JDSU JD745A / GC747a Base Station Analyzer - Repair log

[Qw3rtzuiop]

Hello,

i have a broken JD745A aka GC747A here which i will try to repair.

It is a 4GHz mobile spectrum analyzer with decoding and vna functions.
Unfortunately i didn't find much resources on the internet.  If anyone has some sort of service manual or block diagramm i would be happy.

The major defect is visible on the two images. The displayed level is way to low except for the range from 20 MHz to 105 MHz.
And if the preamp is on that range changes up to 680 MHz.

I started to draw some diagrams to get a better understanding of this device.

Despite this problem the connectors are pretty unreliable. The contacts are dirty or even corroded.
Maybe some IPA will do it. Otherwise they need to be replaced.
A movement of those cable will result in a reboot of the unit which needs some time.

[Qw3rtzuiop]

I was able to remove some loose connections.
Now its time do take a deeper look at the rf-frontend.

Today i've just done some research and collected datasheets for the main components.
Next thing to do is drawing a block diagram.

[Qw3rtzuiop]

After some more experiments it turns out to be a digital problem
Even with a disconnected ADC input the problem remains. Im suspecting broken calibration tables.
Unfortunately i didn't find a way into the memory yet.
Telnet and FTP are secured by unknown passwords. I don't know if have a chance by listening to the JTAG ports of the FPGA and CPU.
My JTAG experience is non existing but I'll try it later on.

[testpoint1]

such meter from the JDSU actually is from Gencomm of Korea, so the RF circuit is quite fair, the performance may not so good (speed and dynamic range etc), also I found some flash eprom corrupt problem, for your case, just check which chip include the data you need, copy from a normal meter is OK.

[Qw3rtzuiop]

Major breakthrough update!

In the meantime i desoldered the nor flash to take copies of them. I didnt succeed to analyze them but thankfully @dxl was able to do it.

He also pointed out some possibilities to get root access. The device runs the file recovery/recovery.000 with root privilege from an attached usb stick at startup in order to start a recovery of a bricked device.

That way i modified the etc/securetty to allow the root access over telnet.
I attached the root telnet "script" and the contents of the /gencomm/cal folder. If someone has a working device i would appreciate if she/he can share their contents of the /cal folder for comparison.

[Qw3rtzuiop]

Looks like the freqcal.dat holds the calibration. Just plotted the content of that file. There is a good correlation to the the traces from my earlier experiments. Time to organize a trustworthy 4GHz signal generator to generate a new cal file.

[Qw3rtzuiop]

Modifying the freqcal.dat indeed changes the displayed spectrum.

When looking at the DANL values from the datasheet it seems like im still missing about 20 dB of sensitivity.
The input attenuators, preamp and the adc itself are proven innocent.
Out of curiosity measured the input impedance and VSWR of the spectrum analyzer with the built in VNA.
I was hoping to see some major offset caused by a loose solder joint at the input. But the values arent that much off from the datasheet values (1.5:1) considering that i just used a home made cal kit and an additional cheapish SMA->N adapter.

Further testing needs another teardown of the rf board. Im hoping for a failure close to the input because i didnt figured out the architecture further down the signal path yet.

[ferdinandkeil]

Hi there,

another GC747A owner here 

What settings did you use for the DANL measurement? I did a measurement according to the datasheet (RBW 1 Hz, 1 GHz < fC < 2 GHz) and got pretty good results (see below).

Do you still need the calibration data?

[Qw3rtzuiop]

Yes I also measured according to the datasheet with RBW 1 Hz and 50 Ohm termination. Thank you for your testing!  Your measurements seem to confirm that there is something off in my unit. I can get either correctly displayed input levels with a too high DANL or a correct DANL with to low levels.
At the moment im waiting for longer mmcx cables to access the rf board while the device is in operation.

If you can copy your /cal folder it would be nice for comparison, but i think not necessary to get the device fully working again.

[ytterligare]

New owner of a JD745A here !!
Did you succeed in fixing your puppy ?
Being radioamateur, I’m enjoying a lot this device...wish I could afford the 985 model ( 8Ghz model ) but also like that, it is really a nice multipurpose machine, and at a reasonable price in the second hand market ( although I payed 1000 € for it and additional 500 for shipping and duties from Korea to Italy...)

[Alenoiz84]

hi, I have a gc7105b that gets stuck in the boot screen with the gencomm logo.  maybe the firmware is corrupt.  I tried to search the network for the recovery file but I did not find anything.  where did you manage to find it?

[Qw3rtzuiop]

where did you manage to find it?

Here:
https://updatemyunit.net

[suj]

New JD745A owner here. Mine has a problem with the signal level in the spectrum analyzer and internal power meter (which uses spectrum analyzer hardware). The decrease in the indicated signal level is +/- inversely proportional to the frequency. At 3..4 GHz the indications are correct, at 1 GHz it is about -4 dB and at 100 kHz it is over - 20 dB. This may indicate a problem with coupling capacitor or one of the input ICs. I want to check the input capacitor and next element (3 dB attenuator?) by feeding the signal from the generator to the input and testing with another spectrum analyzer with the HP85024A probe. I have a problem with removing the mixer module from the case. Where is it still screwed on? Is there a screw under the sticker? The photo shows the state in which I stopped.

[Qw3rtzuiop]

Where is it still screwed on? Is there a screw under the sticker? The photo shows the state in which I stopped.

I marked the srews that hold down the module. There arent any srews under the sticker afair.

[suj]

>I marked the srews that hold down the module...

Thank you, I thought these screws were for these peripheral modules.
My measurements of the input elements (coupling capacitor, attenuator or equalizer, diode limiter) did not show such deviations as I have noticed before. After putting it back together, the spectrum analyzer turned out to be correct in the 600 kHz to 4 GHz range. I'm not sure but maybe it was some PCB stress or problem with the N/SMA connector at the input of the spectrum analyzer? At the moment, this range of correct indications satisfies me, and I will check whether it is a permanent solution from time to time. Alternatively, I will check whether the range below 600 kHz will be possible to correct in the calibration files.
BTW. I weighed the screws used to tighten the PCB screen of the mixers. It's good that I used an electric screwdriver to unscrew and screw it back together. 

[Qw3rtzuiop]

I'm not sure but maybe it was some PCB stress or problem with the N/SMA connector at the input of the spectrum analyzer?

Hopefully your problem wont appear again.

I found a cracked mlcc at the IF path that caused some losses of my unit.

[pieroandreini]

Making a pendrive with the recovery files you share seems don't work on my J745A with firmware 3.120.043
Which version should I install? May I make a downgrade safely?

[Qw3rtzuiop]

Making a pendrive with the recovery files you share seems don't work on my J745A with firmware 3.120.043
Which version should I install? May I make a downgrade safely?

Hi i used the latest firmware 1.122.000. The 3.120.043 is the latest firmware for the newer B model.