Author Topic: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator  (Read 20262 times)

0 Members and 1 Guest are viewing this topic.

Offline Hugoneus

  • Frequent Contributor
  • **
  • Posts: 875
  • Country: us
    • The Signal Path Video Blog
In this episode Shahriar repairs an Agilent E4433B ESG-D Synthesized Signal Generator. This 4GHz unit shows the "UNLEVEL" error for all frequencies and at all output power levels. After a close look at the system block diagram the fault is traced to the main Output RF board.

The Output RF board is equipped with various amplifiers, electronic attenuators, I/Q mixers, heterodyne paths and final PA. By using soldered SMA cables to various points in the signal path, the fault is traced to an internal amplifier chain which is likely made of a MESFET or JFET transistor. Since this part is obsolete, an RFMD GaAs hybrid amplifier is used in its place. The biasing network and matching networks need to be modified to accommodate this change. The repaired unit is then verified for functionality and performance by measuring a QAM constellation output signal.

Watch the video here: [1 Hour]
http://youtu.be/RSCGAY5z_44

More videos at The Signal Path:
http://www.TheSignalPath.com
Shahriar Shahramian
The Signal Path Video Blog
www.TheSignalPath.com
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #1 on: December 22, 2014, 03:41:51 pm »
Wow, Shahriar, that is good timing :)
I've just bought one of these here in the UK for £1k. I'm going to sit down with some mince pies and watch your video :)

Can I ask what options yours has? Mine seems to have an issue with the options fitted to it. It arrived  a couple of days ago in a bit of a state in terms of the way the HW options were (not correctly) installed on the user interface. I've spent the last day or so trying to reinstall the UND, UN8, H99, 1E5  and 100 options on the front panel interface in order to activate them.

It detects the option hardware as being there but it gives a clash if I try and install both UND and UN8 together. I don't suppose you know if UND overrides UN8? The various UND/UN8 installer guide documents from Agilent don't make this clear.

I can get the UND dual arb generator to work OK and the various features of UN8 seem to be there without installing UN8. But it seems odd to see UN8 as not selected/installed on the UI.

Does yours have both UN8 and UND?


 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 7242
  • Country: us
    • SiliconValleyGarage
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #2 on: December 22, 2014, 03:44:39 pm »
this series of machines frequently suffers from this problem.

it's almost always the unlevel error.

i have the adjustment software for these machines. you need an agilent gpib adapter and the agilent i/o library. then you can run the 'falcon' software and set the machine in test mode and do all kinds of self-cal procedures.
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #3 on: December 22, 2014, 04:54:40 pm »
I'm currently watching the video and I have a comment that might clarify things at 26:50.

The reason the two parallel 100R resistors are there is to define the output impedance of the generator at 50 ohms. If you look closely you will see that the ALC detector tap point is immediately before these two resistors. It sits snugly before them.

Because the ALC detector is part of a feedback loop then the impedance at this detector tap point will be close to that of a voltage source. i.e. the impedance at this point (within the bandwidth of the ALC feedback system) will be extremely low (close to zero?)  because of the nature of the feedback. So you have to fit a series 50R resistance after this point to set the source impedance of the generator to 50 ohm.

This ALC + series 50R resistor technique has been commonly used on signal generators for about 50 years (possibly more?) and I think it is used in all of the RF signal generators I have here from various manufacturers :)

« Last Edit: December 22, 2014, 04:56:13 pm by G0HZU »
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #4 on: December 22, 2014, 05:18:13 pm »
OK I'm a bit guilty of skimming the video (I ate the mince pies too quickly) but I think I can advise you about the amplifer device that failed in that generator.

It looks to me like an old school HFET. i.e. a heterostructure FET.

I have used the old Stanford/Sirenza/RFMD SHF-0189 (and the 0289 and 0589) many times as a driver amplifier for broadband systems and the 0189 has an H1 logo on it. So it looks like a good match. However the gain of the device is quite high at LF and I always used feedback to flatten the response.

I can see if I can find some at work if you would like one? I have the 289 and 589 devices here and I probably have a few 0189s but they will be fitted to a dev board. I think the 289 and 589 are now obsolete but maybe the 0189 is still available.

One thing to note with the 0189 is that it needs a negative bias voltage. Sometimes this is provided with an active bias circuit. If this active bias circuit fails then the device will cook itself without a negative bias at the gate pin. They are quite fragile and the specs on these devices got derated on the datasheet over the years. eg there was a change to reduce the recommended Vdd to 7V from 8V on some devices in this series. Also, the MTBF is heavily reliant on a decent layout with good thermal paths (lots of via holes). Without this the device will not be reliable in the long term.

