After a few minutes, I decided to buy it.
Waiting for the packageThe package took almost two weeks to arrive. In those two weeks I discovered something I didn't expect. There are no Anritsu service manuals, neither schematics available. The only thing I found was a maintenance manual, which is only a testing procedure list. It doesn't even tell you how to calibrate the unit, only how test its correct operation.
Also, an Internet search looking for someone who has ever repaired one of these units was unsuccessful. I started to think that buying it maybe was not a good idea. I even started to think how to sell it by parts, just to minimize money loss.
The package arrivedOnce the package arrived the first thing I did was to turn it on and test the unit. It worked. I was able to see some signals generated by myself but as soon as I saw them on the screen something was clear: it was deaf. Deaf as a doorknob. It measured 50 to 60 dB less than the real amplitude.
In the maintenance manual there is the only clue about how the MS2711A works inside: a block diagram. The input goes to the step attenuator and a limiter/protection circuit. Maybe the attenuator was faulty? Maybe the protection circuit was burned out? Maybe, just maybe, it was possible to repair it.
The attenuatorWhen opened, I saw the analyzer is built in sandwich form. The upper board contains the LCD and keyboard stuff, the middle board is the logic board, containing the CPU, some DC-DC regulators and a lot of missing RF circuitry. The lower board, full of metallic shields, is the actual RF board. It was easy to find the RF attenuator section, because there is a small semirigid cable from the N connector to a SMA in the RF board, and that SMA is over a small shield. So it was clear the attenuator was inside that shield.
I soon started worrying. Someone has already removed that shield, and placed it back using a small amount of solder: someone tried to repair it, and that is usually bad. Because the shield was soldered with a few solder points it was easy to remove.
The first thing I saw is some grounding was missing. Surely it was teared when they removed the shield the first time. Anyway the layout was quite clear. The SMA (left-lower bottom) goes to a DC blocking capacitor, then to an attenuator section constructed around a miniature Teledyne relay, a microstrip (low pass?) filter, the limiter / protection, and another two attenuator sections switched by... pin diodes?
A closer inspection with a magnifying glass revealed some worrying things. The DC blocking capacitor was broken in several pieces, the limiter / protection device had a crater in it, making impossible to read the marks: PM-something, and all soldering around the miniature relay were hand made.
First thing I did was to figure out the schematics in this section:
First guessed attenuator schematics (click to enlarge)
All resistors markings were still visible, so it was "easy". Two components are marked with an interrogation. I don't know what are they. They are ceramic, maybe compensating capacitors? Once the schematic was drawn some things caught my attention.
First, the protection diodes seems to work as a rectifier device. At first I thought it could be reverse biased to keep stray capacitance to a minimum but it is not. I see a positive rectifier and a negative rectifier but I don't know how those rectified voltages are used by the analyzer, neither how they can limit input signal if they work as a rectifier. After the coils there are 5.1 volt zeners (not drawn). Maybe an overload sensing circuit?
Second, the three attenuator blocks seem quite classical, except the first one. The second one is a 10 dB T attenuator. The third one is a 20dB pi attenuator. It is obvious the first one, the one with the relay is another 20 dB one, but resistor values do not matched. The first two resistor are 121 ohms in parallel, making a 60.5 ohms one, just like in the third attenuator. But the series resistor differs: 274 vs 221 ohms. And the last resistor also differs, 56.2 ohms vs 60.4 ohms.
Suddenly I remembered these resistor had hand made soldering, so I figured out immediately the one who first removed the shield found those resistors burned and for sure he tried to replace them with the closer ones he had.
I noticed the two 121 ohms resistors were damaged. One measured around 1600 ohms and the other one nearly 10 kohms. Then I decided to remove all parts and check all of them. This was the point where things started to become scary:
Many resistors were damaged, and all pin diodes where killed.
Resistors are easy to find, all of them were 1% in 0603 size. But what about the pin diodes? They all have AF3 marking on top. A quick Google search show the only SMD device in SOT23 package with AF3 markings is a SMV1249-004 hyperabrupt varactor diode from Skyworks. A RF switch with a varicap diode? I don't think so. Also SMV1249-004 is common cathode, and these diodes should be common anode.
A clue come from the switching signals, A, B, C and D in my schematics. When the analyzer wants to baypass an attenuator (for example, the second one) it puts 0 volts on A, and -5 volts on B. When it wants the attenuator to be active, it puts -5 volts on A, and 0 volts on B. These voltages does do not make sense with pin diodes. +5 and -5 volts makes sense, but zero volts? Maybe there was a missing +5 volts supply line...
