Need a VNA...Or maybe not. I'm running more and more into situations where I could use a VNA and/or spectrum analyser with a tracking generator that works up to 1 or 2 GHz. I have a network analyser but that only works up to 300MHz. A spectrum analyser with tracking generator is useful for analysing impedance of components and my current spectrum analyser doesn't has a tracking generator. I could make do with a sweeping RF generator but for quick measurements this process is too slow. Some VNA features to check impedances would be nice as well. I already have a NanoVNA but even with PC software it is still very limited. Especially where it comes to dynamic range All in all I don't need a real VNA but something that is close.
High on my wishlist is the ability to save images on a USB stick and at least a TFT screen.
I looked at getting a Rigol or Siglent network analyser but decided not to in order to avoid firmware related surprises. I want a device with mature firmware; I've been burned too many times now. So I went hunting on Ebay to see what is available on the used market. Initially I found some older but nice high-end VNAs from R&S and Agilent but even faulty ones go for silly money. Also these would require quite a bit of tinkering to add saving screendumps to a USB stick. Looking further I found several base station analysers that also have spectrum analysis and VNA features. At the lower end sits the Tektronix Y400 (should have the YBT250 and YBA250 modules for most flexibility) which seems to sell around $500 (2.5GHz IIRC). In the middle sits the JDSU 745A/B (4GHz) that sells around $1000; at the higher end you can find the Anritsu MS2721B which is around $2000. Shariar from the Signalpath Youtube channel did a repair / review video on the MS2721B and noted that the spectrum analysis mode is rather slow. So even though the MS2721B works up to 7GHz the slow speed doesn't make it a very desirable unit. I opted to get the JDSU JD745A or B and managed to buy a JD745A. There is also a B version but the specs look to be identical. A higher frequency model is also available but these cost more obviously. The unit I bought had a problem with the pre-amplifier so I managed to haggle down to a nice price.
The JD745A has been designed by Gencomm (a Taiwanese company) and the company JDSU seems to have been taken over by Viavi which is a company that is big in mobile and telecom related test equipment. I have come across Viavi before so to me it is a known brand which sells equipment for professional use. All firmware / software and manuals can still be downloaded from the official website so that is a big plus.
A battery and OSL VNA calibrator are included as well.
The sunglasses shown in the datasheet are missing though. Pity, I think they would have looked way cooler on me.
There are several other repair threads on EEVblog for this unit:
https://www.eevblog.com/forum/repair/jdsu-jd745a-base-station-analyzer-repair-log/https://www.eevblog.com/forum/repair/viavi-celladvisor-jd785a-repair/msg3949858/The first link also contains information about getting into the Linux system. That allowed me to make dumps of the flash memory onto a USB stick so I have a backup of the entire flash. Making a backup also showed an interesting effect: at high processing / memory load the display image gets distorted. Since the display memory is likely shared with the main memory, I guess this is set at a lower DMA priority than other memory accesses. During normal operation the display gets distorted every now and then but it is not a major issue.
Taking it apartInside there is a stack with 3 components. First the main board:
This has a processor from Marvel, an FPGA, a GPS receiver + OCXO and the signal processing from the VNA and spectrum analysis module. The little add-on module is used for E1 / T1 analysis. This module / option can only check if there is a valid signal; it doesn't have the ability to do higher layer analysis.
For a minute I was thinking that the mainboard has a whole bunch of Xilinx FPGAs on it but these are flash chips to hold the FPGA configuration. I find this a very unusual setup. I would load the FPGA from the processor so the design is cheaper and the FPGA image can be updated together with the software. Maybe this was easier to split development across two teams though.
The bodge wires from the mainboard go to the power supply:
The soldering joints ain't bad but I applied some hotmelt glue (hot-snot) over the wires as a strain relief otherwise this is an accident waiting to happen. Didn't take a picture afterwards though.
A very loud fan:
I have replaced this with a 20mm fan with the same power rating but lower RPM. This gives the same air pressure to pull air through the casing at lower noise.
Marking the connectors before removal:
The PSU:
Next out is the VNA module:
Last but not least, the spectrum analysis module:
So the interesting part is that the JD745A has seperate spectrum analyser and VNA modules instead of trying to combine these into one.