Can you take an ultra closeup of the H1 device? I'll have a look if I have a SHF-0189 here to compare.
« Last Edit: December 22, 2014, 05:25:35 pm by G0HZU »
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #5 on: December 22, 2014, 05:36:04 pm »
I had a rummage and I found an old X model amplifier I designed about 12 years ago with SHF-0189 drivers.

See below for a shaky image.

I followed all the design guidelines in the datasheet when designing this driver so you can see the heatsink screw and the numerous via holes. There is also a copper heat spreader directly under the PCB to keep the junction temperature of the device low. But this design of mine had to work in a very hot environmment and it was near a 10W broadband power amplifer as well...

See the image below. It looks very similar and you can see that I had to use drain/gate feedback to flatten the gain over a wide bandwidth. I expect that HP/Agilent will have done similar.

See also the link to a SHF-0189 datasheet. However, I suspect that the H1 devices in your E4433B date back to before the days when Stanford/Sirenza (the original device manufacturer) were bought by RFMD.

http://media.digikey.com/pdf/Data%20Sheets/RFMD%20PDFs/SHF-0189(Z).pdf
« Last Edit: December 22, 2014, 05:48:54 pm by G0HZU »
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #6 on: December 22, 2014, 06:07:11 pm »
Hope nobody minds me posting up all this stuff but it would be worth looking at the HP/Agilent PCB to see if there is an active bias network on the back of the board.

I have designed active bias networks for these devices and I base the design on this old app note from Sirenza. By using active bias you lose a bit on P1dB and OIP3 but it is a good tradeoff for mass manufacturing because you don't have to manually bias the device during production test.

http://application-notes.digchip.com/147/147-47819.pdf

Maybe your SHF-0189 got toasted because the active bias network failed (if it has one). See if you can see a similar active bias network on that Output Board. It should have a negative bias voltage at the gate. If the negative bias is not there then the device will draw several hundred mA and cook itself pretty quickly...

« Last Edit: December 22, 2014, 06:15:18 pm by G0HZU »
 

Offline RRobot

  • Regular Contributor
  • *
  • Posts: 84
  • Country: ca
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #7 on: December 22, 2014, 06:26:30 pm »
That was a fantastic repair job. Thanks for posting it.
 

Online PA0PBZ

  • Super Contributor
  • ***
  • Posts: 4159
  • Country: nl
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #8 on: December 22, 2014, 06:37:41 pm »
It detects the option hardware as being there but it gives a clash if I try and install both UND and UN8 together. I don't suppose you know if UND overrides UN8? The various UND/UN8 installer guide documents from Agilent don't make this clear.

I can get the UND dual arb generator to work OK and the various features of UN8 seem to be there without installing UN8. But it seems odd to see UN8 as not selected/installed on the UI.

Does yours have both UN8 and UND?

I have one with both options and I have no problem with it.

1E5    High stability timebase
UN5   Multicarrier multichannel CDMA
UN7   BER Analyzer
UN8   I/Q Generator
UN9   8MB ram for I/Q Generator
UND   Dual ARB

Keyboard error: Press F1 to continue.
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #9 on: December 22, 2014, 07:36:28 pm »
Hi PA0PBZ. I've sent you a PM about the options issue I'm experiencing  :)


Shahriar,
Getting back to the repair, the SHF-0189 was designed to provide very low distortion when driven at a level well backed off from P1dB. I'd expect it to outperform the NBB-310 in this respect because it is designed to give a much higher OIP3.  So I do think it is worth trying to source an SHF-0189 if this is indeed the correct part.

At the very least I can send you a used one but I might be able to find some new 'old' stock gathering dust in one of my storage areas at my place of work. They are cheap devices so you can have one for the cost of postage.

I'd be wary of buying one on ebay because the various chinese copies/fakes will be out there now that this range of devices is obsolete. If I send you one it will at least be a genuine SHF-0189 :)

 

Offline Hugoneus

  • Frequent Contributor
  • **
  • Posts: 875
  • Country: us
    • The Signal Path Video Blog
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #10 on: December 22, 2014, 07:44:33 pm »
Dear G0HZU,

Thanks for all the great information you provided. This is exactly the kind of discussion I always hope to arise from the videos. Your point about the output impedance of the final amplifier is much appreciated. It skipped my mind at the time of the teardown to take the ALC feedback impact on the output impedance into consideration. Thanks you for that.

I agree with you that the replacement part probably has inferior OIP3 compared to the original part and that is why I tried to measure the a constellation EVM at the end of the video to make sure the result is still acceptable. I can try generating two tones also to actually measure the IM3 levels at various output levels.

