That is a very interesting idea but I think the frequency error would
be much larger than what we get with pre-calibrated output.
Take this page for example:
http://jorisvr.nl/arduino_frequency.html
Mentions an error of +105 ppm over 24 hours with a 16mhz quartz. I am
assuming it was in a room where the temperature did not change much.
The maximum error with factory calibration is specd to be lower than
61.5ppm over the full range of temperature of the chip.
How do the pic based frequency counter works? They must have some
pre-scaler and a pretty stable clock?
thomas
On Tue, Feb 11, 2014 at 1:13 AM, Dave <dave@xxxxxxxxxx
<mailto:dave@xxxxxxxxxx>> wrote:
Steve:
It should be possible to write a small routine(microprocessor
pin would be needed) to do the frequency counting function.
Accuracy would depend on the accuracy of the 16 MHz crystal. That
could make the calibration automatic only requiring a button push
to initiate. The program could program a frequency that is low
enough for the micro to count it(say 10 MHz), compare the measured
to commanded and then calculate the Si570 crystal frequency.
Dave - WB6DHW
<http://wb6dhw.com>
On 2/11/2014 12:07 AM, Steve VK2SJA wrote:
Some random thoughts:-
Two calibration methods exist.
1. Manually enter the start-up frequency. Software uses this
value along
with additional values retrieved from the Si570 chip to
calculate the
actual reference crystal value.
2. Instruct Si570 to output known frequency, say 10Mhz and
adjust up and
down the software crystal reference value until the output on
a frequency
counter is exactly 10Mhz.
Both methods have some Pros and Cons.
Method 1:-
You don't need a Frequency Counter
You need to know your Si570 start-up frequency (might prove
awkward
finding this out without a counter - negates not needing one)
Accuracy is good. But Tomas's own results seem to show that
+/- 300 Hz
might not be an unreasonable expectation.
Method 2:-
You DO require a frequency counter of some sort.
Accuracy will only be as good as the counter.
You don't need to know the Si570 start-up frequency.
It is probably reasonable to expect errors of only +/- 3 Hz
after calibration
Not every person owns a Frequency Counter. I will even allow
that it is
possible that someone building a Minima may not own one.
So I would actually suggest having both methods available via
a Firmware
set-up routine. Since most of the extra code would deal with
changing a
value via the VFO pot. And both methods require this. A
significant
portion of the code should be able to be shared between both
methods.
I'm thinking something along the lines of:-
Press and hold Function button while powering up Minima to
enter setup
routine. Use VFO knob to cycle through setup options. Press
enter to
select the one you want. One option can be to exit setup.
Something like
this:-
Setup
Calibrate 1 -> Calibrate 2 -> Other Options we haven't thought
of yet ->
Quit -> Back to Calibrate 1
When you enter either Calibrate 1 or 2 the user interaction is
much the
same. With one you are adjusting the crystal reference
frequency up and
down until you get exactly 10Mhz on the counter. With the
other you are
adjusting the value of the start-up frequency up or down until
it matches
your known Si570 start-up frequency. In both cases you press
the Function
button to accept and save the new value. Which returns to the
top level
setup menu.
There is even a logical progression in doing both calibration
routines in
order 1 and 2. While entirely non-essential Calibrate 2 could
use the
calculated value produced by Calibrate 1 as its starting
point. Which
would put it very close to the correct value.
Anyway just some random thoughts out loud.
73, Steve.
Thomas,
Just to be clear, what I am suggesting is that the calibration
procedure would initially present on the display a 'common'
start-up frequency for the Si570. As an example, the CMOS
Si570 that is sold by KM5H
http://km5h.softrockradio.org/index.php?route=product/product&product_id=49
has a start-up frequency of 56.32 MHz. If it is agreed
upon that
that start-up frequency is commonly used, then that would
be the
frequency value initially presented to the operator. If the
Si570 actually being used has a different start-up frequency,
then the operator can use the potentiometer to change the
initial,
value to the actual value. For example, the CMOS Si570 that I
have has an initial start-up frequency of 10 MHz, so I would
adjust the frequency setting displayed accordingly prior to
pushing the function button to accept the value displayed. I
hope this makes sense to you.
73,
Mac AE5PH
On 02/10/2014 12:24 PM, Thomas Sarlandie wrote:
Dan,
Latest version of my code will always be available in
my github
repository:
https://github.com/sarfata/radiono
The code I was mentioning yesterday has already been
merged by Farhan
and so you will get all my changes in the official
version which is
available here:
https://github.com/afarhan/radiono
The rest of the discussion is about what is the best
way to calibrate:
1 - Using factory calibration settings in the chip -
but we need a
way for the user to find out this frequency and give
it to the program
(maybe through a menu like what Mac suggested)
2 - Starting with no calibration and a fixed output
frequency
(10Mhz). Ask the user to adjust until it is perfectly
10mhz.
Both method will play nicely with my changes. Nothing
was lost ;)
best,
thomas
On Mon, Feb 10, 2014 at 9:50 AM, Dan Reynolds
<on30ng@xxxxxxxxx <mailto:on30ng@xxxxxxxxx>
<mailto:on30ng@xxxxxxxxx <mailto:on30ng@xxxxxxxxx>>>wrote:
Thomas, I would still like to see your code if
possible.
73
Dan -- KB9JLO
https://sites.google.com/site/on30ng/
On Sun, Feb 9, 2014 at 8:27 PM, Thomas Sarlandie
<thomas@xxxxxxxxxxxxx
<mailto:thomas@xxxxxxxxxxxxx>
<mailto:thomas@xxxxxxxxxxxxx
<mailto:thomas@xxxxxxxxxxxxx>>>wrote:
Hi everyone!
Spent some time this week-end on the Radiono
software. I found
out earlier this week that my Si570 had a
very large error
that grew linearly with frequency. We quickly
found out that
each module is different and you need to use
calibration
information that is saved in the module. The
current source
code has Farhan's si570 calibration hardcoded
in. That will
most likely not work for you.
I have updated the Radiono source code to
automatically get
the calibration information and calculate the
crystal
frequency every time we start the program.
The only
information that needs to be defined in the
source code is the
default frequency programmed by the factory.
In my case it was
56.320 but your mileage may vary.
There are two ways to find this frequency:
- Spec: Look on the marking on your Si570
(mine says
CAC000141) and lookup some tables online to
find which
frequency corresponds to your marking (I cant
find this table,
if someone has it, please share!)
- Experimental: just remove the Arduino and
apply power to
the control board. Measure the frequency at
the output of the
Si570.
I have updated the wiki:
http://www.hfsignals.org/index.php/Si570
The updated source code has been submitted to
Farhan in a
pull-request:
https://github.com/afarhan/radiono/pull/4
Full list of changes:
Split the Si570 code in a separate class.
In the Si570 class:
* Use the factory-calibrated output frequency
to automatically
calculate the internal crystal frequency and
use that instead
of hard coded values.
* During initialization detect errors talking
to the Si570.
* Use multi-bytes i2c reads and writes when
possible (ie:
read/write multiple bytes without re-starting
a transaction).
Only works for read/writes in consecutive
registers.
* Use types with explicit size where that
makes sense (uint8_t
instead of char, uint64_t, etc)
* Added some debugging on the serial console
through a "debug"
function
In the main sketch:
* Show a message on the LCD if the Si570 is
unreachable.
* Initialize the serial port for debugging
* Fix some compilation warnings with string
litterals being
casted to char*
* Removed unused variables
And now, back to the soldering iron!
73,
thomas
