E4433B RF sig gen repair

[microbug]

I picked up this sig gen from work before they threw it away ('beyond economical repair'...). I was hoping for the same MOSFET that was blown in many other people's to be the problem, and was very disappointed when I opened the output board only to find a sea of custom Agilent parts. The layout was different too so they aren't just rebadged. These parts all have the prefix 1GG (e.g., 1GG3/1GG7) and appear to be internally 50 Ohm matched.

The instrument fault is an 'unlevel' displayed error, and the output of the unit is fixed at 2.628GHz regardless of the frequency set on the front panel. The output attenuators seem to work (the signal is reduced in 5dB steps) but fine-grained attenuation doesn't work. The output drops slightly to 2.622GHz when the set frequency is 0-250MHz. These sig gens have a different signal path in this low frequency range -- the synthesiser board generates a 1GHz - f_set carrier, and then it's mixed down with a 1GHz reference on the output board.

I poked around the board by soldering coax to different points. The output power amp seems to be doing fine. There is all sorts of crap coming out of the IQ mixer (SOIC-14 part labelled 1GG7 4225): there is always a peak at 2.72GHz. The correct carrier peak is also there when the set frequency is above around 2.95GHz -- below that it rolls off quickly. I probed at the input to the 1GG3 amp after the mixer (i.e., after the two small transformers).

The IQ mixer input looks correct -- single peak at the carrier frequency, like the carrier output on the back of the unit.

This leads me to suspect the IQ mixer (ignore the missing DC block cap in the photos -- I damaged it but will replace). I'm still a bit confused about where this 2.72GHz signal is coming from though -- I don't see why the mixer would need a fixed reference that could leak through (and besides, the fixed reference available is 1GHz).

I'd appreciate any ideas or suggestions. I've attached some pictures. The block diagram of the output board is at page 35, http://www.doe.carleton.ca/~nagui/labequip/synth/manuals/e4400335.pdf.

I might try to get a copy of the CLIP (including the full schematic) from this guy. I'd have to pay for it, but it would easily be worth it to get the sig gen running.

[R-1125F]

Have you seen this video of an E4433B repair?

[microbug]

If you mean the Signal Path videos, yes, I've watched both of them.

[R-1125F]

These generators have trouble with the frac-N chip in the synthesizer board.  Have you checked the synthsizer and reference board?  A bad frac-N can cause all kinds of problems. 

I have bought the CLIP.   I think it for an A model but it is still useful for troubleshooting.

[microbug]

These generators have trouble with the frac-N chip in the synthesizer board.  Have you checked the synthsizer and reference board?  A bad frac-N can cause all kinds of problems. 

I have bought the CLIP.   I think it for an A model but it is still useful for troubleshooting.

I haven't checked the synth board, but its output looks good (as measured at the coherent carrier output and on the output board before the IQ mixer). Carrier between -5 and 0dBm, with some small harmonic peaks.

Edit: I've emailed Milan to buy a copy of the CLIP.

[microbug]

Ok I now have the full schematic. Screenshot of the IQ mixer section where the RF path starts going wrong is attached. Any help would be appreciated to try and identify where this 2.7GHz signal could be coming from.

I'll try to have a poke around some more testpoints and check the synth board this weekend. I'll also check the I and Q inputs to the mixer.

Edit: sadly the CLIP doesn't have much info about the 1GG7-4225 mixer IC. There's no info except what you can see on the schematic.

Edit: any ideas as to what Vdi/Vdq, Vci/Vcq, and Vblo do on the mixer IC would also be useful. My RF hardware knowledge is a bit thin.

Edit: so far I have found that:
- QUAD LOOP is muxed onto ABUS, which is an internal analogue bus used for instrument calibration.
- I/Q DRIVE and I/Q BAR DRIVE come from either the instrument I/Q inputs or another off-board source. An AD7568 DAC and some op-amps provide gain and offset.
- VBLO DAC and QUADRATURE DAC come from the same AD7568 through some LF412CD op-amps.
- PRE LEVEL DET appears to be compared to PRE LEVEL REF DAC, which also comes from the same AD7568 to generate a signal called PRELEVEL DRIVE. That goes off board.
- Based on a public Agilent/Keysight datasheet for a similar part, Vdi/Vdq are probably the detector diode bias current, and Vci/Vcq are probably phase control voltages. Not 100% sure about this though.

