Author Topic: Agilent E4432b unlevel issue again (Read 7203 times)

hackwell · « **on:** May 11, 2019, 06:02:06 pm »

Hi everybody

I read many threads discussing the unlevel failure on ESG D signal generators and watched Shahriar's video about the E4433.

got 2 units off Ebay few months ago. One with the unlevel error and one with a bad display. So I thought I could end up with a working sig gen in the end. Well... nope !

The first unit has the board with the famous 1GM1 4201 and all the RF amps are custom hybrid chips. After some investigations I found a dead output amplifier and the QAM modulator 1GG7 1234 was roasted.
You can get them from UTsource for abt 150 bucks each (NOS)
So after some checkings and measurements I replaced both of them. Got a good signal for about 30 seconds then the QAM modulator died horribly. I bypassed the whole stage from the input to the filter with a jumper wire, losing the benefit of QAM in the process.
It worked quite well for a couple of weeks , in spite I couldn't reassemble the top shielding because of the bypass wire.
Then it went back to unlevel condition again , this time because of the burst modulator!
So I overrid this by soldering 2 wires before and after the PIN diodes as mentionned in the thread about Shahriar's video. I'm wondering what wil be the next failure !

The second E4432b was sold with a bad display, I was quite confident the ouptut board was OK. I easily fixed it, a bad transistor in the VLCD circuit. and guess what? unlevel as well...

After some measurements I foud 3 (yes 3!!!) dead RF amplifiers

All of them except the one that was bad in Shahriar's video

As I had some NBB310 in stock I tried his workaround and it worked for 1 minute before the one before the LPFs died...

Now I have 2 boat anchors!

I ordered a bunch of SHF-1089 but this time I lack confidence dunno why.

In the meanwhile I'll try some reverse engineering.

Does anybody managed to find the cause of the burst modulator failure ?

Or some info about the QAM modulator?

Or traced back the RF amplifiers bias circuits ? They all issued by Q920 but there's additional circuitry in between...

hackwell · « **Reply #1 on:** May 12, 2019, 01:32:40 pm »

quick update :

as I wrote befor , my attempt to replace the bad RF amplifiers by NBB 310 failed. So I tried with ADL5602ARKZ and I have a signal. The unlevel error is gone , but the signal is 8dB below what it should and a bit noisy.
I left one of the NBB310 , gonna replace it as well and see the result.

The ADL5602 provide 20dB gain with a +5v bias and not 8v , so I have to find a workaroud to get 5V from the 8V bias without using bulge wires in order to put the top shielding back into place (maybe the reason why the signal is so noisy BTW)
Maybe a zener diode could do the trick ?

EDIT : the IQ calibration fails at 14% with error 517

hackwell · « **Reply #2 on:** May 23, 2019, 07:33:42 pm »

here are some news:

I replaced the 5 amplifiers by SHF0189Z I finally received and I can call it a success. The output is -0.87 dB below , no more unlevel error , the tone is clean above 250MHz. AM and FM modulation work fine.

I still get the IQ calibration error 517. Gotta investigate this.

Now calibration is the next step. As I used the python script to turn it into a E4433B via GPIB it now can go to 4GHz but it lost its calibration data as expected. (made a backup of U106 and U107 beforehand. If someone is interested I can share the files)

YetAnotherTechie · « **Reply #3 on:** May 23, 2019, 10:24:44 pm »

Hello,
Can you backup the entire memory with gpib? what about cal constants?
Thank you

hackwell · « **Reply #4 on:** May 25, 2019, 12:44:55 pm »

Hi ,
It must be possible as the firmware update from Keysight works by GPIB but I don't know the needed commands. And I think we have to kwon the address of the firmware in memory.
A bit of reverse ingeeniring could do the trick. It's on my to do list.
I did the backup using my TL866 programmer.

