Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator

[Hugoneus]

Thanks G0HZU,

What you are saying is very plausible.

[Radio Tech]

Shahriar,
Another fine detailed video. Thanks

[G0HZU]

Hi Shahriar

I had an initial rummage through my dev parts box at work and couldn't find very many of these parts.

I did manage to find 1 new unused SHF0189(Z) with the H1Z marking (this is the lead free part) and I can remove a working one from an old dev board that is the SHF0189 with the H1 marking.

So I can send you two parts for the cost of postage if you still need them. I assume it will be cheapest if I send them in a slim jiffy bag by Royal Mail. This will probably be just a few dollars postage with about a 7 day delivery time. I'm not bothered if they get lost in the post so no need to pay for tracking unless you would prefer this? I'm guessing it will be about $12 with tracking for about a 7 day delivery.

[Hugoneus]

Hi Shahriar

I had an initial rummage through my dev parts box at work and couldn't find very many of these parts.

I did manage to find 1 new unused SHF0189(Z) with the H1Z marking (this is the lead free part) and I can remove a working one from an old dev board that is the SHF0189 with the H1 marking.

So I can send you two parts for the cost of postage if you still need them. I assume it will be cheapest if I send them in a slim jiffy bag by Royal Mail. This will probably be just a few dollars postage with about a 7 day delivery time. I'm not bothered if they get lost in the post so no need to pay for tracking unless you would prefer this? I'm guessing it will be about $12 with tracking for about a 7 day delivery.

Thanks a lot. How would you like me to pay you? PayPal?

[G0HZU]

I've sent you a PM

[TSL]

Hi Shahriar,

Great video and very timely for me.

I just picked up a 4432B ESG-D with opts 100,1E5,H99,UND for $700AU

It has the same fault indicator. I'm just about to embark on a fault finding mission to see if its the same device that failed.

FYI you can buy SIRENZA new parts from UTsource here...

http://www.utsource.net/ic-datasheet/SHF-0189-1906844.html

regards

Tim

[TSL]

Hi all,

Just finished repairing my 4432 and guess what ?

It was the same fault that Shahriar had in his, that FET had failed. Mine thinks its a 25ohm resistor!

Not having any on hand and thinking about the 2 week delay to get some from UTsource I rummaged around in my box of MMIC's and found a Minicircuits GALI-84+.

The GALI-84+ has the same package and similar IP specs as the FET.

Fitting it was a breeze and now the generator is fully working. Testing it against my Agilent 8935 test set shows its output and modulations within spec.

regards

Tim

[G0HZU]

Sorry for the late reply but I found this thread again on google and I have a few comments about the use of the GALI 84.

If you look closely at the original Agilent PCB layout/circuit with the SHF0189 there appears to be negative feedback fitted in the form of an RC circuit. It's there for various reasons and if you look back at my earlier post in this thread I also used similar feedback when I was designing with the SHF0189 many years ago.

This feedback improves the SHF0189 matching (because it isn't really a 50R gain block MMIC) and also it improves the stability and also it flattens the gain across a huge bandwidth. This is very desirable for a wideband sig gen path.
Without the feedback the gain will be very high at the low range of this amp (eg down at 250MHz) and this will mean the system ALC will have to work over a wider range depending on the sig gen frequency and the amp will be run harder here too.

You might well get away with using the GALI 84 but you must be losing a fair bit of ALC margin at low frequencies because it has lots of excess gain at low frequencies. So there is technical risk here.

IMO it's best to try finding the original part or a better equivalent that runs closer to the 8V? drain voltage of the SHF0189. The GALI 84 looks like a reasonable device to used for a quick/dirty repair but I think these sig gens deserve to be repaired bettter than this. But that's just my opinion

Note that I don't know how well Shahriar got on with the free SHF0189 samples I sent to his office at Bell because he didn't respond to me after I sent them. They could have been lost in the post for all I know...

[TSL]

G'Day,

Yes - you do have a point - and personally I'd have liked to replace it with original parts too, but I wanted it for immediate work and I've not returned to the point of making it right.

I should put its resolution on to the list of things to do !

[drago]

This signal generator deserves a proper repair. Explaining what is going on around SHF0189 is also a good material for a follow up video.

[G0HZU]

Yes, I'm not sure the GALI 84 is the right device to use because it has about 25dB gain down at 250MHz. By contrast, the original SHF0189 stage will probably be designed to only have something in the order of 12dB gain across 250-4000MHz using resistive feedback to flatten the gain to this level across this range of frequencies. Note that a feedback resistor with low self capacitance is needed in order to prevent issues with gain peaking up at several GHz.  So if you refit the SHF 0189 then I think it's best to find and refit the original resistor parts if they have been removed as the stray feedback capacitance they introduce is fairly critical wrt gain stability up at several GHz. Alternative resistors may have higher self capacitance and this will cause unwanted gain peaking up at several GHz and will compromise the overall stability of the amp.

