Agilent E8357A

[joeqsmith]

Attached comparing Caesarv's r4p12 file with mine.  Ceasary's is 12, mine is 21.     Looking at the RAW data, the peak-peak and standard deviation are very close but the mine for what ever reason reads about 5dB higher than expected.  Also notice that mine doesn't exhibit that step at 3GHz.    I was actually thinking to make a cal file that just adds a fixed offset to the raw data.

Maybe mine had some different hardware revision that caused this offset and removed the 3GHz step.

DELTA is the difference from the RAW and the power meter reading.   With the RAW data being so stable,  I suspect most of this error is my Signal Hound and not the PNA's source.  When I made this measurement, I used a short section of semi-ridged to attach the SA.   

[Miek]

Keep in mind all the stuff I've written above is just my best guesses at how it could work, it might not be the power meter response in there. It could be reference receiver response instead, and that's why the phase is normalised to 0? Thinking about it some more, I would expect the power meter trace to be pretty flat from the source calibration adjustment.

Are you able to create traces for individual receivers like in the attachment? If so, you could do that and then fiddle with the values in the mxcalfile to see what changes.

[Forzaman]

Received my HDD and duplicated the drive. Next task is to replace the battery on the CPU board. The system date keeps reverting back to 2088 when I power-up the system. The old battery is a RayoVac BR2335. They seem hard to come buy these days, can only find the CR2335. The battery clip seems to be very tight. Don't want to damage anything by using too much force to pry it. Caesar... can you give any advice on changing the battery? There is no mention of battery replacement in the service manual.

[joeqsmith]

Same here.  Purely guess work as I don't see where it is documented.  Yes, the PNA has a few ways to look at separate receivers.  I mentioned using this earlier. 

My question for Forzaman and Caesarv,  if they rename these cal files, exit and reload the PNA program, display S11 with nothing connected, does theirs display 0, or 6ish dB?   This will at least make sure my PNA is not doing something abnormal.   

Received my HDD and duplicated the drive. Next task is to replace the battery on the CPU board. The system date keeps reverting back to 2088 when I power-up the system. The old battery is a RayoVac BR2335. They seem hard to come buy these days, can only find the CR2335. The battery clip seems to be very tight. Don't want to damage anything by using too much force to pry it. Caesar... can you give any advice on changing the battery? There is no mention of battery replacement in the service manual.

  I just changed mine last month and ran into the same problem.  The cost for the BR seemed like it was over $20.   I tossed in a CR.  Not expecting it to last as long.   

[Forzaman]

Same here.  Purely guess work as I don't see where it is documented.  Yes, the PNA has a few ways to look at separate receivers.  I mentioned using this earlier. 

My question for Forzaman and Caesarv,  if they rename these cal files, exit and reload the PNA program, display S11 with nothing connected, does theirs display 0, or 6ish dB?   This will at least make sure my PNA is not doing something abnormal.   

My VNA is dismantled right now. Going to pickup a battery on Monday. If Caesar doesn't reply by then, I will give it a try and let you know.

[joeqsmith]

No problem, I'm in no rush.  When you get to the point where you are going to align yours, it would be very interesting to use the CW mode, with the output set to 0dBm and measure a few points manually.   
I would expect the meter to read 0dBm but who knows, maybe it's -6ish. 

I was concerned with the BR2335 being so odd, if I bought one how long had it been on the shelf.   A lot of my devices use the CR2032 which is a bit smaller OD and thinner.  I doubt it would pop out but you may only get a few years of service life. 

[caesarv]

Interesting.  I did not even know the difference between a CR and a BR battery until just now.  I have always replaced them with a CR battery because they were much more common and have never had a problem.  Now that I have looked it up, I agree that a BR would be slightly better for the RTC, but I doubt it makes much difference.  Both should last 5-12 years...longer if it is kept plugged in.  I always used the common CR2032 battery Just pry the battery out....I have never damaged the board by prying it out...but I was only working on the 500 MHz board.  I do not remember if the 266 MHz board had a similar battery.

With regard to what I said earlier about the mxcal files being about the same:  I probably should have added the caveat that this may apply ONLY IF you have the same options.  The one I provided was for an E8358A with 3 jumpers on each side of the front panel.  If you only have 2 jumpers on each side, then my guess is that you should use the mxcal files from a similarly equipped instrument.  I can provide a set of those files also.

[joeqsmith]

I believe Forzaman's is also the 3 jumper system.  Mine is the two jumper, no frills.   If you don't mind, please post the files you have for the 2-jumper.  Maybe this will explain the offset.   

Also, mine is the low end computer and uses the same battery. 

[Forzaman]

I believe Forzaman's is also the 3 jumper system.  Mine is the two jumper, no frills.   If you don't mind, please post the files you have for the 2-jumper.  Maybe this will explain the offset.   

Also, mine is the low end computer and uses the same battery.

Yes, I have option 015. Ok, I'll try to pry out the battery. The 266 and 500 MHz CPUs use the larger 2335 size coin battery (BR type) whereas the 1.1GHz uses the 2032 smaller size.

