FreeCalypso and CMU200 update

Tue Jul 11 00:30:36 UTC 2017

Hello FreeCalypso community,

It's time for me to give you all an update on what I've been doing
during the last few weeks of silence.  I still desire to produce more
FCDEV3B boards, but I am currently fixated on the need to have a
proper setup for RF calibration.

Background info for those who have joined the mailing list recently:
every newly produced GSM MS (mobile station) device needs to have its
RF tract tested and calibrated in both Rx and Tx directions for each
band, and the standard test equipment for these calibration procedures
is Rohde&Schwarz CMU200.  I bought a CMU200 on ebay back in March
(these instruments are in fairly abundant supply on ebay), and I have
used it successfully to perform some calibration procedures, first
manual, then automated.  However, there are two issues which I need to
resolve before further progress can be made, and the road toward the
resolution of these two issues is taking quite a while.  The two
issues are:

1: The CMU200 can act as an RF signal analyzer or as a signal
   generator, or both at the same time.  Some units, including mine,
   are also equipped with a secondary signal generator in addition to
   the main one.  On my CMU200 the RF analyzer function works fine as
   does the auxiliary Tx generator, but the main signal generator is
   broken: no signal or an extremely weak signal comes out.

There are two further problems with this broken signal generator issue:

1a: I discovered the fault quite late: I bought this CMU200 back in
March, and because the RF analyzer function works, I thought that the
whole unit was totally fine until some time in late May when I reached
the point of needing the signal generator function.  So that's two
months right there from the time of purchase, ticking away against the
seller's 90 day warranty.

1b: When I discovered the Tx problem in late May, I started the
dialogue with the seller about it, and they seemed willing to help me,
but then I needed to prepare for my presentation at REcon Montreal in
mid-June (not only the presentation itself, but also logistical
preparations for the big trip), and I had no time to deal with the
CMU200 issue.  I was only able to get back to this CMU200 about a week
ago (early July), and by now the 90 day warranty is over.

What I've been doing is I've been trying to troubleshoot this broken
CMU200 signal gen problem on my own, i.e., trying to determine which
internal module is at fault and get a replacement for that module.
The CMU200 service manual explains what the major internal modules are
and what they do; the two modules of interest for this troubleshooting
chase are the RxTx board and the RF front end (FE).  The 4 N-type RF
connectors on the front panel of the CMU200 (where DUTs are connected)
are part of the FE, and the FE has internal connections (made with
pieces of rigid coax) to the RxTx board.  In the Tx path the RxTx
board generates the RF signal from an internal IF input that comes
from another board, i.e., upconverts to the desired output frequency,
and the RF output from the RxTx board goes to the FE via a piece of
rigid coax.  The FE switches this Tx output to one of 3 possible
external connectors (of the 4 external RF connectors in total, two are
bidirectional, RF3 is output only and RF4 is input only), and
presumably does some amplification or attenuation as well depending on
which output is selected, as the specified output levels for RF1, RF2
and RF3 are quite different.

In my most recent experiment done only yesterday, I opened the CMU200
box and disconnected the rigid coax piece that carries the Tx output
from the RxTx board to the FE.  I then took the same N-to-SMA cable
which I've been using to connect the CMU200 to FreeCalypso DUTs,
connected the SMA end of this cable to the now-exposed output from the
RxTx board, and connected the N end to front panel RF ports of the
same CMU200 - I tried both the usual RF2 port as well as the more
sensitive input-only RF4.  I powered the unit on in this state,
commanded it to transmit, and used the CMU200's own RF analyzer to
measure the power level being put out - this is the CMU200 function
which worked OK for measuring power levels put out by FreeCalypso DUTs,
so why not.

What I saw were really low signal levels.  Whether I commanded the
unit to transmit on RF1, RF2 or RF3 at the respective maximum power
levels of -27 dBm, -10 dBm and +13 dBm, the signal coming out of the
RxTx board maxed out somewhere around -70 dBm.  Now we are still
crippled by not knowing what the signal level *should* be at this
point, i.e., how much amplification or attenuation happens in the FE
in each mode, but around -70 dBm max seems too low to me.  Hence my
finger of suspicion points at the RxTx board, rather than the FE.

Earlier today I went to ebay, searched for a spare/replacement CMU200
RxTx module, and found someone in Israel who had one of these modules
for sale.  It was $165 USD including shipping, so I grabbed it.  Now I
need to wait for it to arrive, and the uncertainties galore: we still
have no solid proof, only a suspicion, that the RxTx module is the one
at fault, and there is no certainty that the replacement module is
good either.  But this is all I am currently able to do in attempts to
repair this faulty CMU200.

Now onto the other major CMU200 issue: calibration of the unit itself.
Out of the 3 calibration parts required for Calypso GSM devices (VCXO,
Rx gain, Tx power levels) the last one (Tx power levels) is the
trickiest, and the only one for which I have not yet developed the
necessary automated calibration software.  The biggest difficulty with
Tx power level calibration is that the calibration process absolutely
needs to know what the produced power levels are with high accuracy,
and any not-accounted-for losses in the interconnecting RF cables or
any slight miscalibration of the measuring instrument itself will
throw the process off badly.  Thus the CMU200 itself needs to be in
good calibration standing, and the loss in the interconnecting cables
needs to be measured and accounted for.

The CMU200 that I have came from ebay, and its calibration history is
unknown.  Given that the seller shipped it to me without having caught
the fault in the signal generator function, I feel that it would be
foolish to put blind trust in its calibration status: while the Rx
path (the RF analyzer function) works, we have no way of knowing if
the power level measurements it reports are accurate within 0.5 dBm
like they are supposed to be per the datasheet, or if that calibration
is off.  A related issue is that if I replace the RxTx board inside
the unit in order to find the Tx path problem, the unit will almost
certainly need to be recalibrated before its power level measurements
can be trusted to the required degree of accuracy.

And the problem with getting a CMU200 unit (re)calibrated, whether it
be the unit I already have after RxTx board replacement or another
cheap CMU200 off ebay with unknown calibration history, is that I
still have not found any shop where I can get this (re)calibration
done; I found some places that might be able to do it, but I haven't
got a confirmation for sure, and I am afraid that their asking price
for the calibration service may be astronomical.

So far the only will-work-for-sure option I've been able to find would
be to buy one of these expensive units:

http://www.ebay.com/itm/222146419296

Their warranty is only 30 days, but now that I have already climbed
the CMU200 learning curve, 30 days should be plenty enough to fully
test the unit immediately upon arrival.  And these expensive listings
are the only ones I could find where the seller explicitly warrants
that the unit is properly calibrated.

The problem, of course, is the price.  Midwest ATE (the company
selling the unit linked above) has several CMU200 units with different
option configs (most of these options aren't needed for FreeCalypso RF
testing and calibration), and the one linked above with the asking
price of $2995 USD is their lowest.  Other companies with similar
offerings that I found ask even more.  Now they do have the "make offer"
option, but I doubt that they would budge below $2000 USD, and the
latter figure is still more than I can afford right now.

So if any of the new people on the list (from REcon etc) would like to
help the project move forward faster, and can donate $2000 USD toward
the purchase of a known-good CMU200, here is your chance. :-)
Otherwise I shall continue with first trying to fix the CMU200 which I
already have, and then trying to get it calibrated, and the whole
process may take a few months, during which time there will be no
further progress on FreeCalypso.

Hasta la Victoria, Siempre,
Mychaela aka The Mother