So with 25dB gain at 250MHz in the GALI 84, the output of the vector modulator could be driving the GALI 84 output 13dB harder down at 250MHz. So that means the linearity may suffer and also it may cause issues with ALC control margins.

[nctnico]

Yes, I'm not sure the GALI 84 is the right device to use because it has about 25dB gain down at 250MHz.

I agree. When I looked for a replacement for a different HP RF generator I found the MMIC devices from Triquint had a much flatter gain/frequency response than those from Mini Circuits. With some tweaking I managed to get the frequency response to work reasonably over a wide range. However: if you can fix this generator using original parts then that is the best way to tackle this repair.

[Mosaic]

Thx for the link, I ordered some of those transistors as I have an 'unleveled' E4432B, options 1E5, UN8, UN9, UND, UNA, 100 I took a risk on for $600 based on Shah's youtube video.
I only have 1 active 4 Ghz P6217 tekprobe though for a 3Ghz TDS694C scope. Also, was that an SMA interface used to connect to the probes fed by PCB soldered RG316?

Hi Shahriar,

Great video and very timely for me.

I just picked up a 4432B ESG-D with opts 100,1E5,H99,UND for $700AU

It has the same fault indicator. I'm just about to embark on a fault finding mission to see if its the same device that failed.

FYI you can buy SIRENZA new parts from UTsource here...

http://www.utsource.net/ic-datasheet/SHF-0189-1906844.html

regards

Tim

[G0HZU]

Also, was that an SMA interface used to connect to the probes fed by PCB soldered RG316?

Be careful how you connect your test coax cables to the various locations along the RF signal path if you use a Hi Zscope. I haven't watched this video since the last time but I do recall that Shahriar had connected the Hi Z scope inputs to the signal path via lengths of coax.

This isn't a very good (as in controlled/predictable) method because it will load the signal path differently for different frequencies and this can give confusing results. I guess one can argue that this method was fine because it helped find the dead part. However, this isn't how you are supposed to use RF test gear if you want to make reliable measurements because of the way the coax and scope will load the signal path 'differently' at different frequencies. At some frequencies it will load it a lot and it depends on the length of the coax and the test frequency.

I recall that Shahriar seemed confused why the signal level had lots of ripple wrt frequency changes and this will be the reason. It appeared to me that he wasn't using the test gear correctly.

[nctnico]

I agree. When I worked on the RF amplifier of my E4421B I set it to frequencies between 10MHz and 70MHz to be on the safe side when it comes to reflections while doing measurements to get some ball-park figures.

[Mosaic]

Well, a Hi Z probe will see ripple, but I don't see how a lo Z probe (coax  with a 6dB pad for  better matching?)  will improve matters here when looking for small signals. Certainly if signal integrity was  the subject a Hi Z would not be the way to go with VSWR all over the place.

If I knew what levels to look for I could use the coax/pad approach either with the TDS694C or a spectrum analyzer. Too bad Shah didn't quantify the amplitude of the stages for us.

As I recall anything under 250Mhz used a different signal path earlier on, so going with a sub 25Mhz signal is reasonable IF the unleveled indicator remains on with low freq. sigs.
My unit hasn't arrived yet so i can't say.

From a repair approach would it be simpler to pull the several RF SMD transistors from the module in question (once the coherent feed signal is alive), then measure all the bias  points for correct bias and measure all the transistors out of circuit on an analyzer or using a basic jig  @ say 25Mhz  to 1Ghz for quantifying their function? That way all the transistors and biases are tested in one pass.

I could do a custom jig to 3Ghz and use the VNA on it for the 1dB compression as a matter of interest.

If all those are good then the ALC loop is suspect.

[nctnico]

In my generator the signals below 250MHz are made by mixing 1GHz with the generated signal but that part is before the final amplifier stage and after the auto leveling stage so testing the final amplifier at 25MHz is just as usefull as testing it at 500MHz to get a feel for the signal levels involved.

[Jay_Diddy_B]

Hi group,

I realise that this is an old thread, but I thought it was a good idea to keep all the ESG-D repair information together.

I have an E4432B ESG-D with options 1E5 and UN4. The symptoms are No output and UNLEVELED indicator when the RF is turned on. Having read this thread I was expecting to be change a dead amplifier in the signal chain.

I have confirmed that the problem is on the A9 Output board assembly, by swapping the module from a good generator.


First the board is very different to the one in the video:






I have tracked the break in the signal path to the 'Burst Modulator'.
I did this by soldering wires to the board and running the ESG at 270MHz and monitoring the signals with a scope and 10x probes.

Here is a partial block diagram for an output board (option UN3 and UN4):



And here is a picture of the 'Burst Modulator' on the PCB:




I suspect that the circuit is something like this:




I have confirmed that the right end of the diode string has an inductor to ground. I would expect the that the left hand end would be positive about 6V if the "burst modulator' is on.

If I twist the wires together, to form a small capacitor, to bypass the 'Burst Modulator', the generator seems to work fine, except the 'Burst modulation'. The UNLEVEL indication is gone and I have output.

