K3 Transceivers
Review of Si570
Frequency Synthesizer Kit from K5JHF and K5BCQ
11 July 2009. Original
12 July 2009. Expanded with more measurement data
13 July 2009. Added power output versus frequency
14 July 2009. Revised discussion related to calibration and added frequency
Table of Contents
John, K5JHF, and Kees, K5BCQ, have developed an integrated
manual controller and display for the Si570. John is responsible for the
microcontroller firmware and logical hardware design, while Kees is responsible
for physical hardware design, board layout, and producing the kit. This page is
my review of the kit and assembly process. You may also wish to view my page
Canned Osc Phase Noise for more information on
the Si570's phase noise.
The kit price, including a CMOS Si570 integrated circuit,
and all electronic parts (including the LCD display and the tuning knob) is $40,
For more information on the kit see Kees's web page
My overall conclusions:
- Exceptional bargain for $40.
- Si570 is an amazing chip, with extremely high quality
(low phase noise) output
- The major area of improvement needed is in the
documentation. As it currently stands, this is not a kit for the
- Assembly time: 90 minutes.
- For more conclusions, click
My assembled Si570 kit
What is an Si570 and why do
I want one?
Good question. The Si570 is
described by its manufacturer, Silicon Labs, as an "ANY-RATE I2C PROGRAMMABLE
XO/VCXO." What this means in normal amateur radio language is that the Si570 is
a frequency synthesizer, complete on one very small (7mm x 5mm) integrated circuit chip.
(That's the any-rate, programmable part of the name.) The XO/VCXO part means
that the synthesizer is crystal controlled (XO = crystal oscillator) and that
one variant of the chip accepts an analog input voltage that will shift the
programmed frequency (voltage controlled crystal oscillator. That version is not
used in the kit.) The final perhaps unfamiliar term in the part description is
"I2C" which refers to the method by which frequency commands and other data is
communicated between the rest of the circuit and the Si570. I2C is a specific
serial communications protocol using separate clock and data ports and for the
purpose of this review, it isn't necessary to delve further into the specifics
The Si570's intended purpose is in microcontroller and
computer systems, where it provides very accurate and stable clock pulses with a
frequency capable of very rapid changes. Since computer chips expect digital
waveforms, the Si570's output is a square wave with a fast rise and fall.
For our amateur radio activities, the Si570 has a rather
different use�it can be used as the local oscillator in a superheterodyne
receiver, or it can be the local oscillator in a direct conversion quadrature
receiver such as the Softrock. The primary purpose for which John and Kees
developed this kit is the Softrock, I believe. A number of other adaptations of
the Si570 to the Softrock receivers and transceivers have been made, including
integration with the sound card decoding software to yield a complete multi-band
receiver but will not be covered in this review.
In addition to its use inside a receiver, the Si570 kit
can be used as a stand-alone signal source for test and measurement and for
other purposes, such as a VFO for your old Heathkit DX40 for that matter. Just
keep in mind that the Si570's output is a square wave and may require additional
filtering for some purposes.
A detailed datasheet for the Si570 can be downloaded at
selecting the Si570/571 in the "Resources" tab. I've extracted the summary
description and block diagram below.
As mentioned, the Si570 comes in several variants. The lowest price version is
compatible with CMOS logic and has the following main characteristics. The CMOS
variant has a "single ended" output, i.e., one output pin that swings between 0
and 1 logic levels. Other versions have a differential output, i.e., two output
pins, with inverted signals; when one pin is high, the other is low and vice
versa. The differential output version has a much higher maximum frequency range
and has slightly better jitter performance compared to the CMOS chip.
The chip supplied with the kit is the 3.3V version, but an
on-board voltage regulator allows the kit to work with supply voltages ranging
from 5V to 13V.
My particular CMOS Si570 has a lower frequency limit slightly
below 3.5 MHz and a maximum frequency limit above 250 MHz, comfortably exceeding
the published specifications of 10 and 160 MHz.
The kit includes a
microcontroller, an MC9S08QG8 device, and an outboard non-volatile memory chip,
a 24AA32, as well as an LCD and a rotary encoder. The microcontroller reads the
user's turns and pushes from the rotary encoder and translates those into
I2C commands to the Si570 (and, for that matter, to the 24AA32 as well,
since it is also an I2C device.)
