E4406A Phase Noise in wide span

[dinocroc123]

Hey guys,

i've recently acquired an E4406A transmitter tester. Unfortunately it shows some unexpected behaviour: When i feed it a Signal at 800MHz (frequency doesnt seem to matter though) and 0dBm with the span set to 1MHz and the RBW to 20kHz (AUTO) a noise marker
at +500kHz (left of the screen) indicates -120dBm/Hz. Now the phase noise of my source at a 100kHz offset from the Carrier is at -130dBc/Hz. The E4406A should at a 100kHz should be better than -125dBc/Hz and indeed
when i select a really small span of 10kHz or so i get close to that spec. Why is the Phase noise so bad in a large span? Anyone of you have the same device and observed the same behaviour?

Thanks in advance!

[rfclown]

I don't get that result. I set my E4431B to 800 MHz, 0 dBm, fed to E4406A. Set E4406A to 800 MHz, and everything else default (like you did) 1 MHz span, RBW=20K, marker at far left of screen (set to noise marker). I get -132 dBm/Hz. Set marker to Spectrum Average Trace.

[rfclown]

So get up this morning and realize that I didn't measure with a narrow span. Try to turn on E4406A... and it doesn't come on. Aaaargghh! I've got another buried in a closet with a fan stuck on high speed (very loud), but this one presently has other stuff stacked on top of it, so I'm not going to get the other one out any time soon.

[dinocroc123]

Thanks alot! Its weird, i'll get back to this thread if i find something.

[Scrapcollector]

Hi,

I'm working with OP to solve this mystery. Thanks RFclown for posting this reference pic, sad that your psu now died...

I've tested four different E4406 (three shown in the pics) and so far they all show not better than -120dBc/Hz with 0dBm input at 800MHz with a verified source. With -10dBm, it looks better of course. The calibration of the units tested is "a bit over", so its unclear, if they are all defective, or if I am still missing something.


The specification at 960MHz is
https://www.keysight.com/de/de/assets/9018-03347/technical-specifications/9018-03347.pdf  (p.13)
"Offset 600 kHz ≤ −136 dBc/Hz
Offsets <1 MHz measured with RF Input ≥ −2 dBm; Offsets ≥ 1 MHz measured with RF Input > +12 dBm"

I will keep testing...

[rfclown]

I don't know if this makes a difference... but the person who turned me on the the E4406A, which at the time were being dumped by the boatload after the dot com crash, told me to get the one's with 14 bit DACs, which I did. I don't remember the details, but you can Google it. Forget if it was Malaysian or US made models made in a certain time period used 14 bit DAC cards. This might (or not) be a factor in noise floor limitations.

[Scrapcollector]

Today I tested another unit, and this one tested good, even 2dB or so better than your screenshot.

Its made in Malaysia. You might have pointed us in the right direction... I will check if this issue is consistent with the DAC revision, or if the previously tested units are just bad.
https://edadocs.software.keysight.com/kkbopen/does-my-e4406a-vsa-series-analyzer-have-a-14-bit-digital-if-or-a-12-bit-digital-if-589306038.html

Another question occured.
https://www.keysight.com/de/de/assets/9018-03347/technical-specifications/9018-03347.pdf (p.13)
The technical specification specifies the noise sidebands at 673.6 MHz, 960 MHz and 1990 MHz with offsets of
0.1 kHz,1,10,100,600,1.2 MHz, 6.0 MHz and 10 MHz (Offsets <1 MHz measured with RF Input ≥ −2 dBm; Offsets ≥ 1 MHz measured with RF Input > +12 dBm).

The Instrument Messages and Functional Tests Guide ...
https://www.keysight.com/de/de/assets/9018-01461/user-manuals/9018-01461.pdf?success=true (p.59)
...uses different test frequencies and only two offsets, 10 and 30 kHz.
There is no test limit for offsets over 30kHz. And there is no spec for 30kHz (only 10 and 100 kHz).

At 960 MHz, the spec with 10 kHz is -96dBc/Hz with "something over -2dBm input" and the test limit at 900 MHz is -80dBc/Hz with 7dBm input and 10dB input att switched (while no att setting is written in the spec for -2 and +12dBm).
Is there a third document to make sense out of this or should I just take it as it is, and test for both?

Next issue, the messages and functional tests guide is inconsistent at the procedure. In the text, it says to set Mode to Basic, but the table says Mode is Service. I think It does not work in Basic mode, because I can not set ADC Range to -6, only 0, +6dBm etc.

