HP8568B calibrator noise / spurious signals

[Chris Wilson]

Hi David, thanks for replying again. First screen shot show centre frequency of zero, 1MHz span. Hard to catch maximum noise, it's constantly varying. Second shot shows Max Hold switched in, after about 10 seconds, same settings.


Some more info, noise level is similar on both input sockets. If I had to say, I would venture that switching through attenuation levels is most likely to clear the noise. I just looked at it (been running with the test signal imputed for a few hours) and the noise was gone or much less. When I scrolled through attenuation positions, just for a few times, it came back.

[Chris Wilson]

Occasionally when switching between input sockets, with the input to the N type 100kHz to 1.5GHz one, the signal is shown attenuated. Repeatable, but infrequent.

[G0HZU_JMR]

It looks like there's still something wrong then...

My max hold was much cleaner on the same test and I could leave it for ages without seeing much of a change.

My 8568B definitely has issues with the AC/DC selector if I leave it unused for a long time. The attenuator also has an issue and this is most easily seen when returning to the 10dB setting on mine. If the analyser hasn't been used for ages it will show more than 10dB attenuation for maybe half a second and then it kind of integrates up to the correct level after another second or so. It only does this on the 10dB setting so I'm pretty sure it's the attenuator.

However, it only takes about 5 or 6 clicks in and out to make it totally reliable again for a few months or so.

It looks like your analyser still has an intermittent connection/component but it might not be the attenuator because it shows the spurious at zero Hz.

[Radio Wrangler]

Yes, the aim of that test was to eliminate everything in front of the mixer.

The next logical tests are to go to zero span and look for amplitude fluctuations.

David

[Chris Wilson]

I tried zero span yesterday, but I am not sure if zero span with otherwise the current settings is an ideal check or not. Using otherwise similar settings I get an unwavering flat line.

Leaving the machine on all day (with a very dim display to stop screen burn) seems to result in a more stable general display than when it's first turned on from standby.

I am looking for a source for the service manual to better understand the signal paths.


Thanks again.

[Chris Wilson]

Service manual (set of 4) ordered from Dave at Artek Manuals for download. God help me.....

[Radio Wrangler]

Should be free on the Agilent website. Most non-current HP stuff is.

OK, in zero-span mode, you have a fixed-tuned radio receiver (rather a fancy one)

You can tune to the centre of the zero hertz response and use different bandwidths and also dB/division to look for things that could be gain fluctuations dowm the receiver chain. This would show bad connections.. cracklies

You can tune of of the zero hertz response (right off) and look for things which jump out of the noise and this would find sources of noise not dependent on signals.

You can tune down the slope of the filter response (which is the shape you see for that zero hertz response and use it as a 'Slope Demodulator' to demodulate FM. This will allow you to see FM jumps and things if something noisy is modulating the tuning of any of your LOs. A noisy opamp/resistor etc tuning your YIG or a later oscillator will show up here. The coil driver circuits are a bit prone to running hot and drying out tant capacitors.

A spectrum analyser is a fabulous machine for troubleshooting thiings and diagnosing faults. You can do a lot by turning it on itself

David

[Chris Wilson]

OK thanks David, although I must admit some of that has gone a bit over my head

I am wondering if this noise is electro- mechanical, as on a 10kHz frequency span I can gently tap the front panel and I see immediate noise spikes on the trace. I can also alter or sometimes totally stop the noise at a 1MHz span by tapping. This video sounds like I am beating the hell out of it, but the microphone gain was far too high, they are actually gentle taps on the front panel. It's not the calibrator cable or connectors, it's something internal, and I am suspecting a loose SMA or even a duff input switching relay. I am girding my loins to delve inside the brute. Maybe bypass the input switching relay altogether if nothing loose is apparent. The video is in MP4 format, rather badly focused, but should give the idea:

http://www.chriswilson.tv/8568B_noise.mp4

There's a used microwave input switching relay of the right part number for sale in the US, but is there any chance of a new one still being available?

[Radio Wrangler]

Search for the Hewlett Packard Journal archive and search for the 8568A and you should find some good reading material on how your box works and how it was designed.

If it's a mechanical crackly, responds to tapping, and it shows up on the zero-hertz response, you are well on the way.