TheSteve · « **Reply #5 on:** May 25, 2019, 12:57:33 pm »

I spent some time experimenting with my E4431B last night. I tried various changes via GPIB. I found it was possible to change the max frequency from 2 GHz to 4 GHz without changing the model number or losing cal data. It will boot without error. Measuring the output the unit is operating fine, and in spec up to 2 GHz. From 2 GHz to 4 GHz the output level drops off from the displayed value. At 3 GHz it is 1.6 dB low, at 4 GHz it is 3 dB low. Still an easy mod to double the freq range.

It probably isn't worth trying to perform a calibration on it.

Does anyone know the command/memory location to alter the start frequency?

hackwell · « **Reply #6 on:** May 25, 2019, 01:41:19 pm »

Ok that's interesting !
So you just changed the max frequency using these commands ?

inst.write("SERV:PRODUCTION:CAL 165,2,4000000000.000000;")
inst.write("SERV:PRODUCTION:CAL 165,21,4000.0899047852;")

So it still reports as a ESG-D 2000 in your case (not sure , mine was ESG-D3000B) but goes up to 4GHz ?

Mine is -4dB to 1.75Ghz then -7dB to 2Ghz then back to -4 -5 for the rest of the band. I'm not sure if this is because of the amps replacement.
No more unlevel error anyway

TheSteve · « **Reply #7 on:** May 26, 2019, 05:58:44 am »

I ran these commands to change only the frequency range to 4 GHz:

SERV:PRODUCTION:CAL:BEGIN;
SERV:PRODUCTION:CAL 165,2,4000000000.000000;
SERV:PRODUCTION:CAL:END;
SERV:PRODUCTION:CAL:STORE 165;
SERV:PRODUCTION:CAL:PUP;

My units reports as "Hewlett-Packard, ESG-D2000B, US38440280, B.03.86"

hackwell · « **Reply #8 on:** May 26, 2019, 03:41:53 pm »

Thanks a lot ! I sould have tried that before ruining the calibration data...
I'll restore the unit back to original by reflashing U106 and U107 then try your script.

I had a look inside the upgrade utility and found somme commands :

Code: [Select]

:DIAG:BBG4:FMWR:DOWN 
:DIAG:BBG3:FMWR:DOWN 
:DIAG:BBG1:FMWR:DOWN
:MMEM:DATA "/SYS/CONFIG_INFO.TXT",
:SENS:DEM:FIRM:CDLO:BLOC 
:DM:BERT:FIRM:FMEM:BLOC 
:DM:BERT:FIRM:BLOC 
SOUR:RAD:FLEX:FIRM:BLOC
:DM:ARB:FIRM:BLOC 
:SYST:FWLD
:MEM:DEL:ALL
:SYST:COMM:GPIB:ADDR

I could be a good starting point to access the firmware and calibration data

TheSteve · « **Reply #9 on:** May 27, 2019, 09:37:39 pm »

If you haven't tried to perform a cal and restore the original config the cal data should still be there. Likely easier then pulling the chips to reflash.
I made quite a few changes while experimenting and my cal data is fine.
I still want to know the location of the lower end frequency. I believe they should operate down to 9 kHz.
I also verified if I was to calibrate mine it generates more then the minimum output max spec to 4 GHz.

hackwell · « **Reply #10 on:** May 30, 2019, 11:02:41 am »

I lost my cal data because I changed the model from E4432 to E4433 via GPIB so now the flatness is out of range around 1.8GHz.
That's why I need to reflash the memories. Not a big deal , but I'm still left with my initial issue. It seems the SHF0189 have less gain than the original amplifiers from Agilent so the signal is 5dBm lower thant it should be.
I had to replace 4 of them to get rid of the unlevel error ! 2 of them were utterly shorted to ground and the 2 others have turned into attenuators...
I had a look at the calibration procedure and it's quite tricky. It requires a bunch of test equipment I dont have , all connected to the computer via HPIB and it sort of auto calibrate by retrieving the measured values from the measuring gear and write the calibration data straight into the generator without telling you what's it's really doing

pretty well secured process. Plus some of the test gear needed have to be the exact model mentioned in the service manual. Well...