To enter a frequency (1 Hz steps) push the encoder in and
hold it in and rotate. This moves a decimal point indicator left or right,
indicating the digit to be modified. When the desired digit is identified,
release the encoder and turn it to either increase or decrease the frequency
value. The readout is in Hz, without a decimal point. (The decimal point is only
used to identify digits for modification.) You may save the frequency into
non-volatile memory (retained when power is removed) with 1,000 memory
locations. You may also program in a division ratio to be applied to all
entries, useful with the Softrock receivers where there is a divide by 4 or
divide by 8 factor between the oscillator and the reception frequency. The
division ratio changes the display so that a if a division ratio of 4 is
entered, setting the frequency to 10 MHz causes the Si570 to output a
frequency of 40 MHz. There is also an "offset" frequency factor, which adds or
subtracts a fixed offset to the display, useful if you use the Si570 as the
local oscillator in a receiver.
Other settable options include the number of increment
steps per shaft rotation of the rotary encoder and a direction
normal/reverse option, useful in receivers with an inverting frequency scheme.
Inputs and Outputs
The kit's input and output ports are:
- DC power. Power supplied may be between 5 and
14V at approximately 80 mA.
- RF Output. Depending on options installed during the
build, the RF output will be either single ended referenced to ground or
balanced output. My kit has an output power of approximately +14 dBm, or 25
mW, corresponding to 1.1V RMS.
- Filter Bandswitching. A separate port contains
frequency information suitable for controlling switchable bandpass filters
when the Si570 is used with a multiband Softrock receiver.
- Battery Power. Direct access to the +3.3V internal
power bus for battery operation of the kit. Since this port bypasses the
voltage regulator, it is possible to damage the integrated circuits by
- LCD. The LCD connects to the board through a 10-pin
All connectors are 0.1 inch pin header style, single row,
except for the LCD, a 2x5 0.1 inch header.
My kit included the following options:
- CMOS Si570
- Minicircuits TC1-1T transformer
The kit is supplied with a 6-page "Construction Hints"
document and a schematic. Neither a parts list nor a detailed "Heathkit" style
instruction set is provided.
Most of the parts are surface mount, with the passive
components (resistors and capacitors) being 0805 size. These components can be
soldered without too much effort providing the proper tools are used. A
temperature controlled soldering station with a small tip is essential, as is
the correct solder size. I use a Hakko 936 with the small handpiece and a 2.5 mm
tip. For surface mount soldering, I like Radio Shack's 0.015 inch diameter, 2%
silver solder. Radio Shack has a lot of schlock but their 2% silver containing
solder is a remarkably good product and is available in small quantities so you
don't need to spend $40 for a full pound.
The documentation does not include board photographs, so I
took a series of images as I built the kit.
First, the top and bottom of the board, before assembly.
The board is approximately 1.25 inches x 1.5 inches, and in some areas
components are closely spaced.
The bottom is not silk screened and the drawing provided in the Hints Sheet is a
mirror image of the actual board, which can be confusing at best. To help other
builders, I've prepared an annotated copy of the PCB bottom, shown below. (Note
that U3, the Si570 has eight pads, not six. The two additional pins are on the
ends of the pad layout.)
Assembling the Top Layer
Before starting construction, I believe the builder should
first decide how the output options are to be configured and mark the
corresponding paragraph in the builder's notes with a highlighter. The output
option area is not silk screened and it's easy to get confused by momentarily
looking at a paragraph describing a different output option. My selection is:
- Output transformer (you should order this option
when you purchase the kit)
- No output termination
- Transformer installed for 1:1 (center tap
floats, and is installed with the center tap on the output side.)
In the absence of specific instructions on assembly order,
my approach to building a PCB is to work in the order of smallest components
first. Accordingly, on the top layer, my assembly order was:
- U2, 24AA32
- MCP 1703, the 3.3V regulator. (No U number in
- U4 jumper between pins 2 and 3 (this may or may not
be needed in your kit. With a single ended CMOS chip and the output
transformer, U4 is not used and a jumper between pads 2 and 3 is needed.
- C9 0u1 capacitor
- R2 2K ohm
- R3 2K ohm
- R5 (in my configuration, a 0 ohm 0805 jumper)
- C8 0u1 capacitor
- R4 is not used in my configuration
- R7 is not used in my configuration
- R8 is not used in my configuration
- C7 (in my configuration, a 0 ohm 0805 jumper)
- T1 (I mounted the transformer so that the center tap
is on the output, thereby providing 1:1 ratio, as the center tap pad is not
connected in this orientation.)
- R1 (the through-hole resistor. This is needed only if
you plan to operate with a DC supply above 11 volts. Otherwise a jumper wire
View of PCB top surface after installation of all top mounted
components except connectors and rotary encoder. The output section (upper right
corner of the board) wiring will
differ depending on the options selected.
