53310A Sanity check

[PA0PBZ]

Got myself a new toy, a 53310A MDA which I want to use to check my 'collection' of rubidium and GPSDO frequency standards. The instrument looks almost mint and has all options (001, 010, 030 and 031). I checked the caps in the power supply and didn't see anything alarming. Anyway, after connecting an UCCM GPSDO the std dev was rather high, something like 30 mHz, so I hooked port A to the reference out and got this:



Still 10 mHz, and when I do the same on port B it is even higher (15 mHz). Should that not be (much closer to) zero? Maybe I'm missing something here or maybe the the instrument is not 100%? Can anyone repeat this (reference out to port A and B) and see what you get?

[maxwell3e10]

The resolution depends on the time interval used for each measurement, which is set in the sampling menu. I've recently done a thorough characterization of HP53310 in this thread:

https://www.eevblog.com/forum/metrology/hp53310a-and-other-counter-allan-deviation/

[kj7e]

I believe option 31 (High Resolution 2.5 GHz Input) supersedes option 30 (2.5 GHz Input), I don't think you can have both.

The shorter the Time Base setting, the more deviation will be shown with the same signal.

Ill put mine on my GPSDO later and take a screen shot and post the settings for comparison with yours, but from what I can see, it looks fine.

One test you can do for fun, feed the GPSDO into the external reference in, and also to the A input then play with the Time Base, you will see how it affects the readings.

[PA0PBZ]

The resolution depends on the time interval used for each measurement, which is set in the sampling menu. I've recently done a thorough characterization of HP53310 in this thread:

https://www.eevblog.com/forum/metrology/hp53310a-and-other-counter-allan-deviation/

I see... I should try to understand this Allan deviation stuff better. How did you create these graphs?
So this picture (from https://www.eevblog.com/forum/testgear/new-gpsdo-bg7tbl/msg1417227/#msg1417227 ) doesn't tell you anything without knowing the time interval?

I believe option 31 (High Resolution 2.5 GHz Input) supersedes option 30 (2.5 GHz Input), I don't think you can have both.

Yes, you can have both but you can't have 31 without 30 because 31 adds the high res (basically a down converter, without that it just divides down) to the C port:

One test you can do for fun, feed the GPSDO into the external reference in, and also to the A input then play with the Time Base, you will see how it affects the readings.

I used the ref output of the instrument and indeed, playing with the time base makes a big difference. So posting histogram pictures (like above) without the time base settings are not very useful. Now all that remains is the difference between the A and B port, is that normal?

[maxwell3e10]

If you press the Sampling button and take the same picture, it will be more informative. Its not that unusual to get slightly different standard deviation from one shot to the next or from one channel to another.

[TurboTom]

Just to fill in the "Sanity Check" data base, I pulled my 53310A off the shelf and did some tests as well. First of all, it's good practice to take screenshots (either via GPIB or photo) after pressing the "Status" button since this will display all relevant configuration info of the MDA at the right menu area.

I found that the "quality" of the input signal can make a big difference regarding the standard deviation of the measured frequency. Always chose your input and configure it as required. I used in all the tests the "find 50% threshold" function to get good trigger quality. I also found that a fast square signal at the input will provide a significantly better standard deviation than a sine signal. Since the internal standard timebase is more than an order of magnitude worse jitter-wise than a high precision external reference (I used a Rubidium reference equipped with an Efratom LPRO101), it's always a good idea to use a good external one. I cannot tell for the internal ovenized oscillator option of the 53310A since my specimen came without it.

In order to sanity check the MDA stand-alone, it's sufficient to route the "Ref out" at the rear of the instrument to the input(s). Even without an external reference connected, I found the standard deviation in fast histogram mode to be in the ballpark of 10mHz (timebase is irrelevant in this mode).

I did the following tests as shown in the attached photos in sequence:

1. Channel A connected to REF Out at the rear of the MDA. REF In connected to a 7dBm sine output of Rubidium source. Fast histogram mode.

2. Channel B connected to REF Out at the rear of the MDA. REF In connected to a 7dBm sine output of Rubidium source. Fast histogram mode.

3. Channel A connected to 5Vpp square output of Rubidium source. REF In left open (internal standard reference of MDA used). Fast histogram mode.