After undoing over 100 screws it showed it's inside. At least, half of the inside. There are 2 boards sandwiched between 3 aluminium blocks.
A quick overview: the signal enters at the bottom left corner and enters into a 0dB / 20dB attenuator, pre-amp and fine attenuator section towards the black connector. Then it goes through a lowpass filter down into a switch which allows the signal to take 4 different paths. In order to keep the 1st LO frequency low, the frequency range is divided into 4 bands. The lower band goes through a lowpass filter (bottom of the PCB) while the higher 3 bands go through bandpass filters. The black blobs to the right of the filters are the 1st LO mixers. After some further filtering the signals arrive at another switch (between the blocks with the copper colored flaps on top) and goes into the 2nd LO mixer upwards to the final mixer stage. The LF out is at the left top of the board. The white block in the center top of the board is the 3rd LO mixer.
But the cullprit is the pre-amp so let's take a look at that section:
At the left is an RF switch chip that selects between straigth through or take the preamp path. The device marked with '39' is the preamp itself. The circuitry to the right is a P channel MOSFET and transistor to switch the preamp on/off. Remember the JD745A is a battery powered unit so low power consumption is better. Now to figure out what the '39' device is... From the circuit configuration I can see it is a typical setup for a MMIC. So I feed Google with 'mmic 39' and crossed my fingers. It turned up with a GALI-39 from Mini-Circuits. Pfew... not some kind of obscure chip. So I ordered a few and replaced it.
Because it is not so easy to test the spectrum analyser outside the unit I decided to close everything up and put the device together again. It should work now... what can go wrong?
In the process I decided to replace the cross-head screws with Torx screws because these are easier to work with and several of the cross-head screws where already damaged.
The latter turned out to be a huge mistake:
For some reason the amplitude between some frequencies was too low. Also the preamp still didn't work. One of the things I bought the JD745 is to make scalar measurements. It seems the JD745 is going auto-level control and uses the preamp to boost the signal when it is too low.
This means that the preamp really needs to work for these measurements. This is the result I got when I attenuate the signal. When the preamp is on (and doesn't work), the amplitude will drop because the JD745 thinks the signal amplitude is boosted.
Crap!!! So I had to take it apart again.
This time I tested the spectrum analysis module on the bench using an RF generator and spectrum analyser to trace the signals:
A fan provides some airflow.
A Tektronix P6156 with 20x attenuator (BW: >3GHz !) also works excellent together with a spectrum analyser to look for RF signals:
From here everything looked OK (expected signal levels) except for the preamp.
I managed to trace the control signals for the various RF switches back to the black connector which goes to the main board EXCEPT for the signal that controls the preamp. There is a preamp protection menu option in the settings menu so it could be possible the preamp isn't controlled from the mainboard but through a circuit that does some kind of level detection. However, I have not been able to spot such a circuit on the spectrum analysis board. As I mentioned before the spectrum analyser module consists of 2 boards. The signals to the main board are going to the 2nd (bottom) board which has the connector to the main board through a board-to-board connector. So it could be this circuitry resides on the 2nd board. Time to crack the spectrum analyser module open further and take the 2nd board out:
This has a whole bunch of clock synthesizers to make the various LO frequencies: it is the RF generator board. The signals are fed through 3 pins that are soldered in both boards so I had to desolder one side. The digital signals are transfered using a connector.
No sign of a circuit that controls the preamp and the signal from the connector to spectrum analysis board goes nowhere on the RF generator board. Hmmm.
Now remember I changed the screws... the original screws are 13mm long but the replacements I used are 14mm because 12mm was just too short to have a good grip.
I was expecting the holes to be blind but they are not so I drove a few screws straight into some of the control signals on the RF generator board. Including the signal to control the preamp. So I fixed the preamp but broke it again in a different way
What are the odds of doing that?
I fixed the traces and put some polyimide (kapton) tape over them for protection.
I also noticed that the middle aluminium block had quite a bit of debris on it so I gave that a clean as well:
Some 200+ screws later the spectrum analysis modules was back together again. This time with washers under the screws to make sure they don't hit the RF generator board.
Yes, I also replaced the screws that hold the cover for the RF generator board with Torx but made sure these wheren't too long.