Since the Builder's Notes provide a mirror-image view of
the PCB's bottom, I first made a reversed image so the layout matches the actual
PCB surface. (The mirror image is because the documentation image is extracted
from the so-called "X-ray" view of PCB layout software.) To help future
builders, I've provided a correctly oriented photo of the board with parts
My assembly order of the bottom
- C1 1u0 capacitor
- C2 1u0 capacitor
- C3 1u0 capacitor
- C4 0u1 capacitor
- C5 0u1 capacitor
- C6 0u1 capacitor
- U3, the Si570. Note that the Si570 has 8 pins. Three
on either side with large pads and one on each end with small pads.
The photo below shows the board bottom at this stage
of assembly. (The Si570's two end pads are not soldered yet in this photo.)
The next step is to install the 16 pin DIP socket and the 5x2
header strip to connect with the LCD. Kees recommends soldering the LCD directly
to the header pins, although the kit includes a 5x2 socket. I decided to use the
socket because I will likely mount the LCD separately from the board and because
it facilitates trouble shooting and experimentation. If you plan to mount the
controller card and LCD as a unit, you may wish to make the direct connection.
The 5x2 LCD header strip mounts from the top board
surface. The pins should not extend through the pads beyond the minimum
amount necessary to permit soldering. This is because a DIP 16 socket
straddles the LCD header.
When installing the 5x2 header pins, remember that these come up in the middle
of the 16 pin DIP socket and hence should protrude through the PCB only to the
extent necessary to permit soldering.
Bottom view of microcontroller socket. Note that pin 1 is
oriented to the outside of the PCB.
The remaining step is to install the rotary encoder on the top board surface.
One tab fits through the large hole. The other tab must be
bent horizontally and soldered to the ground pad near the board's outer edge as
As the last steps, I installed 3-place header
sockets on the board for DC input and RF output. The kit does not include these
parts, and instead provides header pins for the purpose. The builder is
instructed to install the pins and bend them 90 degrees to the board. I did not
install connectors for the band decoder nor the 3.3V DC input as I don't plan to
use these features.
I also installed the provided 5x2 header socket on the
LCD. The completed board is shown below.
The controller and Si570 board mounts to the end of the LCD with the 5x2 header
pin / socket.
Post Build Checkout
There's no formal post-build checkout process documented, so
I did the normal things. Resistance check of power supply to ground, check for
shorts on the Si570 output circuit and the like.
When powering the kit up, I used an HP E3610A power supply
with current limiting set to 100 mA, with the thought that any major fault would
trip the current limiting before damage occurred.
I then applied power and found the LCD functioned and I
could change frequency, save frequencies, etc., but the output frequency was
frozen at the default initialization value of 56.320 MHz. To make a long story
short, I found that one of the Si570's two end pads were not well soldered. I
touched up both pads and went through the controller initialization process
described in the Builder's Notes, and all worked as it should. I changed the
default starting frequency to 10.000 MHz.
The image below shows memory position 20 at 10.000000 MHz.
Stability and Accuracy
I looked at the
frequency stability from startup and found the Si570 quite stable. After a brief
warm-up change of about 40 Hz, the frequency stabilized within less than 1 Hz.
This is not a particularly stressful test, as the temperature in my basement
workshop didn't change more than a couple degrees during this time, but it shows
that in a reasonable environment, the output is quite stable. The absolute error
from the commanded 10.000000 MHz is about 48 Hz.
After completing the kit, I wrote in this review that it was
not possible to calibrate the Si570's frequency. Thanks to Alan, G4ZFQ, John,
K5JHF and Kees, K5BCQ, I've learned that is not the case.
The controller's firmware stores the Si570's nominal
startup frequency in memory location 0. As explained by Kees, this is 56.320000
MHz because the Si570 parts sold with the kit have a default of 56.320000Mhz.
This frequency was selected because it yields a good center frequency when
divided by the 20 meter Softrock's divide by for chain�14.080MHz. (If obtained
from a different source, the Si570 might have a different default frequency.
Kees goes on to say "if not specified, SiLabs generally supplies 10.000.000MHz
default parts. This is not a problem at all for the kit, but location "0" has to
be changed to the default frequency of the Si570 part you are using."
After applying the calibration, my frequency error was
reduced from around -4.5 PPM to 0.1 PPM. (Still subject to some thermal drift
and probably long term aging as well, of course.)
The calibration process is:
Memory location 0 holds the default "true" power up frequency. Since I obtained
my SI570 as part of the kit, the nominal value is 56 320 000 Hz matches
the nominal startup of my Si570 oscillator. (I'll add spaces for readability but
the display does not have spaces.)