Has somebody already done that?

But I discovered some intersting things:
-The IQ calibration fails with a 517 error with no further information. But the modulation works fine so it must be relate to the amplifier gain lack. But you can perform a manual calibration and specify a range of frequencies. Tried this with various values and it only fails from 9KHz to 1.5 Ghz and passes from 1.5GHz to 4GHz. I still don't know why tho.

-There is a 93C66 EEPROM on the output board. Apparently it holds some configuation data abt the board model. As I said I have 2 models of output boards. One with the 1GG7 4235 amplifiers with a bypassed IQ modulator and one with the SHF0189.
The synthesizer board doesn't sweep at the same frequency range depending on the output board!! one from 1.5GHz to 2 GHz at -10dBm and the other from 1GHz to 1.5GHz then then jumps to over 3GHz at 0dBm ( I can't see that because my SA only goes up to 2.9GHz ) weird enough!
I wonder what will happen if I swap those 2 EEPROMS ? I have to try this.

-Played a bit with the LF generator and it's a pretty cool stuff. Nearly a full loaded func gen with many waveform choices and a frequency range from DC to 10KHz.

Quote from: TheSteve on May 27, 2019, 09:37:39 pm

I still want to know the location of the lower end frequency. I believe they should operate down to 9 kHz.

I'm not sure it would , as the SHF0189 have poor low freq performance. I'm still trying to hack the firmware tho. What about using the LF gen?

hackwell · « **Reply #11 on:** May 30, 2019, 02:37:43 pm »

I just reflashed U107 and U106 with their original contents. My calibration data are back. The unit complained about a corrupted RAM content on the first boot but everything OK after rebooting.

The calibration data seem to be stored inside U107 at addresses from $40 to $8C and are blanked out to $FF if you change the model number via GPIB.

U106 showed many differences at various addresses too , I'm trying to figure out what they're related to.
here are some differences :

Code: [Select]

address			before				after

$40			15 38 40 70 00			01 40 40 74 10
$BC			FF FF 2A C8 40 70		FF FF 2E B0 50 74
$DA			55 48 15 38 7F 80		41 34 11 40 6F 84
$57A			7F 9C				F6 A7
£581			28 35 76			28 36 7C
$5DB			FF FF 7F B4 35 5E		FF FF F6 8F 36 54
$5FC			7F B4 4A EA			F6 8F C0 DB
$602			35 5E				36 54
$43BE			83 C2				C2 D4
$43C5			33 FF FF			34 00 32
$43D1			E6 5A 0B C0			ED CD 65 00
$4410			40 A7 70 2E 07 FF FF FF		40 AF 40 2E 08 00 00 60
$4468			40 A7 70 2E 07 FF FF FF		40 AF 40 2E 08 00 00 60

and many other locations were affected too.

TheSteve · « **Reply #12 on:** May 30, 2019, 05:05:08 pm »

Colby let me know that the low frequency limit is at CAL 165,1. I lowered mine to 9 kHz but it had no output that low. It had output from 30 kHz and above but it was not a clean usable signal. I wouldn't go below 50 kHz and even that might be suspect and needs further verification.

hackwell · « **Reply #13 on:** June 04, 2019, 04:40:00 pm »

It's not surprising. As I said the RF amplifiers along the signal path have bad performances at low frequencies.

I tried many things to get the ouptut level of my set withing specs to no avail.
So the next step is to get the calibration software running and see what happens.
Not sure at all it would work with my National Instruments USB GPIB instrface tho

Ordinaryman1971 · « **Reply #14 on:** June 12, 2019, 05:22:58 am »

I have two questions about ESG signal generators.
First one is...:
I have a 1GHz version of ESG... E4400... which was working fine.
I've bought a second one 2GHz version with bad screen. I've decided to test it with the good screen from 1GHz one and installed it in the 2GHz one... the screen stopped working.
Now I have two signal generators with not working screens. The screens are both fully turned on, all the pixels are black.
I can hear either one of them turning on and clicking relays but I can't display anything on either screen.
Any experience with problem like this. Not sure how to approach it.
Did I burn the second screen and how did I do it.....
Is there any way to monitor the output out of signal gen and see if there is anything out of it.

