IC9700 frequency lock by Glen English VK1XX
Update blog, Main article, picture , contact
8 May 2019
I have taken a look at the locking method in the IC7610. It's essentially the same thing we are doing-
they have a oddball frequency oscillator for all the clocks (actually it is not oddball, it's a audio sample rate multiple)
and they lock it to a decent internal or external 10 meg reference ...
The approach they need to take is DIFFERENT from the IC7610. Which leads me to think Icom dont actually have experience with doing it this way.
Anyway, we'll see if Icom get this sorted. They might implement a frequency counter in the FPGA (which is easy ), the counter is driven from their 49.152 derived clocks.
At some period, perhaps a 10 times per second , the frequency count is sent to the microprocessor, and the micro runs some filtering and a PID loop, and drives the VCXO as required.
Getting enough error information at 10 times a second on a small error means a rolling window of counts which essentially produces a high sample rate of errors that can be averaged.
The high rates of update are required to be able to follow the drift when the fan starts and stops.
...and Icom if you do a HW rev, you need to isolate the copper pours under the oscillator so it doesnt follow the bottom and top layer copper temperature....
6 May 2019
Rex VK7MO sent me an IC9700 to frequency lock, and I describe it below. It was a textbook case. Easy.
I probed around without schematic for a few hours, figured out how the radio works, measure all the levels , sensitivities etc.
Icom have done an excellent job of this radio. I mean this radio is phenonomal value for money - I think it is cheap at twice the price.
Things to watch : Running wires around the insides is a bit of a no no.
Wires are antennas and this means potentional unwanted signal injection into very sensitive clock and control nodes.
The radio is sensitive to aliasing up to a couple of GHz so almost any stray signals can be a problem.
Because of those sensitivities, this radio needs a bit more care than other radios when retrofitting other circuitry to it.
For those interested, the VCO Kv is 194Hz/V at 49.152 MHz. You'll need that to properly calculate the loop filter components.
This radio does not need the constant fan mod- the loop filter with a BW on 10Hz and TC of around 2mS can easily track the thermal drift due to the fan.
The radio's sensitivity to microphonics is reduced (since the PCB resonances are within the control bandwidth, mostly.)
This was an adapted board, I am using an ADF4157. Chosen for its high resolution divider.
Next version will have a analog switch on it so that when no 10 MHz ext is detected it can switch back to the internal reference control circuitry .
. Also the PCB will be shaped to minimise wiring distances and have a few more ground connections in better places.
It's a difficult job with all the wires going to 0402 parts on the PCB.
There needs to be source resistors in series to isolate those nodes which I have, but it is a micro-surgery operation.
I will eliminate that need on the next version.
I calculated 0.3Hz error at 1296 MHz and I measure 0.4Hz error +/- 0.1Hz. Not bad.
Yes I do this for a living ... I design custom SDRs and frequency stability is part of the course.
In my opinion (not humble) , the internal oscillator stability is not adequate for digital modes. It is barely good enough for SSB.
Icom you did a great job on 99.99% of this radio, why the fail here ? A solution may be able to be implemented in your FPGA and microcontroller.
The 10 MHz goes into the FPGA, there are dual DACs driven from the microcontroller, everything is there for this to be done inside the radio...
Sure the radio meets spec +/- 0.5ppm, but the drift rate v temp makes it unsuitable for narrow band modes in the higher bands.
The next PCB might include an option for a cheap ovenized 10 MHz oscillator on board so that the internal oscillator is ALWAYS locked to a internal decent oscillator and external reference.
Yes it is a prototype !. Lots of Kapton tape.
The board itself is approx 18mm x 40mm double sided 1mm FR4.
I control the 49.152 MHz oscillator directly , and take 24.576 MHz from one of the digital dividers that feeds the '9518.
I take power from the 8V sub regulator, and a tap of the 10 MHz input.
Radio will go back to Rex today, I will think about the mod a bit more.
...- -.- .---- -..- -..-
@ cortexrf dot com dot au