The value stored in memory location 0 should be replaced
by the actual measured start up frequency. This measured value will then be
applied by the controller as a correction factor to all frequency calculations.
The correction factor is calculated as part of the controller's power-on
initialization, so changed values will not take effect until the controller
power is cycled off and on.
Step by step instructions are:
1. Allow the Si570 to warm up, as well as whatever frequency measuring equipment
you plan to use.
2. Select memory position 0. You should see 56 320 000 and the Si570's output
frequency should be close to this value
3. Enter the actual measured value into memory position 0. In my case is was 56
319 742 Hz.
4. Save this value to memory by pressing in the rotary encoder knob until saved.
5. Turn the power off and back on. (The controller only reads memory position
0's value upon power on initialization.)
The before and after results are quite impressive:
100 MHz was 99 999 530.0 and is now 100 000 000.6 Hz
13 MHz was 12 999 942.0 and is now 13 000 002.0 Hz
10 MHz was 9 999 955.67 and is now 10 000 001.00 Hz
As seen in the stability plot below, of course, the output
frequency moves around bit with temperature variation.
After calibrating the Si570, I again looked at stability,
this time for a longer period of 7.5 hours. As the plot below reveals, the
frequency calibration did nothing for the start-up transient (and of course it
should not have done anything) but after a 15 to 20 minutes for stabilization,
the output frequency stays quite stable, with a relatively small frequency
I started the run early this morning while my basement
shop was cool and the air conditioning was not running. By noon, the outside
temperature warmed up to 85°F and the air conditioning was cycling on and off,
and the shop temperate varied three or four degrees during the air conditioning
cycle. This shows up around 250 minutes into the run, where the oscillator
temperature cycles over 1.5 Hz or so and tracks temperature changes in the shop.
As a rough estimate, therefore, the Si570's temperature
coefficient looks to be around 0.5 Hz/°F at 10 MHz. Converting to parts per
million and the Celsius scale, the temperature coefficient is around 0.1 PPM/°C.
The Si570 shipped with the kit has a temperature stability
rating of ±50 PPM over the range -40/+85°C. On a per degree basis, assuming the
full plus and minus range is used, the temperature coefficient is 0.8 PPM/°C,
considerably more pessimistic than observed. I didn't measure my shop's
temperature swing, and I could easily be over-estimating the peak temperature
excursion. It's also likely that the Si570's temperature coefficient is
non-linear, with greater stability in the normal operating range. Crystals are
usually cut with the region of maximum stability in the target temperature range
and one would expect this practice to be followed in the crystal embedded within
the Si570. Finally, it's possible�and indeed probable�that the Si570 has an
on-board temperature sensor and computes and applies a temperature based
correction rather than placing all the thermal stability burden upon the
To be useful in a high performance receiver, the Si570 must
have low phase noise. My first measurement is based on John Miles's PN program
that collects phase noise data with one of a variety of spectrum analyzers
HPIB control. The image below shows the phase noise measurements taken with the PN program and my Advantest R3463 spectrum analyzer.
To see whether the measurements are limited by the R3463,
particularly for close-in measurements, I looked at three signal generators, one
crystal oscillator and the Si570.
The figure below is from the February 1973 Hewlett Packard
Journal feature article on the 8640 signal generators. As the figure indicates,
my R3463 spectrum analyzer falls woefully short of being able to measure the
phase noise of a quiet source, by perhaps 30 dB or more. (As seen at my page
Canned Osc Phase Noise some inexpensive
"synthesizers in a can" are so bad that the R3463 shows phase noise elevation
well above the instrument threshold.)
I've used several alternative arrangements to look at phase
noise, including a crystal notch filter and mixing two oscillators down to 50
KHz and viewing the output with an HP 3562A Dynamic Signal Analyzer.
The figure below shows the result of mixing a 10.050 MHz
signal from one of two oscillators with a high quality 10.000 HP oven
oscillator. Phase noise will appear as broadening of the trace and higher noise
level at frequencies further from the carrier. Of course, this image represents
the composite noise of the oscillator under test and the crystal oscillator.
Compared with the 8640B, the Si570 compares very favorably
close in. (The span is 1 KHz in this image.) Note that the Si570 trace is
narrower than that of the 8640B close in to the carrier and that its noise is
several dB below the 8640B at you move away from the carrier.
