My review of the IC9700 is in the current issue of Practical Wireless, out today, for those of us without a subscription.
Having now upgraded to V1.11 firmware, in order to improve the reference stability, here are my observations.
By now the IC9700 reference oscillator stability is a well-known problem. In fact, for most users of CW, SSB and FM, the stability is a non-issue. It is only with the narrow digital modes that the stability problem shows. It seems to be associated with poor thermal control around the 49.xMHz reference TCXO.
The cooling fan comes on on transmit only. It does not run continuously. Not only that, but it only comes on when required on transmit. When it does come on it causes a small, but noticeable drift in frequency. Without the fan coming on, drift is much less noticeable.
On 2m it is hardly noticeable, even on FT8. It is more noticeable on 70cm and definitely noticeable on 23cm, although not really a big problem using JT65C on 23cm, compared to many other rigs.
Several ‘fixes’ have been proposed, including causing the fan to run continously, albeit at half speed; full reference locking using an add-on board and the possibility of a firmware upgrade to implement a locking system in the internal FPGA.
As designed, the external reference 10MHz input is little better than a modern version of a crystal calibrator. A facility is provided to automate the calibration of the internal reference against the external reference input. Once calibrated in this way, the internal reference is once again free to ‘drift’, controlled only by the TCXO function.
Version 1.06 firmware improved on the correction algorithm and did make a big difference. The IC9700 is able to achieve a frequency accuracy only dreamed of in rigs like the TS2000x without frequency locking mods.
With firmware v1.10 and now V1.11the reference is periodically corrected, once invoked, and now works much like the system used in, e.g. the K3.
However, the period is possibly a little long as there is still a noticeable shortbterm drift under some conditions. I can see some further work, to refine the process, may yet be forthcoming from ICOM.
So, how well does V1.11 work to stabilise the reference oscillator?
On 2m FT8 I could see no change in frequency between transmit and receive, fan operating or not. Well, maybe a Hz!
FT8 on 70cm barely showed showed any change. Certainly not enough to cause any problems.
As FT8 is not recommended on 23cm, I didn’t bother to check for change with these short, 15 second, transmit/receive periods. However, JT65 uses a 47second transmit period followed by a 1minute, 13 second receive period. This puts a much bigger strain on the thermal characteristics of the reference oscillator and drift, even with the wider spaced tones of JT65C, was previous noticed with drift exceeding 20Hz when coming back to receive after the 47 second transmit period. This is important as any difference in the tone frequency from the expected frequency leads to a small but definite degradation in sensitivity. Keeping the tones where they should be gives the highest sensitivity.
So does V1.11 improve the situation at 23cm.
To test the drift I set up one of my signal generators to produce a very weak reference carrier signal about 2.4kHz above the dial frequency I had chosen for my JT65C tests (into a summy load). This carrier produced a nice, straight, line on Spectran. The span on Spectran was set to show just a few hundred Hz either side of the tone. A drift of 10Hz was quite noticeable at this setting.
WSJT-X was then set to give repeated CQ calls of 47 secs on followed by the receive 1minute 13 second od receive, during which the 2.4kHz tone could be seen and measured. This was done at maximum 23cm output power (100% or 10W).
As the signal generator was GPS displine-locked the frequency of the test carrier was very stable.
On test the tone frequency was always within a few Hz of 2.4KHz when returning to receive. It would then wander up to a peak of about 10Hz HF (2.410kHz) over the first few seconds, returning close to 2.40kHz reasonably quickly. Sometimes overshooting to a few Hz below 2.40kHz.
This happened each tx/rx cycle (over) although the peak excursion didn’t always reach 10Hz and averaged about 5-8Hz.
Although not yet perfect, this is a big improvement over the v1.05, v1.06 and v1.10 firmware.
I would expect a further improvement with a refined version of the firmware.
I am quite happy with this result. It would be interesting to do this level of testing with a few of the other transceivers covering 2,70 and 23cm!
I know from experience that the TS2000x drifted about 50Hz between transmit and receive on 23cm JT65C, before doing the fan-always-on mod.
This suggests that it might be worth still doing this mod to the IC9700 to keep the fan turning all the time, as when run at low power on 23cm (1.5W) the fan rarely comes on and drift was noticeably less in these tests.
More testing to do!