My version of the ZPU compiler genrates more optimal code than Sergey's probably
because it's a purely C compiler vs a combined C/C++ one Sergey has. So this
version of production ZPU image has 1829 bytes available for stack, while
Sergey's had only 1764 bytes. Given we're not sure how much stack do we really
need, it's better to have more. My bootloader has 1507 bytes avialable for
stack, while Sergey's has only 1442.
This is a fixup commit for the previous commit (8D7B9D84).
ZPU is a stack machine where all code, constants, variables and function stack
are stored in a single memory block. Compiler doesn't through any warnings
or errors if the allocated memory is exhausted, so we have to check this
manually.
Our memory size is only 16383 bytes and we already had situation when all of it
is exhausted leading to unpredictable behaviour. Unfortunately we can't know
how much memory exactly do we need, but we estimated that 1300 bytes is enough
for us now, so this is what we're checking.
It'c not critical for production use, because it onlu affect time to lock
on startup or on manual frequency correction, but it makes lock faster
when in testing environment, e.g. in out post-manufacturing testing.
We will then disable forced TCXO DAC setting by UHD host library. This way we
avoid de-calibration on every UHD restart in case of operational GPSDO.
Setting GPSDO LPF to the current calculated frequency after setting the DAC
maually should help GPSDO to converge faster. Right now it starts really slow,
because it starts from a perfect target value.
* BUGFIX: Fix GPSDO rounding issue to stabilize frequency down to +-1 Hz @ 52MHz.
* Export measured frequency and last PPS time over the control socket.
* Allow GPSDO debug enable/disable over the control socket.
* Print firmware minor version and git hash on startup.
* Make debug output on wrong packets less scary for users.
Previously we were sending out USRP control ID request, which is wrong, because
we're only interested in UmTRX.
Also lets use our proper compatibility number in the first message to reduce
error output on the UmTRX (it would complain about a compatibility number
mismatch otherwise). And we're now rejecting old compatibility numbers - we
don't maintain backward compatibility.
Users think that these messages indicate some sort of an issue, but
they are part of a normal UHD startup process. So we make them look
rather informational and less scary.
Now we really get to +-1 Hz of the base TCXO frequency precision.
Previously we ignored least significant bits of the frequency due to integer
arithmetics rounding, which meant that we could get up to about 80 ppb of
static frequency error which is not good enough for telco systems.
Added explicit include of #include <boost/cstdint.hpp>
The uhd header wb_iface.hpp has moved from boost cstdint
to regular stdint.h and thus is lacking the boost include.
