HP8568B calibrator noise / spurious signals

[Chris Wilson]

I have suddenly noticed the calibration signal on my HP8568B SA is showing random noise and spurious signals. I don't recall seeing this before. It comes and goes randomly every half or so second. I have had the machine powered up in standby for maybe a year, with occasional use. Is the Xtal oscillator playing up? Thanks

[Radio Wrangler]

20MHz?

I haven't driven an 8568 for some time but I didn't think the calibrator was that low.

David

[PE9ZZ_JO22KI]

Check with a different XTAL stable source (DIL XTO from scrap PC or something). If the spikes remain it's a defect elsewhere.

Tjerk, 9ZZ

[Radio Wrangler]

Yup, 20MHz it is on the 8568B

As Tjerk says, the next step is to look at an independent, clean oscillator.

Another trick is to look at the zero-Hertz response ans see what it looks like

[Chris Wilson]

I have a GPS disciplined frequency standard, and a good divider, if I feed the attenuated 10 MHz signal from the Trimble Thunderbolt to the SA, should that be clean enough as a test source? Or what about my old Marconi 2019A signal generator as a test source? thanks.

[G0HZU_JMR]

It's worth checking the AC/DC input selector switch contacts haven't gone a bit gummy/intermittent through lack of use. So try pressing the two little push buttons (back and forth) that sit between the N type and BNC RF inputs. If this internal RF connection is poor you can get some intermittent level effects that can also introduce spurious glitches. But usually it just makes the level jitter and sag a bit if the connection gets oxidised etc.

Also try scanning through all the RF input attenuator positions from 0dB to 70dB a few times for similar reasons. Also check the condition of the coax cable you are using. Is it intermittent in any way?

[Chris Wilson]

I just decided to try with the Trimble as a source with 30dB attenuation. Here's a single shot screenshot, I am posting the worst from several shots:

[Chris Wilson]

You guys really are the business! I am not sure if it was the AC / DC switch, or one of the attenuator relays, but on the DC input (which I am always terrified to use) the noise was pretty much gone. I rotated the knob set to manual attenuation a dozen or so times, and cycled the AC / DC switch a couple of dozen times and the noise is all but gone. I confess the machine has sat in standby for 3 months as I have been tied up with work and not had time for play. I guess, like cars, it should be used more. Here's hoping the fix is this simple. Thanks Jeremy et al!!

This shot is with the internal calibrator. Still curious as to whether the TB is considered a clean source though, just for curiosities sake.

I took the lid off when I first got the HP8568B, and VERY quickly nailed it back on again <shudder>

[Radio Wrangler]

At some point the thing will fail on you giving horribly scary messages about bits being unlocked. The usual cause are dried-out tantalum electrolytics on the YIG coil driver board.

One of those little ESR meters and a bit of time will sort it for you.

I've got an 8566B in the round tuit pile.

David

[PE9ZZ_JO22KI]

Your Trimble should be squeaky clean. That's the whole point of a GPS DO (Disciplined Oscillator). Have one myself (VE2ZAZ design) and it's clean down to a millihertz. As should the Trimble be.

Have phun with those electrolytics...

Tjerk, 9ZZ

[G0HZU_JMR]

I thought it might be this...

My HP8568B is prone to intermittent issues with the attenuator and the AC/DC switching if it sits unused for a few months.

My Advantest TR4172 is just the same in terms of the attenuator selectors and the AC/DC switchover. In fact the reason I was able to buy the Advantest from the company (really cheaply) about 10-12yrs ago was because someone said it was intermittent and it was written off. All I had to do to get it to work cleanly again was cycle the internal relay switches several times.

Note:
One thing I would always advise you to do before attempting to run the internal auto error correction/calibration is to cycle through all of the attenuator settings several times and also toggle the AC/DC about a dozen times before you run the auto error correction routine.

This is because the analyser will try to calibrate out any intermittent effects and you can easily get a duff calibration stored in memory.

Note: To gain access to the 0dB attenuator setting you have to type this in manually. eg [ATTEN] [0] [dB]

HP deny the user from easy/accidental access to this 0dB attenuator setting when casually using the up/down keys. You can only get it with the above key sequence. They do it like this because the 0dB setting makes the analyser more vulnerable to external DC or RF damage. So they make the 0dB setting a deliberate action by the user by making them press the longer key sequence.

[Radio Wrangler]

Also on the 8566 and 8568 analysers shift-G and shift-H key presses turn on and off trace averaging. The pull-out plastic crib sheets mention this, but if the sheets are gone....

David

[Chris Wilson]

Not sure if I am now paranoid about this thing, this is a screen shot from this evening, looking at my Trimble Thunderbolt through a 30dB attenuator, with Max Hold on for about 3 minutes. Is this OK? Still see the occasional jittering noise either side of the signal. Thanks!

[Radio Wrangler]

I don't know the Trimble specifically, but GPS outputs are normally intended for long term timekeeping and precision frequency calibration purposes. This is not the same thing as sitting continuously on the right frequency with no jitters, it just means that the long term average is very precise.

Most of the 'Atomic clock' frequency standards (Cesium) have got terrible phase noise sidebands though definitive long term precision. In other words, their short term random dither is much worse than a reasonable crystal oscillator.
In this case the cesium standards in orbit will have crystal oscillators phase locked to them with slow loops to give the short term stability of quartz with the long-term stability of a cesium beam device.

Back on Earth, GPS receivers with frequency/time standard outputs have to correct for the doppler shift of satellites and average several together. So the local 10MHz or 1 second ticks are computed locally and tend to have some jumps, though the computation gets their very long term accuracy right.

To do a good GPS freq reference, you need a slow loop locking a top quality ovened crystal oscillator like an HP 10811A

Forget fancy GPS sources, try a medium quality quartz oscillator but beware, some have got especially crappy PLLs inside them to save manufacturers making too many different quartz element frequencies. They just make a few dozen and get hundreds more frequencies by programming. An 8568 will tell the difference between a crystal oscillator and a PLL 'assisted' one. The manufacturers are very coy and don't make thiis obvious.

David

[G0HZU_JMR]

This is how my HP8568B behaves with the same test and analyser span etc. Because the span and RBW is so wide I'm just using a Marconi 2019A sig gen as the source and the MAX HOLD feature has been running for a few minutes.

[PE9ZZ_JO22KI]

@David: It's a bit off topic but I believe you do the Trimble GPS DO injustice. As said before, I have a VE2ZAZ design and attached image shows its performance. I should mention it also contains a high stability Isotemp OCXO. But: note the span!

Tjerk, 9ZZ

[Radio Wrangler]

Thanks, Tjerk. That looks good. So it already has a good OCXO and a slow loop to clean it up.

David

[PE9ZZ_JO22KI]

Yup. OM Bertrand made a nice writeup about it. It's a loop with a twist, if you will. Otherwise you never get the accuracy.

Tjerk, 9ZZ

[Chris Wilson]

Sadly today the noise is back. No amount of switching attenuator settings or between the two input sockets seems to affect it noticeably. With great trepidation I am asking if I should pull the cover and try something? The "something" would require pretty explicit instruction, but if something fairly benign could be tried I guess I'd give it a shot. Thanks!

[Radio Wrangler]

First step, look at the zero hertz response. Does this show the noises?

Replicate the picture in your first post, only with the centre frequency set to zero.

David