Looked at over the full 100 KHz span of the HP 3562A Dynamic
Network Analyzer, the Si570 has a small increase in noise over the 8640B close
to the carrier, but the difference is small. Further from the carrier, the two
noise oscillators have essentially identical noise floors. (It's not clear that
the noise is the limit of the 3562A or is real phase noise from the two
oscillators under test.) The drop in noise below 20 KHz is related to the low
frequency response of the mixer.
By comparison to
the 8640B, therefore, we can say that the Si570 is quite a capable performer.
(The 8640B has for many years been the "gold standard" for low phase noise
signal generators, although some newer synthesized instruments are at least as
The Si570's output is a square wave, with fast rise/fall
times. It is accordingly rich in harmonics, with odd order harmonics
predominating. (A perfect square wave has only odd order harmonics.) The two
spectrum analyzer images below show the Si570's output with a maximum frequency
of 1 GHz with the Si570 set for an output frequency of 100
MHz. (In fact, significant harmonic energy is seen above 2 GHz.) Quite impressive.
I also looked at the Si570's output with a Tektronix TDS-430
400 MHz bandwidth digital oscilloscope, and measured the rise and fall time,
along with the duty cycle. I used the TDS-430's 50 ohm input option to reduce
cable ringing. The oscilloscope's computed rise and fall times, along with the
duty cycle and peak-to-peak voltage are shown at the right side of the display,
computed by the TDS-430's automated measurement feature.
The Si570's rated rise and fall time is 1.0 ns, 80%-20% while
the TDS-430 performs the more traditional 90%-10% measurement. In addition, the
TDS430's internal 90-10 rise time of 850 ps must be considered as the
oscilloscope can only measure and display the composite rise time of the input
signal and the instrument.
Looking at an expanded view of the waveform's rising edge,
and following a bit of work with a ruler and pencil on a printed copy, I measure
the 80-20 rise time as 800 ps. I don't have a figure for the TDS430's internal
rise time on a 80-20 basis, but it's clear that the Si570 meets its 1 ns
I mentioned that it's possible to obtain a sine waveform from
the Si570's square wave by passing it through a low pass filter. The image below
shows the Si570's output waveform (Si570 set for 10 MHz output) after it is
passed through a 5th order 14 MHz low pass filter. As expected, with the higher
order harmonics being filtered, the output is a good quality sine wave.
Looking at the same information, but in the frequency domain
via a spectrum analyzer, we see the 3rd harmonic is reduced from approximately 9
dB below the fundamental to approximately 50 dB down, with all higher order
harmonics in the spectrum analyzer's noise level.
Si570 Output without low pass filter
Si570 Output After Low Pass Filter
Output Power Change With
Jeff, AC0C has asked how the
Si570's output holds up with frequency. The plot below shows the output of my
Si570 kit from 4 to 260 MHz.
The data is taken with the Si570 kit output connected
through 6 feet of RG-316 Teflon coaxial cable to an HP 8482A power sensor, read
by an HP 437B digital power meter. Since the Si570 is not designed as a 50 ohm
output device, I suspect the wiggles in the measured data are related to
mismatches. I have not corrected the data for loss in the RG-316 cable, which is
0.5 dB at 100 MHz and 0.8 dB at 250 MHz.
The linear regression line shows a good fit to a decrease
in output power of 0.023 dB/MHz.
The Si570 is a remarkable performing signal source and
this kit provides an inexpensive way to build and operate the Si570.
As good as the kit is, it could be improved in a few
- This is not a kit for beginners, in part due to the
documentation. Better documentation would allow less skilled kit builders to
complete the kit.
- Bottom silk screening would help as well. (In the
interest of full disclosure, my Z10000 kit does not have bottom silk
screening either. My later kits with bottom mounted components have bottom
silk screen, however.)
- If the builder decides to separate the controller
board from the LCD display, there's no good way to mount the controller
board. One possibility I would like to see is a rotary encoder with a
threaded shaft. Or, a pair of mounting holes in the controller board,
although there's not a lot of extra room on the current PCB.
- It's not always easy to determine whether you are
changing MHz or 10's of MHz or 100's of KHz. The decimal point is presently
used as a cursor to indicate the changed digit, thus requiring the user to
count zeros to determine which digit is actually being changed. If the LCD
supports it, a better arrangement would be useful, such as using the decimal
point as indicating MHz and flashing the digits being changed on and off.
- It would be nice to have an RS-232 interface to the
microcontroller, with a few simple commands such as set frequency, read
frequency, save to memory, recall from memory. The interface could be logic
level with a full bipolar RS232 daughter board as an accessory. I see no
uncommitted pins in the current microcontroller, however.