[R-1125F]

Just to be clear; you are getting the full range 250kHz to 4GHz from the coherent carrier output?

[microbug]

Yes. The coherent carrier output is actually only 250MHz-4GHz. When the instrument output frequency is set below 250MHz, the coherent carrier output is 1GHz - f_set (i.e., 750MHz-1GHz as you decrease from 250MHz towards 0).

I have confirmed that this is the expected behaviour using the CLIP -- the coherent carrier output is a buffered copy of the input from the synth board, which behaves as described above. The block diagram also confirms this.

Edit: added screenshot of block diagram (see first post).

[vaualbus]

If you have an old x86 computer, you can download the service SW and used to access the ABUS reading maybe that help. You need a GPIB adapter (as I think the service SW does not work over the serial port unfortunately).
Also, if you need that part you can search for some spare vector board from the E5515C wireless test set as they use the same output board. (Early models use exactly the same board, newer use a modified board where they remove the final mixer for the frequency below 250Mhz and change a lot of the gold chips to standard SOT something chips, but the IQ modulator is the same! and the spare boards seems to be cheap).
PS early on I suspect I had a failed output board (was not the case ) and I ended up buying all the RF board of the E5515C for like 50$ (indded I spend more on shipping than the boards ) so I have a lot of custom ICs ahah 

[microbug]

That's a good idea. I'll see if I can borrow one from work.

Edit: also hoping I can get away with a VM rather than a real PC.

[vaualbus]

I think I get it to work via serial on a VM, no idea If you can share the USB though.

[vaualbus]

LIke this one is for E5515C, probably there are cheaper boards on ebay but I left this for "reference"
https://www.ebay.com/itm/351978940844?itmmeta=01HTPP8C54WVKYDSHBZ7JT78JM&hash=item51f39469ac:g:ldAAAOSwIgNXwKJ0&itmprp=enc%3AAQAJAAAA8O%2BU5csbRSVAQXQxB%2FArsZ16JBwdUHxXmWYRqoI%2BHYaM0C04oeYIYNFx3ClTyG7zLPVz8ZaI2DpnORLI6caCHqO7cwaiVSt6ksVGZtXJqggOz7sBRG6%2FlY13DW2IiKsfbpGwx2cLughO1YyJAF7RxCiWV7eiO1rjcA2gTTAtledSB2wC5S%2BhOALahS%2FZeBT7k4CJ1bAVeATjajLIraIsQAvIomqzSKiAnRftFhRhYMjxI9JN%2FNWbNscP3i2U52f3VnXtLV2kPBtKKMGBCwffxTJJGYtJeWAMHz%2F%2BMXTUA81PqvbGiAGkeIvou2N7U%2FbcBQ%3D%3D%7Ctkp%3ABk9SR9TCodbVYw

[microbug]

Thanks, good to know there is a source for the part. I did find a few on grey market sites, but who knows if they're genuine (and the price is similar to the whole board). I'll borrow a GPIB adapter and a spectrum analyser this weekend and try to dig a bit deeper. This 2.7GHz signal doesn't make much sense, I think I must be missing something...

[vaualbus]

This one for 50 bucks is like good anyway as you get way more money just from the parts in it (excluding the custom IC, I guess  )
https://www.ebay.com/itm/264970900545?mkcid=16&mkevt=1&mkrid=711-127632-2357-0&ssspo=a_NoMhZNTx6&sssrc=2047675&ssuid=&widget_ver=artemis&media=COPY

[microbug]

Thanks, I've asked about shipping to the UK.

[microbug]

I must apologise to R-1125F. On checking again, the coherent carrier output is looking pretty suspicious.

After boot, the sig gen is set to 4GHz. Notice the peak at 2.72GHz, which is the same as what I'm seeing at the main RF output.


