Advantest R3477 9kHz to 13GHz spectrum analyzer

[nctnico]

Both traces are averaged 10 times but the picture shows measurements from two different sources.

[Wuerstchenhund]

The R3477's manual doesn't specify the phase noise (at least not directly in dBc).

Apparently it's -100 dBc/Hz (10kHz offset):
https://www.custom-cal.com/ModelInfo.aspx?kn=1157&m=ADVANTEST_R3477&srv=Calibration

I think it is needless to say the E4421B's output is noisy indeed but that may be caused by my self made output amplifier.

Why don't you just use your 10MHz reference as input signal? Rb references and a good OCXO has a much lower phase noise than a vector RF generator like the HP ESG.

[dschiedsch]

Hello ntcnico.
I grabbed one of these analyzers from ebay as well.
Working fine but I found out that the calibration above 3.5GHz is off.
I guess the cal for the YIG Filter is not good anymore. (Tested against a RS-signal generator)
I think one can calibrate this through the service menu. However I also don't have the password.
Did you have any success finding the password on the flash disc?

regards
gernot

[nctnico]

I already posted it earlier:

The code to enter the service mode: 17320504

Use at your own risk though! I'd check if there isn't some frequency compensation enabled in one of the many menus.

[dschiedsch]

Hi
yeah thought about that and tried to put everything back to default

also do you know if the calibration is stored on the disk?
if yes then a backup of the disk should minimize the risk of bricking the device

[nctnico]

I have found some files which seem to have calibration data in them in the C:\Advantest\R3477 directory on the disk.

[dschiedsch]

Did you ever check your analyzer in the range above 3.5GHz?
Is it accurate there?
I've made a table with my measurements and attached it.

Also coud you check the values of your YTF Tune adjustment in the service menu
because mine reads 0 for parameters 1 to 6 which I find odd

thanx

[nctnico]

Can you post a picture of which settings you are referring to so I can make sure I'm looking at the same?

BTW I found something odd: the frequency calibration file on my R3477 goes to 15.4GHz and it is indeed possible to select frequencies up to 15.4GHz!

[dschiedsch]

Hi
I used a R&S ZNB8 network analyzer as signal source and compared the measured signal between my R3477 and a R&S FSP analyzer
Settings were ref +10dbm attenuation 20db span 100MHz and RBW=VBW=1MHz
Signal was 0dbm.
I measured the peaks with a marker and put them in a table.

And as you can see from the measurements at around 6GHz there is quite a discrepency between the measurements.

[dschiedsch]

Oh sorry
i realised that you meant the YTF settings 
and i saw that you have to go to the band select menu to change between the ytf harmonic modes.
and only the base band has TC whereas the H1,H2and H4 have parameters 1 to 6
Also if you select H4 band the setting of the upper screen is centered at 15.2GHz.
So it seems tha calibration goes beyond what the analyzer is officially working
anyway here is the picture

[nctnico]

Mine also shows numbers in that menu. Unfortunately I don't have the means to test at over 3GHz. I do have an ADF4350 board somewhere but I doubt that will be any good to check a difference of a few dB over a (relatively) wide frequency span.

[dschiedsch]

Yes you're probably right.
I'm only lucky that I can use some equipment at work which I know is calibrated well.
I only used the analyzer below 3gig for a few months and didn't see any issues. Only to find out it was off when I measured a 5.8ghz video transmitter.

So next thing would be to find a calibration manual. Because changing values in the service menu and looking what happens is probably too painfully slow :-)
Any ideas where I might get this?

[suj]

Hi guys. For some time i have Advantest R3681 32 GHz Signal/Spectrum Analyzer with integrated 6 Ghz generator. I suppose that this is the common platform with R3477 (Windows XP Embedded, touch scren, similar performance). And finally, the password for the service menu is the same - thank you guys . I don't have the calibration instructions, but the search base is expanding now. Each instruction  (R3477, R3671, R3681) will be useful - service menu looks identical.

[dschiedsch]

Hi
The a1freq.csv file (changed filetype to .txt for attachement) in the Advantest\R3477\Fact_cal folder seems to hold the calibration values over the whole band.
It gets loaded at boot (I moved it to a different folder and the analyzer throws an error) and I think this is where I want to change values so
the dB values in the high band are correct.
Now the question is which column is what
Obviously the first is frequency and I would guess the second is the harmonics mode (aka if it's in base band or using one of the YTF modes H1,H2,H4)
The values after that I guess are correction values in dB.
The question is now what the different columns are for
different RBW settings maybe?

If anybody can make a better educated guess then please enlighten me.

[nctnico]

You could try and change a few values with extreme numbers (+6 or -6) and see what happens.

[nctnico]

Today I got round to (somewhat) checking the response on my R3477 to see if it has the same inconsistancies as dschiedsch showed. I tried one of the cheap ADF4530 generators but it seems I have received a defective one. So I resurrected an ADF4351 frequeny generator / sweep board I made a couple of years ago. The ADF4351 can generate frequencies up to 4.4GHz and I think my R3477 behaves OK since it doesn't have the sharp step around 3.2GHz. The second graph shows the output from 10MHz to 15GHz in order to give an idea about the amount of harmonics the ADF4351 spits out.

[dschiedsch]

Hi
I started working on the a1freq file and indeed it does what I thought it would.
Its a calibration file for the frequency band.
the format is as follows:

frequency, mode of YTF, 5x 0db attenuation correction, 5x 5db attenuation correction, ..., 5x 55dB att correction, 6x 60dB and up correction
its a total of 66 correction factors but only 13 are unique. there are 12 sets of 5 and one set of 6

so now since I know what the values are for I can try recalibrating with a known source. Luckily I know someone from university ...

[videobruce]

Three questions:
1. How heavy is this?
2. What is the power consumption (watts)??
3. If the unit is powered on, can you change the battery w/o loosing calibration??

[dschiedsch]

Im Not at Home for the next four weeks. However
The calib Data is on a cf Card so there is No Problem with a battery.
On the question of how Heavy it is i would have to Check  but there is a datasheet online so please Look it up.

[nctnico]

Three questions:
1. How heavy is this?
2. What is the power consumption (watts)??
3. If the unit is powered on, can you change the battery w/o loosing calibration??

You can find the weight and power consumption in the manual you can download from here: https://www.advantest.com/service-support/electronic-measuring-instruments/operating-manual-of-discontinued-products
The calibration is stored on the hard drive which is a standard compact flash card. I have made a (working) backup of the CF card in my R3477. It is a PC so changing the battery only affects the BIOS but the standard settings work just fine.

[nctnico]

Today my R3477 threw me a nasty surprise. I have not used it for a couple of months and it rewarded me with a 'self test failed' message. Great, just when I need it for some tests.

Checking the diagnostics info showed the digital IF board wasn't responding.  Great!
Time to pull the digital IF board out. Since the boards aren't labelled I just pulled the one furthest away and it turned out the be the right on. The plugs are numbered according to the position they are supposed to be plugged in. Number 4 goes into the 4th connector.

Now here is the board:

What the hell is wrong with it?
Let's apply the standard fault finding tactic: check the fuses!

Bingo! Fuse F271 which seems to be the 5V rail is blown. Now to find the problem. An easy way is to push some current (with the PSU set to a low voltage) into the board and use the thermal camera to find the hot spot.

Which is this capacitor on the board:

I desoldered a similar looking capacitor and measured it. It turns out the be 22uf. Since I don't have those at hand I placed 2x 10uf/50V which should do it and botch repair the fuse.



After checking the short on the 5V rail was gone I powered the R3477 on and.... yes it works! Pfew... 

[AF6LJ]

Nice catch.

[nctnico]

Indeed. It just goes to show that boards which look very complicated and unrepairable can in fact be fixed. Ofcourse the approach Dave uses in the 'Data IO' repair video (EEVblog #1061) to find a shorted component also works but the thermal camera is just quicker.

Meanwhile I got the right fuse, soldered it in and put everything back together.

[DaJMasta]

Got one of these analyzers and have been working to quiet it down, but I had a look around at some things which may be useful to others:

The fan monitoring is done by an FPGA under the CPU fan blower which overhangs from the main computer module - but it seems to be tolerant down to pretty low tachometer signals.  I hooked up the signal gen to the two board's fan connectors and verified both to be happy down to at least 50Hz.  The CPU fan is a 5V job, and the default blower is stupid loud... but what baffles me is that the computer module is designed to be run fanless (I suspect with some ambient circulation), so why throw in such a loud blower if anything?

The clearance from the board to the top cover of the module is around 35mm, so it's not easy to get a regular fan to blow across the fins, so what I opted to do was to get a 12V rated 60x60x15mm fan and mount it on the opposite side of the fan module attached to the top plate (three new screw holes, but easy to do and plenty of clearance to keep the intake from being obstructed).  Since there's space for a small hard drive there, it's unpopulated with the CF card for the OS and is open space.  At 5V instead of 12, this blower is quite quiet, and still manages about 70Hz on the tachometer signal.

The PSU I'm running into the same problem for, as best I can tell, it's the primary source of noise now.  I tried swapping in a fan rated for 4500rpm and it still wasn't happy, and yeah there's no tachometer signal from the PSU, just the alarm signal, so you'd have to find a similar RPM fan or spoof it somehow.  I looked around for settings for what's required in terms of RPM in the software, but it's all done in the FPGA it seems like, and with a hex editor I can see a command prompt call following the fan stopped error message to allow a shutdown to be called by the program, but not the shutdown call itself.  And really, I don't want to stop the thing from monitoring the fans altogether, I'd like to know if one fails legitimately too.

Anyways, for some other notes about the instrument:

The basic CF card is 50x speed, moving to a 140x or so card was a notable load time improvement

The processor board is an SOM-4470F with a 700MHz Pentium 3 and 512MB of PC133 RAM installed (but only running a 100MHz FSB).  It's designed to run fanless, so should be just fine with the ventilation for the card, and it's an ETX form factor board, so there's some chance a different board could be used (though with the FPGA interface and the various outputs mapped to the bottom connectors, it could be difficult to find a match, I don't know the spec.)  As it's a soldered on chip and a maxed amount of RAM, there's no upgrades to be had here.

Fan 1 is the CPU heatsink fan
Fan 2 is the computer module ventilation fan
Fan 3 is the 120mm rear fan
Fan 4 is the PSU fan

It appears that Fan 3 does not send a tachometer signal back to the computer module, it could be monitored on the board it connects to and then sent back as an alarm signal like with the PSU.

PAX.exe is the main screen you see in the operating system and it handles all of the rebooting from the fan monitoring, so closing it prevents the reboots if you're tinkering with the fans.  I was unable to see exactly how the FPGA talks to the program, the device manager is not installed, but it seems to be a parallel interface from the routing on the board, and using a terminal to monitor COM1 yields no results.

Most of the directories are hidden if you access the CF module on a different computer (you just have to check view hidden files in explorer on the instrument itself), but the important ones are C:\Advantest\R3477 where factory calibration information is along with some other stuff and C:\Program Files\Advantest\R3477 is where the application itself is, as well as text error logs.  For me, there's one more, C:\FIDTSERV , because that seems to be where the touchscreen recalibration software is.


I've got a replacement screen and touchscreen coming for my unit, but otherwise it seems happy.  Had to replace the lithium battery on the CPU board for it to keep time, but it seems to be a real solid instrument and I'm looking forward to having it around.

[nctnico]

I replaced most of the fans in mine (same airflow/pressur rating) and used the rubber fan mounts (see earlier post). I did disable the fault signal from the PSU. I think the fan on the CPU is the loudest and I did not replace that one. However I think it might be possible to just remove it because the fan on the CPU cage is already providing airflow.