Heres the short version for those who do not want to download the pdf to look at the test procedure:
Step 1: Align All 2: Prest Step 3: Set UUT to measure the carrier: Mode:Basic, Marker:Trace:A+Spectrum Avg, Freq:600 MHz, Span:800Hz, InputAtt:10dB, ADC-Dither:On, ADC-Range:Manual:-6dB, Pre-ADC BPF:Off, AVG-Number:5, Avg Type:Pwr Avg(RMS), TraceDisp:Avg.

Step 4: Preset the source to 600 MHz 7dBm. %. connect. 6. Adjust amplitude until UUT shows 7dBm.

Step 7: Reset the UUT to measure noise: ResBW 5Hz, ADC Range:Auto, Pre-ADC BPF:On, MarkerFct:Noise.
Then the offset Frequencies 600.01 and 600.03 (and 900.01 MHz etc) are measured with an 800Hz span at 5Hz bandwith.

When doing this, there's a spurious response at -92dBm right next to the center, but looking at the spurious response spec this seems OK.
The test manual btw only tests residual response, with nothing at the input.
The specification says, with input CW between 700 and 1678.6 MHz the spurious signals are below -59dBc.

[G0HZU]

So get up this morning and realize that I didn't measure with a narrow span. Try to turn on E4406A... and it doesn't come on. Aaaargghh! I've got another buried in a closet with a fan stuck on high speed (very loud), but this one presently has other stuff stacked on top of it, so I'm not going to get the other one out any time soon.

Take the small fan controller PCB out of both of your E4406As and replace all of the 100k resistors in 1206 package. They are probably a green colour and they say 1003 on top. I think there are about 6 of them but I can't remember.

Next, replace the two 150k resistors, also in 1206 package also probably green in colour and these are on the other side of the board.
My guess is that one or more of these resistors has gone open circuit. This is a common problem on this board.

In the time I've owned it, my E4406A (14 bit) has had both of the faults you were describing. The stuck/noisy fan was caused by one of the 100k resistors going open circuit.
A few years later it died shortly after not be used for a while. It powered up once, then died.

One of the 150k resistors had gone open circuit and replacing this meant I could power it up again.

I think these resistors are tight tolerance types so best to fit sub 1%  (I think I fitted 0.1%)

[G0HZU]

Just to clarify, when my E4406A died, the green standby LED was still lit OK but the unit would not power on. Also, you have to remove the 150k resistors to measure them as I recall that there is some resistance in the circuit that masks the fact that one of the 150k resistors had gone open circuit. I think it measured about 180k in circuit but >100M ohm when removed.
I think it was the 150k resistor nearest the edge of the PCB that had failed on mine. I've got pictures somewhere but there's only two of them on the board. Best to replace both of them anyway.

[G0HZU]

I tried the 1MHz span test on mine at 800MHz and 0dBm and got the plot below. My unit has the 14 bit ADC and was made in Malaysia.

[G0HZU]

It's a while since I did anything with an E4406A but the first LO dominates the phase noise of the instrument. I think the first LO uses a similar RF card to the very early ESG signal generator range, so it's no surprise to see a fairly flat -99dBc/Hz on a 10kHz span at 800MHz.
I recall that the first IF of the E4406A is up at about 321.4MHz so the system does sometimes use a divide by two on the LO output. So at lower frequencies, eg 14MHz, the close in phase noise will be at least 6dB lower due to the division. So you might see -108dBc/Hz on a 10kHz span at lower frequencies. Obviously, it will get worse up at higher frequencies because the LO gets multiplied.

You might see a fairly flat -87dBc/Hz on narrow spans at the higher frequencies as the LO can be doubled twice in some cases. I see -89dBc/Hz at a 3kHz offset at 3.8GHz for example.

[rfclown]

Just to clarify, when my E4406A died, the green standby LED was still lit OK but the unit would not power on. Also, you have to remove the 150k resistors to measure them as I recall that there is some resistance in the circuit that masks the fact that one of the 150k resistors had gone open circuit. I think it measured about 180k in circuit but >100M ohm when removed.
I think it was the 150k resistor nearest the edge of the PCB that had failed on mine. I've got pictures somewhere but there's only two of them on the board. Best to replace both of them anyway.

Thank you for this info. When my unit didn't turn on the other day, the standby LED was still lit. Years ago I used these units ALL the time. I had one at work (day job), and one at home (contracting). By the time the fan on the one unit went constantly high, my day job had allowed me to get a FieldFox which replaced what I'd been using the E4406A for, so I brought it home and stuck it in a closet where it still sits. The main reason I don't fire up the E4406A much anymore is that I've been working on more wideband stuff the past 4 years. At my present day job I have MXAs, and at home I use a Pluto for wideband signals. For narrowband I usually turn on my spectrum analyzers; only usually turn on the E4406A when I want baseband I/Q.