The control lines come from LMC6772 dual CMOS comparators. One for each attenuator block. This comparator has open drain output, so if I found the pull up resistor, I could find the faulty +5 volts line. In the MS2711A these comparators were supplied from +5 and -5 volt lines, and their outputs had a 100k pull up resistor tied to... ground.
There was not a faulty +5V line. The control lines were intended to provide 0 and -5 volts. How those pin diode switches could work with 0 and -5 volts?
The answer was easy, they couldn't because they are not pin diodes. I realized there is another electronic device, usually used to switch signals, that could work with these 0 / -5 volts control signals: The FET transistor.
Suddenly everything had sense. They were not pin diodes, they were N-FET transistors. I was happy., but the joy lasted only a brief moment. What N-channel FET in SOT-23 has AF3 marking? I really have no idea. Even Google had no idea. Except for the small number of 0603 resistors (easy to find), I had six unknown FET transistors, and an unknown DC blocking capacitor. How to replace them if I don't know what components are they?
At this point, concerns increased even more. With no components in attenuator PCB, I injected signals at the output, this is, at the input of the next stage. This is the same as set the attenuator value to 0 dB. The analyzer was still deaf. There must be more components broken in the next stage, under a really big metallic shield.
Under the mixers shieldAccording to the block diagram, the MS2711A uses two different mixers, each one for a different frequency range. They are switched by some kind of device, and taking into account the entire attenuator block were damaged it was easy to expect that switching device to be also damaged.
The problem is the mixers shield is really big. I was afraid to damage the PCB removing it but using hot air it was easy to remove.
The mixer block (click to enlarge)
I immediately saw the switch. This time it was not made with FET transistors, but with a SOIC-8 integrated circuit. The only markings on it were W2.
Pins 1, 3 and 4 are ground. pin 2 is RF input. Pins 4 and 8 are RF outputs, and pins 6 and 7 are control signals with the same levels as the signals used in the attenuator block: 0 and -5 volts.
With all these clues it was not difficult to find what IC is. The SSW-224 appeared as a candidate: it is marked as W2, it has the same layout, it has the same packaging... no! The SSW-224 has a heat sink plate at the bottom and has ceramic encapsulate. This one has no heat sink plate and is made of plastic.
It was not the SSW-224, but I found two promising candidates: SSW-108, and SSW-208, but all of them have the very same problem: They are all obsolete and impossible to find.
I removed the switch and applied signals directly to pins 5 and 8, and I saw the signals on the screen This seems to be the last component damaged in what there is no doubt it was an excess of power applied to the input. At this point I had all the damaged components delimited, but... What model of FET transistor are those AF3? Where to find a SSW-108 or SSW-208?
The logic board and its missing RF componentsThe logic board caught my attention from the first time I saw it. One side has a lot of RF traces, but only passive components mounted. There were six large delimited areas intended to be covered with RF shields, and many components soldered, but not all. It was clear all this stuff is just the tracking generator, which is an option in this spectrum analyzer. It is not an installable module, just the PCB has or has not the required components soldered.
The upper side of the logic board
The block diagram of this tracking generator were quite clear. It seems to be very similar to the receiver side. In fact, you can see the places for the two mixers, the microstripline filters, almost identical to the ones at the RX side. I could identify the "tracking generator power amplifier", a chain of 3 MMIC, and after it, something that at this point was quite familiar to me:
Yes!, the tracking generator uses an attenuator block almost identical to the one at the RX side, and fortunately for me, it was almost complete!
There are nine (!) AF3 marked transistors, the same 1% 0603 resistors with the same values, Even the DC blocking capacitor was soldered. The only part missing was the limiter or protection device, the one marked PM-something at the RX side.
I desoldered the DC blocking capacitor and measured it. It was a 47nF one. Because the maximum DC voltage that can be applied both at the input and tracking generator output is 50V, this capacitor must be at least a 50 volt one, and a very good one to work up to 3 GHz. I soldered it again in the input attenuator.
I also desoldered all resistors needed and resolder them in the correct places at the input attenuator. Then, I desoldered one of those AF3 transistors and tested it using my chinese transistor "texter":
Just as expected, they are FETs transistors!. I desoldered another five and placed them at the input attenuator. At this point I had the complete final schematics of the input attenuator:
The final input attenuator schematics (click to enlarge)
The only missing components were:
- Two 121 ohms, 0603 1% resistors. In the tracking generator attenuator a 60.4 ohms resistor is used. At the input attenuator two 121 ohms, 1% resistors in parallel are used, presumably to dissipate enough power ( up to +20 dBm ). These were easy to find.