The questions is where on the board is the 'BURST MODULATOR DRIVER'?

Thanks for your help with this challenge.

Regards,

Jay_Diddy_B

[nctnico]

I'd look in the block diagram where the burst modulation is coming from. A DMM with instant continuity is handy because you can swipe the probe along the pins quickly.

[Jay_Diddy_B]

I'd look in the block diagram where the burst modulation is coming from. A DMM with instant continuity is handy because you can swipe the probe along the pins quickly.

I followed this advice, and ended up in this area of the board:



I reflowed some of the soldering joints and the problems have gone away !!

I was lucky, I didn't need a single part, especially some of the unobtainium amplifiers.



Regards,

Jay_Diddy_B

[radar_macgyver]

@Jay_Diddy_B:

Thanks for the insights, I have an E4433B that had the same issues, it too has the '1GG3' MMICs instead of the HFETs used on Shahriar's unit. I had tracked down the loss of power to the burst modulator, as you had, and then was stuck because of the multilayer board and not being able to flip the board over (due to the output MMIC being screwed down). I noticed that the transistor where the burst mod drive signal is sourced had a dry joint (mottled gray appearance instead of shiny silver). Also, one of the diodes next to that transistor were not properly soldered (the bead wasn't formed well). After some light rework, it seems to be stable; no more UNLEVEL errors! If I were to guess, I'd say the two toroids next to the diodes were probably shielding the diodes from IR during reflow.

Thanks all!

[TK]

I got an E4422B unit with the UNLEVELED error and no output.  I followed different repair tips from this and other threads.  I tried the 20K resistor bypass fix and I got -30 dBm signal, but unit still showed the UNLEVELED error.  I also tried reflowing the area near the BURST MOD at the same time.  After I removed the 20K resistor, I got consistent -60 dBm signal with the UNLEVELED error still on the display, so the "no output" error was fixed, but signal did not go higher than -60 dBm... it goes down when I set levels lower than 0 dBm, but it does not go higher than -60 dBm even when I set any level higher than 0 dBm.  The output amp is OK, if I remove it and bypass the input to the output, the output level drops by 30-40dBm.  Then I tried to test the level before and after the BURST MOD section by soldering 2 pieces of jumper wires, one before the first diode, one after the last diode (this BURST MOD section is just before the output amplifier section).  I set the output frequency to 50MHz, level to 0 dBm and connected the scope probe (10X)... magically the generator outputs the correct level... it even goes up to near 20 dBm and no more UNLEVELED error!!!  After 2.5 GHz starts to give the UNLEVELED error again... Then I decided to remove the 2 wires, thinking the generator was almost fixed, but I get the same -60 dBm signal and UNLEVELED error... I solder the wires, no error... remove them, error... solder the wires, fixed...

It is the most bizarre "repair" I have experienced... The wires are not even connected, they are not twisted... they are just hanging in the air... Only 1 wire doesn't fix the problem, I need to have both wires soldered

[edpalmer42]

Transmit and Receive antennas?

Try soldering a capacitor between the open ends of your wires.

Ed

[syau]

I got an E4422B unit with the UNLEVELED error and no output.  I followed different repair tips from this and other threads.  I tried the 20K resistor bypass fix and I got -30 dBm signal, but unit still showed the UNLEVELED error.  I also tried reflowing the area near the BURST MOD at the same time.  After I removed the 20K resistor, I got consistent -60 dBm signal with the UNLEVELED error still on the display, so the "no output" error was fixed, but signal did not go higher than -60 dBm... it goes down when I set levels lower than 0 dBm, but it does not go higher than -60 dBm even when I set any level higher than 0 dBm.  The output amp is OK, if I remove it and bypass the input to the output, the output level drops by 30-40dBm.  Then I tried to test the level before and after the BURST MOD section by soldering 2 pieces of jumper wires, one before the first diode, one after the last diode (this BURST MOD section is just before the output amplifier section).  I set the output frequency to 50MHz, level to 0 dBm and connected the scope probe (10X)... magically the generator outputs the correct level... it even goes up to near 20 dBm and no more UNLEVELED error!!!  After 2.5 GHz starts to give the UNLEVELED error again... Then I decided to remove the 2 wires, thinking the generator was almost fixed, but I get the same -60 dBm signal and UNLEVELED error... I solder the wires, no error... remove them, error... solder the wires, fixed...

It is the most bizarre "repair" I have experienced... The wires are not even connected, they are not twisted... they are just hanging in the air... Only 1 wire doesn't fix the problem, I need to have both wires soldered

Sorry for digging out this old thread, my unit has the same issue and I am finding a way to fix it (besides soldered 2 wire)

[Bud]

I fixed mine by washing the board with isopropyl and a toothbrush. While it was also a bizarre fix, my speculation was the fault could had been caused by conductive dust or tin whiskers. The unit operates flawlessly ever since.