A replacement part would be much appreciated if you have some. I can purchase a few from you and pay for the shipping.

As for the other suggestion for the Falcon software, it is a great debugging tool to monitor all the analog bus voltages. I have used it before but this time I thought it would be more interesting for the viewers to do a teardown. :)

I'll look at the options for you as soon as I get home.

Thank you all for your contributions.
Shahriar Shahramian
The Signal Path Video Blog
www.TheSignalPath.com
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #11 on: December 22, 2014, 08:00:47 pm »
One thing to note: If you still have the original H1 part then it might be worth keeping it safe.

If the external active bias circuit has failed and the drain voltage at the device falls to 0V as in your video then maybe HP/Agilent designed the active bias to not cook the device if the active bias wasn't there. So the HFET might be OK.

So maybe the only faulty part is in the active bias circuit. But I'll see if I can find some new ones at work. It won't be until Jan 5th though because I'm on xmas holiday :)

 

Offline Hugoneus

  • Frequent Contributor
  • **
  • Posts: 875
  • Country: us
    • The Signal Path Video Blog
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #12 on: December 22, 2014, 08:02:02 pm »
One thing to note: If you still have the original H1 part then it might be worth keeping it safe.

If the external active bias circuit has failed and the drain voltage at the device falls to 0V as in your video then maybe HP/Agilent designed the active bias to not cook the device if the active bias wasn't there. So the HFET might be OK.

So maybe the only faulty part is in the active bias circuit. But I'll see if I can find some new ones at work. It won't be until Jan 5th though because I'm on xmas holiday :)

Great! I still have the old part... Unless my cat ate it because I left it on the table.

Enjoy your holidays!
Shahriar Shahramian
The Signal Path Video Blog
www.TheSignalPath.com
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #13 on: December 22, 2014, 08:35:56 pm »
Thanks :)

I guess you could look at the other devices on that board to see how they are biased and how the bias behaves when healthy.

On my design I included an interlock that meant that the 8V feed would get switched off if the negative bias failed. This prevented damage to the HFET. Maybe HP designed an overcurrent trip on the 8V feed to do a similar thing?

I also had to make sure that the circuit behaved itself during power up and power down without stressing the SHF-0189 devices during this brief phase. In all I had to do quite a complicated design to keep the devices reliable electrically and thermally but it was worth it for the RF performance these devices deliver.
 

Offline HighVoltage

  • Super Contributor
  • ***
  • Posts: 4303
  • Country: de
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #14 on: December 22, 2014, 08:53:14 pm »
Really great repair video again.

The way you are guiding us through the steps is very easy to understand, although I have not much HF experience at all.
Thank you!
There are 3 kinds of people in this world, those who can count and those who can not.
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 7242
  • Country: us
    • SiliconValleyGarage
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #15 on: December 22, 2014, 09:51:48 pm »
who , knows , maybe after traveling through the cats digestive system that part may work again..

if only it were a cat-whisker diode ...  :-DD
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #16 on: December 22, 2014, 10:04:35 pm »
I managed to dig out a really old Stanford Microdevices datasheet for this SHF-0189 device and it shows the Id vs Vgs vs Vds.
Stanford were the company that made this device many years ago.

You can see why the device will get fatally stressed if it loses the negative bias and Vgs falls to 0V and Vds is 8V. It will draw about 300mA from 8V and this will cook it. It will normally run at about Vgs = -1.3V for 100mA Id although this will vary from device to device.

This is why it is wise to design in a cutout/interlock circuit for the 8V feed and I suspect that HP will have done this. It's a shame that they don't produce a schematic for this sig gen...
 

Offline Hugoneus

  • Frequent Contributor
  • **
  • Posts: 875
  • Country: us
    • The Signal Path Video Blog
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #17 on: December 22, 2014, 11:36:32 pm »

if only it were a cat-whisker diode ...  :-DD

Brilliant.
Shahriar Shahramian
The Signal Path Video Blog
www.TheSignalPath.com
 

Offline Hugoneus

  • Frequent Contributor
  • **
  • Posts: 875
  • Country: us
    • The Signal Path Video Blog
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #18 on: December 22, 2014, 11:41:35 pm »
...
You can see why the device will get fatally stressed if it loses the negative bias and Vgs falls to 0V and Vds is 8V. It will draw about 300mA from 8V and this will cook it. It will normally run at about Vgs = -1.3V for 100mA Id although this will vary from device to device.
...