Second question is:
I just bought another ESG yesterday... 4GHz version and it's turning on, the display is working fine but I get two errors;
513, 1GHz oscillator unlocked
521, YO loop unlocked.
I've measured the output of the refference and it's 9.9981kHz... well 20Hz out, the signal is peaking at 2.7V
I've connected external refference... (oxco) and I'm getting the same errors.
Where is the 1GHz refference coming from... is it refference board or some other place.
Is it a refference board that's bad?

If I will be able to fix either.. all of them, wanted to give it a chance and possibly hack it to higher frequency.

Kean · « **Reply #15 on:** June 12, 2019, 12:01:05 pm »

Quote from: Ordinaryman1971 on June 12, 2019, 05:22:58 am

The screens are both fully turned on, all the pixels are black.

Sounds like an LCD contrast control issue.
Maybe watching Dave struggle with something similar will help you diagnose

Ordinaryman1971 · « **Reply #16 on:** June 12, 2019, 01:43:18 pm »

In my case they are both exact fit for this application. When the screen was installed back in the original frame it didn’t work either so... something went bad.
The service manual says something about the reset after the screen replacement, I just wonder what do they mean by that

Kean · « **Reply #17 on:** June 12, 2019, 01:59:31 pm »

I don't have an ESG, so I'm not familiar enough with the internals to give much guidance.
LCD's generally go dark because they are either not getting initialised properly, or the contrast settng is wrong.
Both are typically cabling/connection issues - especially if you're seeing other signs of life from the unit.

roxbox · « **Reply #18 on:** June 02, 2021, 05:22:54 pm »

Quote from: TheSteve on May 26, 2019, 05:58:44 am

I ran these commands to change only the frequency range to 4 GHz:

SERV:PRODUCTION:CAL:BEGIN;
SERV:PRODUCTION:CAL 165,2,4000000000.000000;
SERV:PRODUCTION:CAL:END;
SERV:PRODUCTION:CAL:STORE 165;
SERV:PRODUCTION:CAL:PUP;

My units reports as "Hewlett-Packard, ESG-D2000B, US38440280, B.03.86"

read
:serv:prod:cal? 165,2;

set E4436B start and end freq
:serv:prod:cal:begin;
:serv:prod:cal 165,1,25000.000000;
:serv:prod:cal 165,2,4000000000.000000;
:serv:prod:cal:end;
:serv:prod:cal:store 165;
:serv:prod:cal:pup;

set freq cal table.

the last freq has to be put in two locations to mark end of freq list.

:serv:prod:cal:begin;
:serv:production:cal 65,0,25000;
:serv:production:cal 65,1,50000;
...
:serv:production:cal 65,115,3960100000;
:serv:production:cal 65,116,3980100000;
:serv:production:cal 65,117,4000000000;
:serv:production:cal 65,118,4000000000;
:serv:production:cal:end;
:serv:production:cal:store 65;
:Serv:production:cal:pup;

Then I ran TME Power Level Accuracy (Adjustment)

seems that when it switches from the power meter to the specan that it starts at freq 100000 and not 25000!

I will look for another setting related to the start freq.

vladsol · « **Reply #19 on:** July 23, 2021, 07:08:24 am »

Could you share the dumps of U106 and U107?

TomT · « **Reply #20 on:** September 23, 2021, 10:47:12 pm »

Hi Ordinaryman,

I used to repair the RF boards on the ESG series to component level, and have lots of experience with the reference boards. The only repairs to the reference boards that were ever needed was to replace the Reference Oscillator, the 1GHz oscillator transistor, one of the VCO voltage drive transistors (only used on some boards), Ext. Ref. Input transistors, or diodes. I personally own an E4432B ESG-D, and a recently acquired E4437B ESG-DP. There was a problem with unlevel on the E4432B, but with the knowledge of repairs of output boards I did at work, I purchased a few of the HP proprietary RF amps before they became unavailable.

