Aeroflex IFR-2398 / calibration / repair help

[G0HZU]

Note: It's risky to make (fairly precise) conclusions about stage gain when probing 'in circuit' especially up at 3.5GHz.

If you are using the 470R probe then you are measuring voltage gain so you can get confusing results if the amp has different source and load impedances. You could try probing/comparing at the low impedance tap points of each of the little printed BPFs in the first IF. But you can probably still get confusing results when probing here.

Have you ruled out the possibility of gain loss due to instability somewhere?

There do appear to be lots of little bits of (what looks like) RF absorber stuck on the signal path in various places. I assume these are there to suck up unwanted LO harmonics at extremely high frequencies in order to try and reduce the amplitude of internal spurious terms caused by the undesired meeting and mixing of 1st and second LO harmonics at extremely high frequencies. But they might also be there to improve stability in some cases. Have any of them fallen off? Are there any more of them stuck inside the lid?

[andy_silicon]

The cover is full of shielding and black RF absorbers.

Also the 1st LO is 3.4-6.2GHz 

I made an improved probe which is what allowed me to do *any* probing.
( the best / most reliable results were achieved by probing the BPF strips )

Nope - I'm not 100% sure that instability somewhere isn't causing the loss.

[G0HZU]

In case you haven't already done this,  I'd strongly suggest you take a decent photo of the inside of the lid in case any of the absorber bits fall off the lid during the faultfinding

[KJDS]

The cover is full of shielding and black RF absorbers.

Also the 1st LO is 3.4-6.2GHz 

I made an improved probe which is what allowed me to do *any* probing.
( the best / most reliable results were achieved by probing the BPF strips )

Nope - I'm not 100% sure that instability somewhere isn't causing the loss.

The 470R resistor is great in a 50 ohm system to give a 20dB ish loss.

However as you work your way along those filter resonators you get further and further away from 50 ohms, and at the non grounded end then you'll see much higher voltages. I'd disconnect the bias to each amp in turn and measure the level with the 40MHz cal signal again. If it drops further then you know the amp is ok. If it doesn't change much then that amp is likely to be the issue.

[G0HZU]

about 30-40dB down at 2.7G compared to 40Mhz

I think you need to focus on this issue rather than poke around the amps in the first IF because the fault above is most likely to be something wrong in the first mixer or a LO1 drive level issue into the mixer or a fault in the RF signal path leading to the minicircuits mixer (or it could be linked to instability somewhere). Can you trace/probe the RF path through the attenuators up to the first mixer RF input port across 40-2700MHz to look for 30dB loss here at the high end?


Note: If there's lots of absorber stuck to the underside of the lid then this is probably there to prevent (GHz) signals propagating around the unit by reflecting off the underside of the lid. This can cause a feedback path for oscillation so maybe the unit hoots if you try and just fit an alternative bare metal lid (with no absorber fitted).

[andy_silicon]

So - focussing on the high end loss.

I've retouched a couple of dry looking joints and firmly pressed all the rf absorbers down.

Now I have 8db loss at 40Mhz and 22dB at 2.7ghz.

I.e. I lose 14db from 40Mhz to 2.7Ghz. Getting a little better.

BTW all my measurements are with the E4406a VSA at the target frequency and NOT using a wideband power meter.

( the source is an E4437B - and I've accounted for losses in the cable ).
The generator and vsa are known to agree closely within 0.2dB
( when connected by a very low loss N hardline )

[KJDS]

That big frequency slope suggests it's a problem up to and including the first mixer and it's LO.

[andy_silicon]

Last time I measured at the attenuator output it was flat.

So that leaves the first preamp and that 'bodge'

What about the carrier null at the first mixer - could that show as this ?
( I've left that area untouched - but no guarantee the previous owner hasn't had a tweak )

[G0HZU]

I'd advise against meddling with the LO null circuit just yet... I would expect that its function is to boost the isolation (at around 3.4GHz) between the LO port and the 3.4GHz IF1 port on the mixer. It looks to have two trimpot controls in a pair of microstrip tuned sections.

The net idea of this is to presumably to have control of both amplitude and phase at ~3.5GHz in order to deliberately leak the 3.4GHz LO energy across to the first IF such that the leaked signal cancels the internal LO1 leakage in the mixer itself. Without this circuit, if the LO1 was able to leak into IF1 at a reasonably high level (when the LO1 is down near 3.4GHz) it could get amplified in the two IF1 amplifiers and overload the second mixer.

Also the noise sidebands from the LO1 (when tuning down near 3.4GHz) could leak across into IF1 and mask weak signals. So this null circuit is presumably adjusted to null the net LO leakage into IF1 at about 3.4GHz.

Note:
What is a bit worrying about this analyser is that it appears to rely on NVRAM to hold all the factory calibration data. By contrast, all the analysers I have here have a hard coded (default) factory calibration that doesn't rely on a battery. But you can run a user calibration at anytime to allow the analyser to self calibrate all its attenuators and filters etc to allow the storage of an enhanced 'user' calibration and this is backed up in battery backed memory.