4. Channel A connected to 5Vpp square output of Rubidium source. REF In connected to a 7dBm sine output of Rubidium source. Fast histogram mode.

Findings: For all practical means, there's no difference between A and B inputs on my MDA. Standard deviation if measuring REF Out differs only slightly if external or internal reference oscillator is used (~10mHz vs. ~7.5mHz). If internal reference is used and external highly accurate signal is measured, standard deviation due to jitter of the reference is way worse (~250mHz in case of the standard internal reference). Signal quality at the input, especially level and rise/fall times (obviously) make a difference in signal jitter (7dBm sine ~7.5mHz, 5Vpp square ~5.3mHz)

Previous experiments and measurements with this instrument had revealed that the MDA is so sensitive that the BNC cables used can make a difference, even though relatively high level signals are measured. For certain, highly accurate timing measurements, there are few - if any - modern replacements available that perform better. Especially considering the "street prices" one has to pay for the MDA... 

Cheers,
Thomas

P.S. Options 30 and 31 are mutually exclusive, yet Option 31 provides the functionality of Option 30, though with different hardware.

[PA0PBZ]

Thanks Thomas, I'll try to do the same and come back with the results.

[PA0PBZ]

Well, that was interesting. I now get the same results as Thomas but at first it was not even close. Then I noticed that his MDA has an auto sample interval of 1 mS while mine uses 500 uS. I have no idea why there is this difference but when I changed the interval on my MDA to 1 mS the results are equal to his.

[TurboTom]

Of course, the sample interval will make a big difference. Funny thing is that the automatic selection apparently doesn't always chose the same value. During consequent tests, my MDA also defaulted to 500µs (even though the same input signal was present) so I had to change the parameter manually. Anyway, since your readings are similar to mine, we've either both got good units or both faulty ones 

Cheers and happy Easter,

Thomas

[kj7e]

Ran the 10Mhz OCXO internal ref output to input A and B to compare the inputs.  All other settings the same, 500ms sample interval, 1mHz span, 100 samples.

Input A;




Input B;




And then my GPSDO output, 500ms sample interval, 50mHz span, 100 samples on the A input;

[TurboTom]

@kj7e

I can duplicate the figures with the input fed from the REF Out at the rear of the MDA with your configuration. Since I used my Rb standard as a reference for this measurement, the figures were slightly better (about 3.5µHz standard deviation on both channels). Using the internal standard reference of my MDA, and the instrument fully warmed up, and the Rb oscillator signal fed to input A, I get a standard deviation of about 50mHz. This shows how much better the ovenized reference oscillator (Option 10) is compared to the standard one. Yet, using a really good external reference can improve the clock jitter by almost the same factor as the step from the standard reference to the internal ovenized one.

Cheers,
Thomas

[PA0PBZ]

I'm still missing something here  Same setup as KJ7E's first picture, the histogram looks a lot more stable but the std dev I get is a lot larger (96.4 compared to 6.4).
So any suggestions where I screw up please?

[kj7e]

Try a tighter span (in the Vertical menu), IIRC I had a 1mHz span when comparing the A and B inputs while measuring the internal 10MHz ref output.

Edit, looking at my post I mentioned it was set to 1mHz span.

[PA0PBZ]

Mine was also on 1 mHz span, look at the min and max frequency of the histogram.

[kj7e]

Mine was also on 1 mHz span, look at the min and max frequency of the histogram.

You may want to double check that as the plot does not look right.  If you were at 1mHz span, the plot would be much wider for deviation indicated.  Its harder to go by the min max on the display, go to the Vertical menu and look at what it says for Span.

The min max on your display in the previous post indicates a span of 100mHz, not 1mHz

[PA0PBZ]

Thank you for pointing that out again. I noticed that you said 1mHz but I still set the MDA to 100mHz, obviously not my day
What also doesn't help is that the setting is per channel... It's now a solid 4.2 and 2.9 uHz for ports A and B.

[kj7e]

Just made a quick video showing me using the 53310A to adjust its internal reference against a GPSDO.  May give you some hits on the use of the MDA.

I posted the video here, which seems like a good general info repository for the 53310A;
https://www.eevblog.com/forum/testgear/hp-53310a-modulation-domain-analyzer-79866/msg2359143/#msg2359143