Sweeping the output through the 250kHz-4GHz range, it got worse. The 2.72GHz peak is always present, and is particularly bad around f_out = 2.434GHz.



The correct carrier does always seem to be present, and is always the loudest signal (amplitude of 1-3dBm).

I have forgotten to bring an N-type to SMA adapter home, so can't check the actual RF output today. I'll have a quick look at the synth board...

If anyone knows what the coherent carrier output should look like (in terms of images present and carrier amplitude), do let me know. I will have a dig through the service manual if not.

[vaualbus]

I think you can look at the coherent output in one of te TSP repair video. it had a failed generator with ALC error and it foe sure measure the coherent output.
The video should be around 4 years old(never mind it's already 9years, wow time is passing so fast!)  let me post here the link:
https://youtu.be/RSCGAY5z_44?si=z0MPqTE5SF31ma4G

[microbug]

Found the coherent output discussion, at 18:30 in the video. Looks like my levels are ok (his were -6 to -3dBm, mine are 1-3dBm).

There are spectral images in the video, but otherwise his spectrum looks fairly clean. My spectrum is generally bad with the 2.7GHz signal, and at 2.434GHz is particularly bad, with multiple peaks that shouldn't be there. There's nothing like that in the video. That does lead me to suspect the synthesiser board.

The fact that this 2.7GHz signal is always present narrows the search down to areas of the synth board that are common to all frequencies. There are some switched bandpass filters on the synth board, in tight bands between 250MHz and 1GHz. They are quite aggressive (11th order). I would expect these to cut the 2.7GHz signal out. Either they aren't working, or the signal is appearing after the filters. I need to sweep the frequency slowly through the full range and monitor the 2.7GHz signal level. This should reveal whether different filters are switching in and out.

Another clue: the PLL out-of-lock LED (DS1) is on in some frequency bands. There are at least four different bands where the PLL is out of lock. I will look closely at U201 (UPB584G -- NEC 2.5GHz div2 prescaler, marked Agilent 1GC1-4209), because its datasheet warns of unwanted self-oscillation if the input is too small.

I also have been trying to install Windows XP to get the Agilent Falcon service software working (using this post). Have had some difficulties with GPIB-USB-HS drivers and the Agilent software crashing.

[microbug]

Here are some pictures of the synth board. The right side is the phase locked loop and splitter. The middle generates FM baseband to mix with the VCO tuning voltage (I think). The left includes a /2 divider and two doublers, and a lot of switched filters. The gold part near the connector is the output buffer.

Edit: the label that says 'microstrip to output section' should be 'microstrip to filters/dividers/doublers'. The output section comes after these parts.

[vaualbus]

Well more interestingly is that, if the PLL get unlocked, why the instrument never report such error?
Maybe check the reference 10Mhz signal and the different supplies on the board.

[microbug]

It now reports error 508, frac-N loop unlocked.

This definitely wasn't there before -- the only error was oven cold. UNLEVEL was displayed but no error was present. I have seen someone online mention that the frac-N error can be intermittent, but it's also possible I've caused it. It goes away if cleared when the out-of-lock LED is off, so I think the error reporting is working.

I checked the supply rails on the synth board (had to solder wires to testpoints and re-attach the board), and the 10MHz reference output on the back of the unit. All were fine.

Edit: the frac-N error has now disappeared again. It is quite likely that it appeared and disappeared before and I didn't notice -- after I first checked the error queue and found only oven cold, I didn't check it very often again.

[vaualbus]

Oh I see! Anyway the interesting thing is that the Frac-N locks the signal before the doubling and filtering stage, I rather would check, as you said, that part.
So If something is wrong on that path it would never unlock the PLL. Maybe also having the board case removed cause more unlocking due to  the board not been shielded?

[microbug]

Yes, it could be that the board being unshielded is causing the PLL to unlock. I will put the shield back on and do a frequency sweep. If I don't get the unlock error then I think I could ignore it for now.

[vaualbus]

Any new regarding the repair? I am curious  if you found something bad on the filtering  and doubling section!

[microbug]

I’ve been quite busy this week. Don’t worry, I’ll have a look at the weekend!