So in my case (HP or Advantest analysers) it's no big deal if the (rechargeable) battery fails because you replace it and load the hard coded factory cal data and then run a user cal after that...

But your analyser looks like it will lose all calibration once the battery goes flat... Not good!

If I were you I would contact IFR in the USA with your serial number and ask for all the hardcoded passwords and if it is possible to have access to the original factory calibration data file (for your serial number) so you can reload it into your analyser if your battery has gone flat.

If your battery is NOT flat then be careful not to short it accidentally because you will lose all cal data on the processor board!

I doubt they will let you have their little software program that lets you dump/load the cal data so you may need to reverse engineer the commands to allow a read or load of cal data. If you know anyone who works in a decent cal lab in the UK you might be able to sweet talk them into giving you their copy of the IFR cal software because I assume they must need a copy themselves.

[G0HZU]

In my case I found IFR UK and USA to be very helpful with my old Mi2024 sig gen.

The IFR UK office put me in touch with the IFR USA office that deals with passwords etc and they asked me to phone them at a specific time with my serial number on hand.

They then told me the unique/permanent 'master' password inside my sig gen that overrides everything else. They also sent me the full service manual in pdf format at a time when it wasn't available on the net.

[andy_silicon]

As previously stated - they won't give me the program for calibration.

I have already worked out the passwords ( one of them is in the service manual )
And I'm also quite far into reverse engineering the firmware  O0
I've found the undocumented GPIB calibration table read/write commands but haven't tried them.

Once the factory menu is entered you could manually write down the 300+ calibration values 

Anyway watching the Tour de Yorkshire this weekend - so the IFR will have to wait.

[andy_silicon]

I've confirmed that the the rf path from input to the 1st mixer has a maximum loss of 4-5db (40Mhz to 2Ghz )
Yet by the time we get to the 10.7Mhz IF we have a minimum of 16db loss.

40-500Mhz is flat as shown on the SA. ( but the power on calibration will probably correct the 16dB loss )

After that we have upto 16db roll off til 2.7Ghz.

I note that the Yig is supposed to provide +15dBm but I see about -10dBm.

Unfortunately my equipment only goes upto 4Ghz , and the Yig is 3.4-6.4Ghz.

Would a low level LO cause a problem ?
( unfortunately the service manual doesn't give levels for the 1st LO )

[KJDS]

I'd expect between +7 and +17dBm as the 1st LO drive level. if you can identify the make of the first mixer then it should be easy to check what the drive level should be.

Low LO drive level on a mixer causes the conversion loss to increase substantially from a typical 6dB to a lot more than that.

[andy_silicon]

Minicircuits MCL-1 SKY-60

http://217.34.103.131/pdfs/SKY-60LH.pdf

It's a +10dBm mixer
Or it could be +7dBm
( I'll have to look more closely at the exact model )

[dbloong]

this yig is    micro lambda .

http://www.microlambdawireless.com/

you see yig services pdf.

my  ifr-2398 spectrum analyzer  .  RBW FAIL.........

i need rbw cal

[andy_silicon]

I would suggest RBW fail is a fault rather than calibration issue.

If you look through the thread you will see the passwords necessary to get you into calibration menus.
Use at your own risk.

[dbloong]

My machine I changed the battery backup, backup battery without electricity

[dr equipment]

Hi,

I have a same problem in my IFR 2398. The problem is in YTO or in MCL-1 SKY-60?

Can you send me a zoom photo of MCL-1 SKY-60 and the circuit around?

Thank you 

[WernerFalken]

Old thread, I know, but just wondering if you got this sorted.  I repaired one with level/rbw fails and it was right at the front of the input - one of the AH002R2 switches.  Diagnosing that fault wasn't as straight forward as it sounded.  A lot of wild goose chases!  I went through much of the same frustration with making a riser card/RF adapters to get it functioning so I could probe with another SA.... quite fun.  And locating a NEW replacement of that chip was a PITA....

Depending on the rev, there are two different Ax002R2 chips: AH002R2, and AW002R2, and they are not drop-in compatible for these units.

On a calibration note: IFR told me that these units are hardware designed to self-cal within 1db across the spectrum.  After that, a cal table is used.  From what I can tell, the statement was truthful.  I used a HP 8664A & HP 437B to check it after intentionally wiping the NVRAM, and the repaired 2398 was pretty darned close for hobby work!  Going in and manually calibrating works, but isn't entirely intuitive, so I haven't nailed it down 100% yet.

Anyhow, I started a Google Group to gather info about this model, along with the LG SA-7270A, which appears to nearly identical to one revision.

https://groups.google.com/forum/#!topic/aeroflex-ifr-2398/3C1R7NfneEk

I have a bunch of content there - including some firmware files.