Thanks for the info. It is possible that the biasing circuit is what has died. The new device does not need the input negative biasing scheme. I can measure the IV curve of the device I took out and find out if it still works!
Shahriar Shahramian
The Signal Path Video Blog
www.TheSignalPath.com
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #19 on: December 23, 2014, 12:22:49 am »
Probably the simplest check is to see if the gate has shorted to the drain with a DMM. Normally the device measures just a few ohms drain to source on a DMM but the gate to drain should he high impedance.

If the device fails and the drain goes low Z to the gate then the device will override the biasing (active or passive) when in circuit and put a positive bias on the gate and it will probably cause the 8V feed to trip out a protection circuit as the device will take a lot of current in this faulty state.

Maybe this is why you saw 0V on the drain and the gate? Because it tripped something? But I'm just guessing really...
 

Offline Hugoneus

  • Frequent Contributor
  • **
  • Posts: 875
  • Country: us
    • The Signal Path Video Blog
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #20 on: December 23, 2014, 06:36:02 am »
Probably the simplest check is to see if the gate has shorted to the drain with a DMM. Normally the device measures just a few ohms drain to source on a DMM but the gate to drain should he high impedance.

If the device fails and the drain goes low Z to the gate then the device will override the biasing (active or passive) when in circuit and put a positive bias on the gate and it will probably cause the 8V feed to trip out a protection circuit as the device will take a lot of current in this faulty state.

Maybe this is why you saw 0V on the drain and the gate? Because it tripped something? But I'm just guessing really...

Dammit! I can't find the part! I knew my cat would do something to it...  |O
Shahriar Shahramian
The Signal Path Video Blog
www.TheSignalPath.com
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 15169
  • Country: za
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #21 on: December 23, 2014, 07:07:46 am »
Look tomorrow in the litter box, it will be there somewhere............

But in any case you did in testing see a negative bias then nothing, so likely it was short circuit eventually. As is you probably could leave the MMIC in there as it does provide enough gain, and for most applications does the job. Might not be perfect at edge cases, but for everything else it works.
 

Offline bktemp

  • Super Contributor
  • ***
  • Posts: 1623
  • Country: de
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #22 on: December 23, 2014, 07:58:31 am »
Great video! I always enjoy the teardown/repair of expensive equipment.

A bit off topic, but I wonder how the IQ modulator works. A typical IQ modulator needs a 90° phase shifted signal, but how is this done in the signal generator over such a wide frequency range?
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #23 on: December 23, 2014, 12:06:29 pm »
I watched the video more closely this morning and I think I can explain why the device biasing seemed confusing during the DMM probing.

At 42:53 you actually measured the gate bias at -2.1V (not the drain). The drain was close to 0V. So I think you accidentally got these measurements back to front and this caused the confusion. So I think the active bias was valiantly trying to bias off the (damaged,shorted) SHF-0189 device at 42:53, and maybe -2.1V is the most negative the bias generator can go before it hits its circuit limit and gives up. It was never going to succeed because I now think the HFET is indeed faulty.

If we then fast forward to 45:50 when you remeasure the gate voltage we can see that it has risen from -2.1V to +0.2V.

I think this is expected because you removed the bias choke in the drain and so there is no drain current. So the active bias circuit will raise the gate voltage to the other end of its range to 'try' to get the target drain bias current of maybe 100mA. Obviously it is going to fail because there is no DC path to the drain because the bias choke is removed and it probably hits its endstop at +0.2V gate bias voltage.

So I think the active bias circuit may be healthy and it probably has an operating range of -2.1V to +0.2V at the gate.

This should be more than enough range to bias a healthy HFET to the target bias current (probably 100mA) with enough range to cover the spread in device charactersitics across a large batch of SHF-0189s.

The other thing I noticed was that the final HFET device at the sig gen output looks like it may have already been changed in the past as it looks like it has been hand soldered. See 28:09 in the video because Q2600 (H1) looks like it has been replaced/resoldered in the past and it also doesn't have the extra red dot on it like the other H1 devices.

So maybe these devices fail quite often in these sig gens. It's a bit of a concern for me because I now have two of these 4GHz E4433 sig gens here at home. I bought the first one earlier this year and bought another one a few days ago. They can be bought quite cheaply if you are patient. I bought the first one for £780 and the second for £1000. Both have lots of options fitted and they probably cost over £20k new.
« Last Edit: December 23, 2014, 12:21:27 pm by G0HZU »
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 15169
  • Country: za
Re: Video Teardown and Repair of an Agilent E4433B ESG-D Signal Generator
« Reply #24 on: December 23, 2014, 02:15:42 pm »
If that then at least we know you can replace some with MMIC devices with little drop in performance.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf