Agilent E4406A Repair

[Tony_G]

Hi All,

I thought I'd share the repair that I'm doing on my Agilent E4406A Vector Signal Analyzer. I picked up this unit as part of a great test gear auction haul (you can catch the videos here: Part 1, Part 2 which includes an awesome 7090A Measurement System, and Part 3).

I really wanted to get the signal analyzer because I'm hoping that the demodulation capabilities of this and it's PC software will help me in a project I have planned.

I should apologize for the image quality as I didn't actually take any images (I should have) but I recorded a video of the work - These images are stills from that video.

Turning the system on showed the self tests which all passed - Pretty positive:


So I tried putting a 50 Mhz, 0 dBm signal into it and saw this:


Maybe it was something to do with the input attenuator:


Nope - Bummer - It still looks like it is 15 dB or so down and I've exceeded my knowledge of the machine (FYI The HP Yahoo group is an awesome place for guidance on any HPAK gear - Highly recommended) - Based on suggestions, the first step was to check out the block diagram  which shows a 7dB attenuator connected to a relay and then into the mixer:


A few people suggested that it might be that attenuator or it could be the relay - Time to start pulling the machine apart - The internal shield has a cool diagram on it showing roughly the layout - Well thought out design:


First thing to check was that 7dBm attenuator - You might be able to see it where my finger is pointing here:


It's more visibile with the front cover off in the bottom right:


Here is what it looks like taken out - 7dB 5W made by Weinschel:


Unfortunately this turned out to be working fine - Here you can see it connected to my E4418B & 8657B:


Now it was time to look at the actual RF board - Here it is out of the machine but still in the enclosure:


And with the enclosure removed - The input part is in the bottom left corner of the board:


Time to check out the schematic - This shows AT1 being the solid state attenuator, U38 being the solid state relay and Z2 being the mixer:


The key piece of information on the schematic is that a 50 MHz Cal signal is switched in to do the self tests. So if the unit is passing the self tests then AT1 should be in good shape. Let's flip the board over and take a look at it - It's pretty funky looking, check out the legs:


Interestingly there were some scratches on the surface of the ICs - I can't see any visible rework on the board so someone either had the board out but couldn't find the problem or they were made during production. The other side of the board shows them on U38 as well:


With nothing visibly wrong I needed to stop and take a think about it, so I produced the first part of the video series on the repair - You can watch it here if you'd like:

https://youtu.be/D7jdur-MGQI

I'll do another post on Part 2 shortly.

TonyG

[Tony_G]

By now I was pretty certain that the issue was with U38, the solid state relay:


so I removed the relay from the board:


And decided that I might get a bit medieval on it:


Booting the unit without any bridging gave this result in the self test:


If you look at the failures, they're all related to measuring the calibration signal either directly or at a stage further down the RF signal path. However the unit 'works' so I bridged the RF input to the attenuator:


and saw this:


Outstanding - it's now showing closer to the actual value of the signal (measured in the first video as 0.61 dBm) and not too bad given that the highly optimized, custom relay has been replaced by a hand held piece of hookup wire. I was interested to see if it would pass self test so I bridged the cal pads:


and did a full alignment:


Everything passed so now lets bridge the RF input again and then see what we read:


Excellent - 0.76 dBm, this means my hookup wire is a 0.15 dB attenuator at 50 MHz This pretty much conclusively indicates that the issue is based around what state U38 was actually in. However, if you remember the schematic there were two control lines:


I measured Cal_SW_A & Cal_SW_B at +9V and -9V, the datasheet for the relay indicates that the control voltages only need to be 0V and 5V. However looking at the scehmatic again I see this:


So the control line of cal_sw should drive the relay control lines towards +/- 10V so overall I think +/- 9V is within reason and shouldn't be the cause of the relay failure. The feedback that I got from the HP Yahoo User Group was that these GaAs relays have a habit of failing in a way that puts 10-20 dB of attenuation on a port so I think I'm good to move on with a relay replacement.

I ordered some cheap, lower spec'd relays from Ebay so I can put them on the board and see if that really does solve the issue before I buy the proper part from Europe for 10x the cost.

You can watch the video of part 2 here:

https://youtu.be/iBonBIo-pKU

I don't know when I'll get the relays in but I'll update the thread when I give that a try.

TonyG

[Tony_G]

Got replacement parts but still not working - Here is the video:

https://youtu.be/C864ibb65YI

TonyG

[4thDoctorWhoFan]

Excellent repair video.
Keep up the good work.