Proof of the pudding:
First a test of a microstripline filter I created many years ago:
In earlier measurements (with the broken preamp) using the scalar measurement, the part below -50dBm would drop down and reside at around -75dBm making the JD745A less usefull for these kind of measurements. The VNA could take over but this has less dynamic range and needs a more cumbersome calibration. Getting the preamp fixed really was important for useability.
And the impedance of a big 1nf surface mount capacitor:
No, there is nothing wrong with the image saving. I tried the reverse color option which appearantly reverses all colors and not just that of the trace area. A quick check at a couple of frequencies shows that the new preamp chip has about the same gain as the original one. The levels with / without preamp enabled are the same. I dreaded to mess with the calibration but it looks like that isn't necessary. I need to do further testing though to make sure.
A remaining issue is that the touchscreen doesn't seem to work for the function buttons but it turns out that the touchscreen is not enabled for those. The touchscreen is only used for using the onscreen keyboard (and maybe some other things that I have not investigated).
Bias tee optionThe JD745A has an option for a bias tee which be used to apply voltage on the RF input of the VNA. This would be handy to measure the impedance of capacitors at various DC voltages. It would be nice to get this to work somehow.
I found this connector on the VNA module:
Hmm. What can be done with this? The unit is so compact that there is no room for an external bias so my guess is that the BIAS tee is inside the VNA modules. Measuring on the connector revealed that one pin is ground and that there is a reverse polarity protection diode on this connector. I tried to apply voltage to this pin and turn the bias-tee on but this didn't work. I modified the license file to enable the bias tee but the option is still listed as 'not installed'. The menu item is accessible though but nothing happens. So either the hardware for the BIAS tee is missing or the license code must be valid. I have not investigated this further. I don't want to open the VNA module if I don't have to.
What can the JD754A do?According to the datasheet, the JD745A is quite versatile. After all it is a basestation analyser that is supposed to be able to do RF troubleshooting as well.
The spectrum analysis mode is pretty quick. Certainly usefull for doing EMC pre-compliance work and use as a general purpose spectrum analyser. In spectrum analysis mode the tracking generator can be freely set to a certain frequency and signal level (0dBm or between -50dBM to -30dBm) so it can be used as a CW (constant wave) generator as well. As expected the signal isn't very clean. When set at 500MHz, there is a 1 small harmonic at 1GHz but when set to 1.1GHz, there are a few strong harmonics.
The VNA (antenna / cable analyser mode) can do both vector and scalar analysis. In vector analysis mode the VNA input is used and measurements have a dynamic range of 60dB. In vector mode an OSLT (open, short, load -50 Ohm- and through) calibration is required. In scalar mode a through calibration is sufficient. It is also possible to show a Smith chart but unfortunately this doesn't have an inductance / capacitance marker.
I took some measurement from a 1uf 0402 and a 100nf 0402 capacitor using the PCB design made available by oz2cpu (see
https://www.eevblog.com/forum/projects/component-tester-board-for-sa-na-impedance-caps-inductors-filters/ )
This graph shows a small bump right at the level where the JD745A starts using the preamp so the preamp may need some calibrating after all.
It also has various cable analysis tools like loss and 'distance to fault' measurements. The accuracy of the distance to fault is quite good (down to centimeter level).
An interesting feature is to use dual spectrum mode. This allows to look at 2 spectra at the same time (see image above). This is handy for basestation use because there are seperate frequencies for uplink and downlink signals. I can see this becoming handy for EMC precompliance work as well. Looking at the entire spectrum and zoom in on problem areas without losing oversight. Or have a split with the 150kHz to 1Mhz range and 1MHz to 30MHz for LISN measurements.
And it has a 'waterfall' display under the 'interference analyser' mode. I did some testing with a 434MHz keyfob:
Less usefull are the various 3G/4G signal analysis options. My unit also has the E1/T1 analysis module installed but it has been ages since I have come across such an interface.
Last but not least, the JD745A has a built-in GPSDO using a uBlox module and a decent OCXO. No 10MHz and 1PPS output though but these should be easy enough to add if you want.
All in all I think the JD745A is a good buy for the money. Maybe even a replacement for my higher end Advantest spectrum analyser but let's see how it holds up for some real world use.
to be continued...