David

[Chris Wilson]

Centre frequency 20MHz and zero span isn't noisy, and tapping makes no difference to the line's amplitude

[G0HZU_JMR]

I'm not sure how hard/soft you are tapping the RF unit but it's unrealistic to not expect some microphony effects on the first LO when you tap the RF unit. The loop bandwidth is quite wide but I don't think this will make it immune to microphony.

However, in your video I can see a few spurious terms that appear even when you are not tapping the unit. Presumably these are due to the original spurious issue.

So most of what you see in your tapping test might be 'normal' LO1 microphony and not the same thing that is causing the original spurious issue.

[Chris Wilson]

A fear of making matters worse, and a lack of time has seen me ignore this machine, plus a lack of true understanding of its very complex operation at the component level, of course!



However, I had to use the 8568B again yesterday, noise and all, and I have seen
an issue that may, or may not, shine more light on things. I was
trying to measure a signal level at 137kHz. I found the -29dBm hard to
believe, it should have been more. I then used the inbuilt 20MHz
source which is supposed to be -10dBm as a check. It showed a -10dBm
signal, as expected. I then used two different good quality sig gens
to inject signals at known levels. I used plain sine waves. Once you
get down around 5Mhz and lower in frequency the display and readout
gets more and more incorrect, reading lower and lower as frequency
gets lower. I then tried 1Ghz into it, and that was correct. So the
lower the frequency (and I could see if it changes at a particular
level, suddenly), the more low it reads. Does that give any additional
clue please?

May I wish you all a happy Christmas and a prosperous and healthy New Year
and thank you all sincerely for your patient help over this and other matters during 2014. All the very best to you and yours.

[Chris Wilson]

Further info

Accurate sig gen, sine wave out all at -10dBm

20MHz - -10dBm shown on 8568B
10MHz - -9.9dBm
5Mhz - -10.6dBm
1MHz - -14.4dBm
500kHz - -19.1dBm
100kHz - -32.6dBm
10kHz - -52.5dBm
1kHz - -5.2dBm

At 1Khz centre marker showed 0Hz, and frequency counter also showed 0Hz. Centre marker and counter correct for previous frequency inputs. When I removed the signal lead from the HP8568B with 1kHz going into it, the 8568B display kept running as if the signal was still present!

[dave cox]

20db / decade is 1st order RC. Is the analyser DC coupled ?

dc

[Chris Wilson]

I am running it with a DC block on the input Dave. Is that wrong for these tests? Running into the N type input on the right.


This one, or one very similar:

http://www.rfshop.co.uk/dc-nmnf0-5g100v50-dc-block.html

[Radio Wrangler]

It all depends on the value of the capacitor in that DC block adaptor. It will definitely cause a 20dB/decare roll off at low frequencies. THe question is where does it start?

The advert has no data about the thing itself, only the price and their stock level.

You could measure the capacitance from inner socket on one end to the pin on the other.

David

[Alistair D]

This is the datasheet for the one mentioned. From the figures quoted it is almost certainly the DC block that is causing the roll off you are experiencing.

http://www.jyebao.com.tw/pdfpath/DC-NMNF0-5G100V50.pdf

Al

[G0HZU_JMR]

A lot of DC blocks like that are spec'd from 10MHz to a few GHz. So the LF rolloff you are seeing is probably normal.

However, the N type input port on the HP8568B is already DC blocked to 50V internally so you don't normally need to worry about fitting a DC block here. Can you try without the DC block (assuming you have < 50V DC on your input connection)

Note: The alternative BNC input port to the HP8568B does not have a DC block so you can easily damage the analyser with external DC if you use this port.

You can get DC blocks that work well down to just a few kHz. For my old 22GHz HP8566B I bought a high quality DC block (N type) that works from about 100kHz to 18GHz with very low VSWR. It wasn't cheap but it does prevent damage from external DC as the HP8566B has a 0V DC limit on the input port.

[Chris Wilson]

Thanks everyone, it was the DC block, it gives readings that tally with the sig gen, level wise, with it out. I have never had cause to use it at below 10Mhz before, but my new found interest in the LF bands means I need to either be very careful with it or buy a better block like Jeremy's.

I was half hoping it was a part of the noise issue and would help to locate the problem. Mea culpa, apologies.

[Radio Wrangler]

A block with a larger capacitor will be fine for your LF work because you already have that >10MHz job for when you want to go high. THe bigger capacitor will go bad at a lower frequency than the smaller capacitance one will. A lot of my blocks and accessories are in APC3.5 which makes N look economical.

David