[Tony_G]

Thanks - Next step is to put one of those relays onto a bread board and checking to see if it does in fact work.

My current theory is that the devices do work but the control voltages need to be within spec so I should be able to increase the value of R333 & R334 to cause a bigger voltage drop in the control lines.

TonyG

[Jay_Diddy_B]

Hi,

Check the 1000pF capacitors on the control lines. I had a similar problem in an RF generator. The capacitor was short.

Regards,

Jay_Diddy_B

[Tony_G]

Thanks Jay - I'll take a look at them but I suspect that they're going to be 'ok' because I'm seeing +8.21V and -9.16V on the control lines as measured at pins 1 & 3 on the relay - I would expect to see those lines at 0V if the caps shorted.

[Neganur]

oy there's a CLIP for the E4406A? I remember some talk on the Keysight forum but the link was never posted.

[Armadillo]

Did you verify that the cal switch control actually toggled [MC33178 flip flopped] ?

Or broken traces going into Pin 8 of the switch [unlikely but should check it anyway]

[Tony_G]

Thanks Armadillo - I did check that the Cal_sw line does switch from 0V to 4.5V and that both Cal_SW_A & Cal_SW_B also switch and are opposed as expected. U26 is operating correctly and I see opposing voltages on U38 pins 1 & 3 as expected.

The thing that isn't expected is that the datasheet for the 71018 says the max control voltage is -8V but I see -9.16V as the max -ve voltage - That was the point I was making when I mentioned binning the U38 parts. I'm hoping that the datasheet max is just conservative. When I get a chance to put the IC on the breadboard I'm going to see if it works with 0V & -5V as per the datasheet and then if so I'll see what happens to it at -9.1 - If it's just overdrive that is causing the failure then I should be able to put bigger resistors (R333 & R334) in line for more control line voltage drop.

TonyG

[Tony_G]

oy there's a CLIP for the E4406A? I remember some talk on the Keysight forum but the link was never posted.

I saw that same comment on the Keysight forum but it seems they never posted it - I posted to the Yahoo HP Test Gear Group and a member sent me the CLIP for the RF board. I've hosted that here is you want to grab it https://1drv.ms/b/s!Amqar8_XQ9UzkKoGH1kGOfX_Zxj9WA - This is just the RF Board, I didn't get the rest of them.

TonyG

[Armadillo]

Thanks Armadillo - I did check that the Cal_sw line does switch from 0V to 4.5V and that both Cal_SW_A & Cal_SW_B also switch and are opposed as expected. U26 is operating correctly and I see opposing voltages on U38 pins 1 & 3 as expected.

TonyG

Thanks for the reply. But I imagine you measured the voltage in static. Did it "FLOP" and able to "FLIP" correctly?
We will wait for your external test jig result and your Part 4 video then.   

[Tony_G]

I did measure it in operation (you can see the process in the video at the 21 minute mark) and it swaps appropriately. The IC alone will be interesting I think.

TonyG

[Armadillo]

Did you notice the absolute max rating?

[Tony_G]

Did you notice the absolute max rating?

Yes - That was the point I was making regarding binning the components. If you look at the schematic you can see that the MC33178 is being operated without feedback so it should drive to the +ve or -ve rail depending on the Cal_sw voltage compared to the R271/R269 voltage divider value (2.19V in my case). Now the MC33178 isn't a rail to rail op amp so it will be some amount off the +/- 10V rails - According to the datasheet for a +/- 15V rail the swing voltage is going to be -13.8/+14 into a 2K load.

So having +8.21V & -9.16V isn't outside of what could be expected.

I'm hoping that the E4406A engineers didn't wander over to the MGS-71018 engineers to find out that they could run some instances of the chip at -9.2V but they'd have to be selected (or 'binned' in other words).

This is why I'm looking forward to when I put the IC on the breadboard and try an in-spec voltage according to the datasheet because if it works then I can just up the values of R333/334 to give more of a voltage drop and that should bring the control voltages back into line. However, I'm wondering why the designers didn't just do that in the first case - I suspect things will not be this easy.

TonyG

[Armadillo]

This is why I'm looking forward to when I put the IC on the breadboard and try an in-spec voltage according to the datasheet because if it works then I can just up the values of R333/334 to give more of a voltage drop and that should bring the control voltages back into line. However, I'm wondering why the designers didn't just do that in the first case - I suspect things will not be this easy.