First, knowing the full model number, and options would be very helpful in diagnosing this problem. If you can supply the Reference Board part number, even better.

The 1GHz VCO, is located on the Reference board, at the opposite edge of the board from the connectors It is easily identifiable by the hard line coax soldered to most boards, and is the fixed LC for the oscillator. Others have a small VCO module. The 1GHz is divided down to 10MHz, compared to the divided down, frequency and phase compared to the 10MHz Reference Oscillator, and low pass filtered to create an error voltage, which is then applied to the voltage control of the 1GHz VCO. If the Reference Oscillator is too far out of tolerance, the 1GHz oscillator is unable to lock. Likewise, if the 1GHz free run frequency is too far off, it won't lock.
The YO loop may lock once the 1GHz is locked. It’s just best to start with the reference board first, as it is the source for the entire generator.

As for the Ext. Ref. not locking the 1GHz VCO, the best place to start is at the Ext. Ref. input to the board. The input circuitry may have been compromised from an improper input supplied, which I have repaired in the past. It’s also possible that phase detector circuit has issues, but I did not experience any problems in this area when I was working on them.

As for the display, play with the contrast controls. That was one problem that messed with me and others that I worked with, especially after replacing the display. Also, if I recall, there are two different part numbers for the displays, and I don’t recall if they are interchangeable. I believe that one of the part numbers was used in newer units as the older display became unavailable, or a better or cheaper display was found.

If you plan to attempt RF board repairs yourself, when reassembling the clam shell, be careful about over-tightening the screws, as they are stainless steel screwing into aluminum. Start by starting all of the screws before tightening, and be careful not to cross-thread them. Then start tightening from the center screws, and work your way out to the edges.

grendle373 · « **Reply #21 on:** November 11, 2021, 03:46:06 pm »

Great info!

Question, I have an e4432A, I repaired a display issue ( Bad LCD, Blown Q601, Q602), Display is fine, everything else works correctly and is in spec. , except I get a "Fail 601 power search" when running the IQ mod cal.

I replaced the output board (usually the culprit) with a known good one ( passes IQ cal in a similiar unit), and still get the failure, Also replaced the Refrence board with a known good one, no change. Unit has option UND (BB generator). Output level drops about 11 db when the IQ mod is switched on. Scratching my head on this one......

mbielman · « **Reply #22 on:** March 30, 2023, 09:43:26 pm »

I know this is an old post but are these from the CAL software package?

SERV:PRODUCTION:CAL:BEGIN;
SERV:PRODUCTION:CAL 165,2,4000000000.000000;
SERV:PRODUCTION:CAL:END;
SERV:PRODUCTION:CAL:STORE 165;
SERV:PRODUCTION:CAL:PUP;

They are not in the SCPI docs.

Thanks!

vaualbus · « **Reply #23 on:** April 02, 2023, 12:59:38 pm »

So basically this commands invalidate the cal data If I got that part correct from previous comments? I too recently purchase a 2Ghz version and maybe I wanna "upgrade" to 3 or 4ghz. Also are there any way to dump the ROMs from the serial/GPIB interface? It's also strange that we don't have a debug serial port somewhere in the intruments!

Speaking about hacking, as any one ever tried to hack the software options to add them to the instrument?

mbielman · « **Reply #24 on:** April 03, 2023, 02:07:05 am »

What I'm trying to ask is 'Where do these commands come from? How do you know about them? Are they documented?'

Maybe they are from the software .log files? Or the test window itself?
Assume there is NO docs available for these.

(what does the PUP command do???)


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: Agilent E4432b unlevel issue again (Read 7203 times)

Share me