- The PM-something SOT-23 protection/limiter double diode. The most probable candidates are BAR66 and BAR67, both of them marked as PMs. These two diodes are very similar. Even the datasheet show identical values for the properties in common. It will not surprise me if these two devices are physically the same, but one is sold as a rectifier diode, and the other one as a protection diode. Why? Nobody knows...
- A small piece of copper foil, needed to restore the missing ground. It will be sewed with copper wire through the existing vias to minimize parasitic inductance to ground.
After one week waiting all components to arrive I finally could complete the input attenuator. To test it I placed an small piece of wire between pins 2 and 8 in SSW-108 / SSW-208 footprint and applied some known signals. Everything worked as expected. The input attenuator worked and I could see signals with reasonable levels. Wonderful!
The RF switchThis spectrum analyzer uses two different mixers in its operation. One for the 100 kHz - 2 GHz range, and the other one for the 2 - 3 GHz range. The switch over point is not exactly at 2 GHz, but around 2060 MHz.
I know the switch must be a SSW-108 or SSW-208. The difference between them is what they do with the unused port. SSW-108 connects the unused port to a 50 ohms load resistor (this is why it is called non reflective switch) and the SSW-208 just left the unused port disconnected.
There are some RF switches that can replace these SSW-108 / 208, but none of them have the same pinout or the same packaging. Modify the PCB to accommodate one of these new switches is a possibility, but not an easy one: both the input attenuator and the mixer compartment are made in a multi layer PCB, and the ground plane is an inner layer. So you can not access it on either side.
Without the possibility to reach the ground layer, the grounding of the new RF switch will be poor, and performance on high frequencies will suffer. A possibility is to dig into the PCB, removing with care layers of fiberglass until the ground layer is fully exposed in an area enough to accommodate the new chip, but that is a very delicate procedure and if something goes wrong there is no turning back.
I spent some days thinking about it. I just could place a piece of wire between pins 2 and 8 and use it as a 100 kHz - 2060 MHz spectrum analyzer, but...
Looking in detail the SSW-208 datasheet I noticed some interesting things. The switching voltages are 0 and -5 volts, just like the ones used in the attenuator section. In fact, these voltages come also from a LMC6772 dual CMOS comparator, just like the rest of the attenuator section. To connect pins 2 and 8, you must apply 0 volts to pin 7 and -5 volts to pin 6. To connect pins 2 and 5, -5 volts must be applied to pin 7 and 0 volts to pin 6. It seems the SSW-208 switch is also a FET based switch.
Most likely SSW-208 internal diagram
Homemade RF switchThe days went by and no solution was found. It was very frustrating. It was almost repaired, but only a small SOIC-8 integrated circuit was missing. I even suspect what is inside the IC, but I can't find one anywhere. One day by pure chance I saw a kit of a home made discrete 555 chip.It was funny, but it ignited the spark: what about a homemade discrete SSW-208? I know what should be inside the chip, and I still have three spare AF3 marked FETs from the tracking generator attenuator.
It seems it can be done
First I placed the two FETs, then I add an small piece of wire from the pin 2 to the two FETs. I soldered two small 0603 10kohms resistors to the gates, and with small pieces of wire, I connected the resistors to pins 6 and 7.
My homemade version of SSW-208 RF switch
After mounting all components in the small SOIC-8 footprint I tested the new switch. I used a small ADF-4351 board with a 20 dB attenuator. I generated a signal a bit under 2060 MHz and a bit over 2060 MHz (2010 and 2110 MHz) and the measured level was the same at both sides of the switchover point. The switch works!
Then I only need to solder the shields back, check again all functions and voila! Analyzer repaired.
ConclusionThis is the first time I found spare parts inside a device although not intentionally put there for that. The most logical thing to do if no tracking generator will be mounted is not mount a single component, isn't it? But almost all passive ones were there, and fortunately for me, those mysterious AF3 marked FETs where mounted also.
They allowed me to repair the complete attenuator block and made a discrete version of the SSW-208 switch. Isn't it wonderful? I have been lucky this time. Very lucky. It's clear repairing this kind of equipment has inherent risks, like finding faulty obsolete components that are no longer available, difficult to find (and usually expensive) custom parts... a lot of things that will push you into what I call "creative repairing" a technique that sometimes gives good results.
You have to bet from time to time to win something.
I still wonder what are those AF3 marked FET transistors. I'm curious to see what specifications they have to be used as 3 GHz RF switches with great performance. If you know what transistors are, please tell me.
Miguel A. Vallejo, EA4EOZ