[Forzaman]

Caesar, are you familiar at all with the configurable test sets available for the E835xA PNAs that allow 4-port measurements?
There are quite a few N4416A test sets available for the E8357A, but none come with the N4425A Balanced Measurement Software required to control it. I was wondering if the software requires a license, and if so, is it node locked to the PC. If it is node locked to the PC, then I will give up my search, but if it is somehow locked to the test set, then I will continue my quest to get one with the software. Thanks!

[caesarv]

Attached is the 2-Jumper version the mxcal files from my E8358A.  The previous one I uploaded was for the 3-Jumper version.  Again, these are only approximate for any unit but the one it was created for.

Sorry, I do not know anything about the N4416A or the software that was used for it.

[joeqsmith]

Shown with 2 jumper on 12 and 3 jumper on 21.   There 's about a 1dB shift between them but still the big offset.   

My question for Forzaman and Caesarv,  if they rename these cal files, exit and reload the PNA program, display S11 with nothing connected, does theirs display 0, or 6ish dB?   This will at least make sure my PNA is not doing something abnormal.   

Another thing I would like to see is after alignment (assuming both of yours are),  could you also plot S11 and S22 with an open. 

I installed the 2-jumper files and swept S11/22 with the an open.   While its much worse than my attempt to align it, it addresses much of the offset.  Seems to suggest there is something else going on with the alignment and what is actually being recorded.   

[joeqsmith]

Just for completeness, shown with Caesar's 2-jumper files using an old HP cable for a thru and measuring S21/12 (no cal).   

[Forzaman]

Joe, I attached my screen captures with and without the mxcal files loaded. Note that the scale was changed to 2dB/div on the plots without the mxcal files loaded due to the larger offset. For the through measurements I used a 3-foot flexible SMA test cable with adapters on the ports. The insertion loss characteristics of the cable is also attached.

[joeqsmith]

Thank you very much for running that.  This answers the first big question.  Seeing that mine exhibits roughly the same offset, the next question is do they reduce the power level when running the alignment.   If you plan on using one of the called out power meter/sensors with the internal alignment routine,  I would be very interested in knowing if the power level is reduced to -6ish dBm or if they run it at 0.  I assume you can just read it right from the power meter.   I'm not too worried about the entire data set, just the first data point would be good enough to answer that.   

That HP cable I show was only used to a couple of GHz.  Good enough to see if the files Caesar provided would produce an offset or not.

One step closer...

[Forzaman]

Yes, I plan on performing the alignment with the proper equipment, but only for the existing 6GHz configuration for now.  I'm not going to do the frequency upgrade until after the warranty expires. I can round up all equipment except for the 8482A sensor so far... still asking around.

Will let you know the power level during calibration.

[joeqsmith]

That's fine and will tell me what I want to know.  Because mine created the files even without anything attached to the GPIB, I wonder if it goes though the adjustments first.   I'll try that again while monitoring one of the ports. 

Just an FYI, if it wasn't clear, you can backout that range change or any other option if that was your concern about the warranty.   I was thinking changing the hard drive and battery would be more intrusive as you have to open it up where the range option is just software. 

Mine came with a return grace period.  Maybe a few months.  More than enough time to check it.  I hoped to ship it off to Caesar before they left KeySight.  They posted about having some of the CPU boards in stock and I knew that sending it in would solve the alignment and uncover any other potential problems.    Mine had been worked on before by some monkey in a cage which increased the risk.  By the time I was running down the clock on that grace period, I had used the PNA enough to be hooked.   No more 70's thumb wheel switches.   
 

[Forzaman]

Thanks for the info Joe. I didn't realize you could revert back to the original HostID and frequency range after the upgrade. PA0PBZ provided me with the key for the time domain which I installed and is working well.  If I need to send it back, I can reinstall the original HDD. I doubt they will realize the battery has been changed. Even if they do, I don't think they would have expected me to ship the unit back just for that. There were no cal or warranty stickers on the case, so they shouldn't be able to tell if I opened-up the unit.

[joeqsmith]

At least with mine, I am able to just copy the original license back into the PNA and it's resorted.   

I've been playing a bit more with the receiver alignment and reading what I am able to find.  Looking at the service guide for the N5221A, they use the power meter to measure the cable loss.  They don't mention measuring the power from port 2.  What's interesting about this is the data you both supplied, the second file has a slight loss.   

I had tried to run the alignment without the power meter and monitoring the signal from PORT1.  It seems to suggest that the run it at 0dB.   They do call the mxcal files, mixer calibration.   It would make some sense if they are using R1 as part of this measurement, we would see the 6dBish loss.

I wrote simple program to parse the ASCII characters from the alignment software.   I further filtered the data by showing only characters with adjacent ASCII characters (see attached).   