TonyG

I would like to differ. The U26 output is a darlington pair. The control switch input impedance of the U38 is not in the datasheet but I reckon should be high enough.
Initially I thought making the U26 load to about 2K by changing the resistor value and adding one to ground [see attached] would be a sure thing to limit the voltage from exceeding the absolute max value. Then I question myself that in reality, they could have easily not drive the U26 with +/-10VFFE if exceeding the max is their utmost concern if at all. So actually, the series resistor should not be increased or it will retard or delay the response time of the switch, hence the initial design was to drive it with higher voltage to overcome input capacitance. If you see the datasheet, when it comes to frequency response graph, they are using -7v as the supply voltage. So I don't agree with my own modification shown in the schematic.   ! 

[Tony_G]

I'm not sure that is why they're driving it at that voltage as the slew rate of U26 will swamp any benefit of increasing switching speed via higher control voltages. I'm starting to wonder if there were different values based on whether U38 was a MGS-71018 or a 1GG7-4218. I can't find a data sheet for the 4218 to confirm though.

[Armadillo]

What's up with that Part 4.
At least a thriller.!

[Armadillo]

Oh BTW in your next video, don't wear that Cordless Anti-Static Wristbands, it jinked!

Say who!... LOL.

[Tony_G]

The wristband is actually an ESDProduct fabric wrist band (http://esdproduct.com/esd_wrist_straps.php) - The blue color is common which I guess is why those fake wristband makers picked it.

I got new original parts but they we dead on arrival so I've purchased some different switches (from Minicircuits) so I'll be working on that when I get back from my business trip and hopefully have part 4 "It lives"....

TonyG

[Armadillo]

Great!. I am sure we will be looking forward to your Part 4. Keep us all informed of the exact timing.

BTW would you be tongue tight under "LIVE" condition especially when you need to brainstorm at the same time. 

Oh yes, I would really appreciate it, if you would kindly post some clear pictures of that front Type N connector "Bracket", .... you know the one near to the 7dbm attenuator you took out in Part 1. Because I was wondering if I can fabricate that piece of bracket to shift the input from the back to the front, when I have the time of course. I just need a pictorial image of how it is done. That semi-rigid cable, I think not so semi-rigid with 1 or 2 dbm extra loss won't makes a different for me, anyway.
Thanks in advance.

[Tony_G]

Still not fixed but maybe moving in the right direction:

https://youtu.be/osxz-1mzcpQ

TonyG

[Tony_G]

Great!. I am sure we will be looking forward to your Part 4. Keep us all informed of the exact timing.

BTW would you be tongue tight under "LIVE" condition especially when you need to brainstorm at the same time. 

Oh yes, I would really appreciate it, if you would kindly post some clear pictures of that front Type N connector "Bracket", .... you know the one near to the 7dbm attenuator you took out in Part 1. Because I was wondering if I can fabricate that piece of bracket to shift the input from the back to the front, when I have the time of course. I just need a pictorial image of how it is done. That semi-rigid cable, I think not so semi-rigid with 1 or 2 dbm extra loss won't makes a different for me, anyway.
Thanks in advance.

Not ready for Live video yet I think.

I'll get a shot of that bracket shortly for you - I wanted to get this video out first.

TonyG

[Armadillo]

I'll get a shot of that bracket shortly for you - I wanted to get this video out first.

TonyG

Thanks TonyG.

BTW, i was thinking about it, for IF related matters, it needs 2 sources of energy. You looked at one, why don't look the other one. Have you verified those LO references up to spec to those mixers? I mean, If the switch switches, so it switches, the dbm may be marred by the downstream like mixing pin diodes etc...
Just really my 2 cents, or maybe not even half a cent cause I have not look at any circuit so to speak. Pardon my interference.

Good, tenacious nice work up there, keep it up....   

[Tony_G]

Have you verified those LO references up to spec to those mixers?

Thanks - Given that everything works if I manually bridge the pads with a wire, I think the issue is really in the switch and associated circuitry. I looked at the datasheet for the op amp and it should give the values I'm seeing on the switch lines (the graph isn't great resolution but it seems to be around 8-9V for a 10V supply).

I'm going to take a trip into looking at the power supply to see if there is an issue there, 900mV of noise seems way to much to be handled with the spec'd caps.

TonyG