[BravoV]

Okay - the service manual mentions keying in the password "27329780" to get to the factory setting which allows the calibration tables to be cleared before recalibration.

For the life of me I couldn't get this password to work. ( no enter key! )

So after much reverse engineering I now see the bloody obvious.
( once you see that they used a generic routine for entering numbers it was obvious the termination of input was one of the Hz/Mhz/GHz buttons )

It's the "Hz" key to enter the password.

Sigh....

At least on the way I discovered another password for the system menu "72237908"

Needless to say don't mess with these extra menus unless you wan't to wipe out you calibration.


There do appear to be undocumented gpib commands - the work continues.

Aware this is an old thread, just want to say thanks as these both passwords worked at my LIG NEX-1 NS-30, as its originated from Korean LG, before sold to Aeroflex and renamed to IFR-2399

Using that, enabled few 'extra' software features. 

Anyone have the abandoned diagnostic/calibration program for this Aeroflex spectrum analyzer series ?

Few shots of my NS-30 when unpacking while ago.

[WernerFalken]

After many years of on/off work, I just completed the Backup/Restore process for the 2398 (python). As it is fairly similar, you might give the backup a shot. If the unit has its calibration data intact, I'd advise against attempting the restore (of course), but getting it backed up is important.

For the 2398, specifically, the calibration process should be trivial to workout now that the backup/restore is done. Pairing the commands with the service manual should be pretty straight forward, but I am no longer in possession of a fully working 2398... sigh.

PS Much of the seed work credit for this outcome goes to andy_silicon, btw. Without his work, I wouldn't have had to spend time doing this Much of the blame goes to him....lol

[BravoV]

Colby, appreciate the work. 

Is it possible to backup the cal data through the serial port instead of through GPIB ? As I don't own one. 

[WernerFalken]

If you can issue the commands and retrieve the data that way, it should be possible. The returned data is simple text/floating point data that I dump to a file. It isn't some special binary formatted file dump from the machine or anything. Issue a command, read the result, store the data. Restoring the data is similar - just commands repeated for each calibration data element.

I'd like to say that I'd try it soon, but I doubt it. I did keep my one, barely functioning unit out, so maybe some day I'll try.

[tech255]

I have been reading this post with avid interest as i have a 2399b with rf problems the yig test fails in the self test routine so im just about to start having a go at repairing it, i think ive identified the white blob with 101 if you follow this link theres several on the page.

https://www.rf-microwave.com/en/transistors-rf-fet-gaas-fet/

http://www.hp.woodshot.com/hprfhelp/4_downld/products/xrs/atf10136.pdf

Ive got access to a spectrum analyzer that will get to 6ghz but no probe just wondered if i could connect it directly just to measure Whats needed with maybe capacitor or resistor in line.

Any help would be useful
Thanks

[zed dez]

Hello all,

@tech255: From my previous experience with my 2395 (same as 2398 but with a frequency extender 3-26.5GHz) I'd say that YIG Fail is due to the YIG itself. It is part of a PLL and the error is triggered if unable to lock. My unit was overheated (faulty cooling fan) and inside the YIG found some gold wires disconnected (the main coil and the supply in my case). Carefuly inspect them with a microscope. When OK I can confirm the YIG will lock even when RF Synth module is opened and the only conection needed is the 10MHz reference. Also, you can check the 3.2-6.2GHz directly, it is outputed at second upper exit, counting from the IF output. A simple capacitor probe will be enough, and the LO level is very high, you can't miss it. Set the SPAN to zero.

@Colby: Many thanks for your work, helped me a lot for my 2395. It is same as 2398 (even the passwords are the same). I can confirm the GPIB link can be substituted by the RS232. Inspired by your work wrote my own calibration utility and a TRAALL? to display the trace on my laptop and play with it. The 2395 is going up to 115200. At this moment all calib data in my unit is erased, only zeroes.
Still I have 2 problems:
- For the attenuator calibs 0 to 50dBm I'll have to guess a reference for the input test signal, I don't have FCAL2398 to know how exactly is setting the RF generator during calibration steps.
- It is still unclear to me how the YIG slope is determined. I guess the offset is based on the internal test signal by tuning to 40MHZ and then gently sweeping the YIG till it gets a max level value. For the slope a second test at the other end of the band will be required, let's say 2GHz. With a single point it is assuming the GHz/volt curve is perfect linear, that is close but not perfect.
In plus, the 2395 is fitted with a frequency extension re-using the same LO (x1, x2, x4) with extra sweeps to cover the 3-6-13-26GHz bands. At it's input there is an YIG filter that must be calibrated to go in tandem with the LO YIG.
Maybe extra GPIB commands, since I don't have PSOP adapter to read the flash I'll try to boot a linux or DOS, with external keyboard pressing ESC at boot the unit is entering BIOS like a PC

A good new year to everyone!