Attach one end of  cable to Port Then attach )IJ(Rx sensor to other end of the cable. Use adapter if necessary.Please wait...measuring cable loss valuesAttach the  sensor to the end of the cable

*************
Receiver Adjustment
The receiver calibration is used to adjust the network analyzer receivers for a flat response across its full frequency range:
1. A power meter/sensor is connected to Port 1, as shown in Figure 3-16, to establish a reference for flatness.
2. A cable is inserted between the power sensor and the test port, as shown in Figure 3-17, to establish a reference for the cable.
3. The same cable is connected between test port 1 and test port 2, as shown in Figure 3-18, and a signal from Port 1 is used to adjust the “B” receiver at Port 2.

The adjustment is repeated using a signal from Port 2 to adjust the “A” receiver at Port 1.

Data obtained during this adjustment are stored in the mxcalfile_pxx files inflash memory on the test set motherboard, with a backup copy stored on the hard disk drive. The data are used in subsequent measurements.
********

[joeqsmith]

Again, pure guess work...

Measured power directly at port1 and also R1.  Insert test cable on port1 and measure power at end of cable along with R1.   Subtract R1 and zero the phase.  This seems to get us very close on the two lower graphs for NOCAL.PNG.    Attach cable to port2 (thru).  Measure S21,12.    Compensate the mag only with new levels.

S21/12 with the cable used to align it is now very good.  S11/22 also seems decent but procedure could be totally wrong. 

Next step is to try a patent search.   

[Forzaman]

At least with mine, I am able to just copy the original license back into the PNA and it's resorted.   

I've been playing a bit more with the receiver alignment and reading what I am able to find.  Looking at the service guide for the N5221A, they use the power meter to measure the cable loss.  They don't mention measuring the power from port 2.

Do you not have the E835xA Service Guide?

In the Source Calibration procedure prior to the Receiver Calibration, it mentions not testing the Port 2 power since the difference from port 1 is negligible.

Again, pure guess work...

Measured power directly at port1 and also R1.  Insert test cable on port1 and measure power at end of cable along with R1.   Subtract R1 and zero the phase.  This seems to get us very close on the two lower graphs for NOCAL.PNG.    Attach cable to port2 (thru).  Measure S21,12.    Compensate the mag only with new levels.

S21/12 with the cable used to align it is now very good.  S11/22 also seems decent but procedure could be totally wrong. 

Next step is to try a patent search.   

[joeqsmith]

Do you not have the E835xA Service Guide?

In the Source Calibration procedure prior to the Receiver Calibration, it mentions not testing the Port 2 power since the difference from port 1 is negligible.

I had read that but note that it was not specifically called out in the receiver calibration.   The reason to look at the embedded comments in their program was to solve some of the ambiguity.   

When extending the range, both need to be ran.  If I plot the source, I can see they only characterized it to the range it was sold for, not considering it would be upgraded.  Its pretty bad.

[Forzaman]

[/quote]

I had read that but note that it was not specifically called out in the receiver calibration.   The reason to look at the embedded comments in their program was to solve some of the ambiguity.   

When extending the range, both need to be ran.  If I plot the source, I can see they only characterized it to the range it was sold for, not considering it would be upgraded.  Its pretty bad.
[/quote]

That is bad news about the calibration not going beyond the original frequency configuration. I may not upgrade my unit now. I don't have any need to go beyond 6GHz at the moment. Obtaining the time domain option was my main goal. 

[joeqsmith]

Worse, I am not able to find any information on where the source leveling data is stored.   I had also backed out the ASCII data for the sourceadj.exe program (attached).  It looked like it may have stored it into the srccalconfig.dat file.  Odd thing about that file.  I had backed up the discrete files from the original hard drive.  After installing the SSD, I upgraded the firmware.  This file now has a 2005 rather than a 2004 date.  They are different but the fact they overwrote it, I don't believe this is where the source data is stored. 

When looking at the ASCII dump, note the "SENS1:CORR:CSET:STAT ONSENS2:CORR:CSET:SEL 'SENS2:CORR:CSET:STAT".   Looking these up in the programmers manual, we can see they are for calibration.    The one I was expecting was the SOUR:POW:CORR but we don't find it in the EXE. 

That is bad news about the calibration not going beyond the original frequency configuration. I may not upgrade my unit now. I don't have any need to go beyond 6GHz at the moment. Obtaining the time domain option was my main goal. 

The attached graph shows the source set to 0dBm, sampling at each frequency that the receivers would be aligned at.  You can see how the performance degrades once  we move beyond 6GHz. 

In my case leaving the alignment may be the best answer.  I don't like not knowing with 100% certainty how these adjustments are made and how they effect the end result.   I understood you're planning to use all of the correct equipment.  In that case, I don't see a reason not to open up the range and realign it.   

[joeqsmith]

I don't have any need to go beyond 6GHz at the moment. Obtaining the time domain option was my main goal.

The only reason I was wanting to increase the frequency range was to increase the time domain resolution.   I wanted to use the PNA as a comparison to the $120 LiteVNA.  They limit the LiteVNA's firmware to 9.3GHz.   Attached are two of the circuits I demo'ed  (splitter with an LC tank and a low pass filter).   The data from the PNA is pretty much what I would expect.   

You can see the demo here if interested: