HP 5335A Timer / Counter - Anything i should know?

[tkamiya]

Ah, thank you. 

I have naked rubidium oscillators as well as cased ones from Efratom.  On one version, adjustment hole wasn't available as it was completely covered with heatsink (without an access hole!)  So I did the same thing.  Glued a multi-turn pot to inside and punched a small access hole on top cabinet.  Just for a good measure, put a sticker on it.

I need to do your modification to my 5335... 

[0culus]

I might have exponent expression wrong.  My measurement was : 10,000,000,00x where x changes.  That's 10^-9? or 10^-10?  Input is 10MHz.  I always mess this up!

The GPS I have is PRS-10 and FRS-C.  I intend to use former and slave it to GPS.  To avoid any interactions, power supply will be linear and external.  So it'll be shielded from temp change and magnetic field.   

I am revamping my house standard.

I was also surprised how much Rb drifts.  It took good 24 hours to completely stabilize.

That change would be x * 10^-10.
The un-modified 5335A has 4s max. Gate Time, so only 10^-9 will be displayed/resolved.
So maybe you should have a 2nd look, what you really measured.

My FRS-C needs about one week, until it has stabilized to parts in 10^-11.. that drift-time always depends on the limit criterion used.
I have provided an external potentiometer to adjust the Rb to 2*10^-12, but drifts are on the order of 1*10^-11.

You can open the gate manually to get the full digit capability. I think there might be a way to control that over HP-IB?

[Dr. Frank]

Yeah, both ways work.

It was a pleasure for me to discuss and measure, as I learnt a lot new about the  behaviour of the 10811.

PS.. and my 5335A now got a genuine Cal Sticker, as it's now freshly adjusted to 3*10^-10

[tkamiya]

You had me confused for half a day....

The reason I have 10 digit number is because my setup is as follows.

DUT is HP5335A with 10811.  It's clock out is connected to ANOTHER COUNTER
This ANOTHER COUNTER gets its clock from redundant GPSDO OXCO and/or Rb

So I am just treating HP5335A as an oscillator. 

I've also done the other way around as well.  But I gain an extra digit doing it the way I described it.  Also it's newer and faster. 

I've never ever looked into behavior of oscillator this much, and witness its drift.  Case open/close thing was an eye opener.  Surprising how sensitive it is!  So, Dr. Frank, are you satisfied with performance of my 10811 now that we know it's actually 10^-10, not -9?  Should I pursue this further?  Like I said earlier, in a week or so, I'll have another HP5335A.  It will be fun to compare responses side-by-side.  Also in planning stage is Rb with GPS.  I'll have to do something with ambient temperature changes.  My lab is the hottest room in the house.  I'll have to make a small compartment and insulate it to see what it will do.

Thank you very much for your instructions.  It helped great deal.

[Dr. Frank]

You had me confused for half a day....

The reason I have 10 digit number is because my setup is as follows.

DUT is HP5335A with 10811.  It's clock out is connected to ANOTHER COUNTER
This ANOTHER COUNTER gets its clock from redundant GPSDO OXCO and/or Rb

So I am just treating HP5335A as an oscillator. 

I've also done the other way around as well.  But I gain an extra digit doing it the way I described it.  Also it's newer and faster. 

I've never ever looked into behavior of oscillator this much, and witness its drift.  Case open/close thing was an eye opener.  Surprising how sensitive it is!  So, Dr. Frank, are you satisfied with performance of my 10811 now that we know it's actually 10^-10, not -9?  Should I pursue this further?  Like I said earlier, in a week or so, I'll have another HP5335A.  It will be fun to compare responses side-by-side.  Also in planning stage is Rb with GPS.  I'll have to do something with ambient temperature changes.  My lab is the hottest room in the house.  I'll have to make a small compartment and insulate it to see what it will do.

Thank you very much for your instructions.  It helped great deal.

Sorry, that I confused you, I was not aware of that. 

Anyhow, your 10811 is definitely working fine, as thermal drift is down in the 10^-10 ballpark. At that resolution it was more or less a go/nogo test of the oven, what we have discussed so far, and a bit beyond, very interesting also for me!

If you want to dig deeper into the performance, you'll have to use different techniques and evaluations.. It's then all about 'Stability'.
So you might compare the 10811 against the GPSDO or Rb, using Allan Deviation (ADEV) statistics. This gives you the relative stability figures on different timescales.
"Relative" is important, because on short time scales, most free running OCXO are superior over GPSDO, and even Rb, due to the high jitter of the GPSDO signal, and of the noisy Rb signal. See ADEV diagrams from different time-nuts, how the different standards perform, relative to MASER or Cs standards.

http://www.leapsecond.com/museum/manyadev.gif
http://www.leapsecond.com/pages/adev-avg/

The residual x*10^-10 variance might also arise from GPSDO instability, instead from the 10811 itself, if you were using effective Gate time of 1sec, or less, only.

You may instead use time interval methods, on 1pps signals derived from these standards, and a resolution of 1ns for T.I. (like the 5335A) is already  sufficient for that.

Do you have a Fluke /  Philipps counter, or an agilent 53131/132A, or even a KS 53220/230A? These feature better interpolation techniques, like 100ps, or down to even 20ps single shot, like the very old HP5370B, or the SR620.. That would improve / speed up the ADEV measurement.

Oh yes, again, TimeLab from kfe5x is highly recommended for that. There you'll find also several ADEV graphs for various standards.
http://www.ke5fx.com/timelab/readme.htm

Looking forward to your next 5335A, and picture of this mysterious 10811 3010 series.
Frank

[tkamiya]

I have HP53131A hooked into GPSDO. 

Problem with Allen deviation is I really don't understand it and I lack facility to actually measure that kind of precise and repetitive measurement.  I've purchased GPIB interface, so I might try it at some point in time.  For this experiment, I used HP53131A in frequency A mode and used long gate time.

Here's a picture of 3010.  Other than these two mysterious stickers, they are identical to everyday 10811-6011.  It came out of parted out HP8568A upgraded to B.  Right now, these are available for like 60 dollars so I may buy a few and keep them in my stock. 

I wish I can make use of my Rubidium stuff but just their size and heatsink requirement, and power is stopping me.  It works with switcher but it is really recommended to use a linear supply.  I had 3 of them running concurrently and it definitely made the room much warmer, so that's a problem, too.

I think I'm becoming a time-nuts....  I'm already a test equipment anonymous.

[FriedLogic]

Plotting ADEV and similar is only part of the analysis. They are designed to look at random noise processes, but there are a lot of systematic processes also going on. These may show up on an ADEV plot, but it's not usually the best way to investigate them.

Oscillators can be sensitive to temperature, pressure, humidity, magnetic fields, etc., and the best way to check for this is usually to keep a log of them to add to plots of the frequency and phase. These can be processed in various ways by averaging, removing drift, measuring correlation, etc. to understand the behaviour of the oscillator in different situations, and so give the best chance of optimising its performance. The effects can also be delayed, which further complicates things.

Although ADEV is often quoted for short times like 1-100s, it's valid over any time period, so you can still get some use from it even it you don't have a very fast and high resolution counter.

One slight complication with the 53131A/53132A/53181A is that their frequency modes do not do a single 'Gate Time' measurement, but rather do what is effectively a lot of overlapping 'Gate Time'/2 measurements. This adds some filtering, and also a triangular weighting to measurement, and the ADEV of measurements done with them will be low compared with the actual ADEV. It is still useful for comparing with measurements done on the same or similar counters.

[tkamiya]

OCXO is good at short term.  Not so at long term.
Rb is good at long term.  Not so at short term.
Cs is great but not so with my budget.
Hm is super great but not even a possibility at home.

I am going to experiment with GPS trained Rb trained OCXO.  That's the best I can do.

I wonder how do we even know Cs is wrong when it IS the standard....

[Dr. Frank]

OCXO is good at short term.  Not so at long term.
Rb is good at long term.  Not so at short term.
Cs is great but not so with my budget.
Hm is super great but not even a possibility at home.

I am going to experiment with GPS trained Rb trained OCXO.  That's the best I can do.

I wonder how do we even know Cs is wrong when it IS the standard....

Cs, like Rb has a small dependency on magnetic fields, so you have to cancel these externally.
This dependency is much, much smaller than Rb.

The primary clocks were all compared against each other, and that median will define the UTC.

Comparison is done by bi-directional satellite links, or by GPS.

PS: You already have all tools on hand to make stability studies, and to analyse the collected data on any wanted way, ADEV, MADEM, freq. deviation.. all inside TimeLab, which also has a complete data collection routine for these commonly used counters.

GPIB access, and a few BNC cables, and you're done. One of your standards has 1pps for sure, so that would be a good starting point.

It's much easier to do, than you think!

[tkamiya]

What I really need is a GOOD understanding of time device management.  That inter-dependency of nature (Cs vs Rb vs XO) of medium (short term vs long vs very long) also needs to be clear.  I'm going to have to get the setup ready and follow along existing resources such as TimeLab.  Since there does not seem to be one "best method", I'm thinking there won't be one device that does all. 

Maybe I'll make two identical setup and do a controlled comparison study.  I have a stack of Rubidium modules.  It should be quite interesting. 

[tkamiya]

A new HP5335A with OCXO and 1.3GHz option just arrived.  It is in beautiful shape and seems to work as it should.  I will be keeping it plugged in for a day or two and see where it settles.

My goal is to have two of each so I can do A/B testing easier. 

[FriedLogic]

Maybe I'll make two identical setup and do a controlled comparison study.  I have a stack of Rubidium modules.  It should be quite interesting.

That's the way to go. Once you have a basic test rig set up you can build on that, then decide what performance you want to aim for and how to measure it. Once you get it up and running you'll get a better idea of where to go next.

If you're trying to squeeze the best performance from a rubidium, temperature is normally the most important thing, followed a fair bit after by air pressure - but that's more a discussion for one of your rubidium threads rather than this one.

[tkamiya]

Dr. Frank,

As I mentioned earlier, I acquired another HP5335A and this time, it comes with factory oven crystal, 10811-6011.  After overnight warm up with case closed, I opened it this morning and immediately touched top and side of the module. 

They are only very very slightly warm in certain spots.  Other parts are absolutely same temp as air around it.  This unit still has a factory fan.  (a turbo jet!)

So it is confirmed that IS the expected behavior.  HP has designed a duct system that keeps the Xtal cool.  I am not going to keep the case open to validate how hot it will get.  I think it's unnecessary.

I thought I'd report in and leave this, in case someone else wonders in future.

[tkamiya]

Fried Logic....  what an interesting handle!

To do an A/B comparison, I'd like to keep everything else the same except the element I'm testing.  I know each counter has certain characteristics since they are so heavily software involved.  I've learned they do quite a bit of massaging and post processing before the data gets on display panel. 

Besides....  this unit was very inexpensive on eBay.  I couldn't refuse....

[LazyJack]

Yes sorry, I'm a bit late.
I had to study all the change history in manual 05335-90044, up to Serial Prefix 3154.

V1.1 had been introduced in June 1984, for all instrument with 2424A06001 and above.

V1.1  has one decimal point in the HP-IB string, as shown with my instrument.
V1.0 shows no decimal point in the GP IB string at all.

I already have the newer prescaler  in the C-channel (05335-60034), which divides by 64.
The old prescaler (05335-60009) divided by 20 only, therefore the firmware has to match.

It's not clear, if there exists an intermediate firmware which corrects all the bugs of V1.0, but is made for the old prescaler, i.e. multiplies the result by 20 instead of 64.

The difference may be inside the EPROM A4U22, as 05335-80012 (as in my unit) refers to the new /64  prescaler, and 05335-80006 refers to the old /20 prescaler.
That's change 7 in this manual, and affects prefix 2820, exactly my counter, and also for retrofitting the newer C-channel option 030.


Obviously, I have an intermediate A4 version, 2 EPROMs - but V1.1 and the newer prescaler. Perhaps my instrument had been updated, or the C - Input had been replaced.

I could not find a comprehensive table, which EPROM code refers to which firmware version, and/or to which prescaler.
Frank

Sorry to resurrect this thread, but I found this table. It is in the latest service manual (05335-90444, recently uploaded to KO4BB). It's on top of page 6-9 .

[mindcrime]

Hey guys, I was just looking at buying an HP5335A that comes with options 010 and 040, but not 030. I'm tempted, but am wondering - if I were to buy it, how hard would it be to find the parts (and anything else needed) for the option 030? The device is probably worth the money even without it, but that would be a really nice addition...

Or would I be better off holding out for a 53131A or similar, for which there seem to be plenty of option 030 cards (including clones) that are readily available?

[LazyJack]

You can sometimes get the opt030 kit on ebay. It is a board plus the front panel. But be careful. There are two versions of it, one with a divide by 20 and the other is a divide by 64 prescaler. The later is for newer models, with v1.1 firmware.
I have one 5335a with all options. A good instrument, but quite large and somewhat loud.

[mindcrime]

You can sometimes get the opt030 kit on ebay. It is a board plus the front panel. But be careful. There are two versions of it, one with a divide by 20 and the other is a divide by 64 prescaler. The later is for newer models, with v1.1 firmware.
I have one 5335a with all options. A good instrument, but quite large and somewhat loud.

Yeah, now that you mention it, the size factor is probably the biggest reason I didn't already pull the trigger on buying the beast. I'm kinda out of room for test equipment anyway, especially something that big. I think I'll hold off for now.

[OH2LIY]

Hi, I did build this opt.30 using what I could find my bonk storage. I works nicely


73 Ramppa

[LazyJack]

Very nice! Thanks for sharing the design.

[TERRA Operative]

So my turn to get one of these up and running again... 
I picked one up from Yahoo Auctions with options 010, 030 and 040 fitted. So far it seems to work ok, pending the usual relay/fan fix and whatnot.

I also need to replace the red 7-seg display cover as someone in the past must have used the wrong cleaner which caused it to warp severely and fog up badly. If anyone has a spare one of these (plus a few of the little black clip thingos), I'll be happy to buy it from you.

My question at the moment though, is my unit seems to have some extra blanked-off positions on the rear panel that don't seem to exist in any of the manuals I've found online.
I assume the 4 in the center are for moving the front panel connections to the rear, but does anyone know what the ones I circled in red are for? Maybe for rear panel connections to option 020?

[LazyJack]

Yes, they are the rear panel options for the Voltmeter option. These options apparently were only available on later units, as older ones do not have the cutouts for the rear connectors.
I think the rear connections are option C10. They are connected in parallel with the front connector, so they will decrease the bandwidth, there is a service note about it.

[David Aurora]

(Necroposting just to possibly keep useful 5335A info in one place)

I just picked up one of these last week from a forum member, thought to be working but it's a bit iffy on first tests. Just trying to work out whether these are normal quirks or if I should make some time to dive in properly and do some repairs.

First up is the startup condition- manual says it should boot straight into counting frequency on input A. I *think* I've seen it do that once maybe, but generally it doesn't. I get nothing until I manually press both COM A and auto trigger, then it works fine. Is that normal?

The other thing is rise/fall/slew modes. Period, duty etc all seem to measure and display fine, but I get nothing from rise/fall or slew modes. I shoved in the example signal in the manual (I think it was something like 400nS rise/fall, 10µS period, 1VPP, it's not in front of me right now), set it up as described and nada. Tried playing with all sorts of settings and I still got nothing. If I check trigger level it shows the 10/90% trigger points, so obviously it's measuring/calculating something at least, but for whatever reason it doesn't display the rise/fall/slew rates on the display. Am I missing something about using these modes?

[David Aurora]

(Necroposting just to possibly keep useful 5335A info in one place)

I just picked up one of these last week from a forum member, thought to be working but it's a bit iffy on first tests. Just trying to work out whether these are normal quirks or if I should make some time to dive in properly and do some repairs.

First up is the startup condition- manual says it should boot straight into counting frequency on input A. I *think* I've seen it do that once maybe, but generally it doesn't. I get nothing until I manually press both COM A and auto trigger, then it works fine. Is that normal?

The other thing is rise/fall/slew modes. Period, duty etc all seem to measure and display fine, but I get nothing from rise/fall or slew modes. I shoved in the example signal in the manual (I think it was something like 400nS rise/fall, 10µS period, 1VPP, it's not in front of me right now), set it up as described and nada. Tried playing with all sorts of settings and I still got nothing. If I check trigger level it shows the 10/90% trigger points, so obviously it's measuring/calculating something at least, but for whatever reason it doesn't display the rise/fall/slew rates on the display. Am I missing something about using these modes?

Scratch that, just popped it open and things started making sense. Channels were cross patched, not sure why but I'm guessing because of a fault, other bodges present too. Looks like I've got work to do

[TERRA Operative]

Have fun, and be sure to do the relay/relay socket/12V fan mod for reliable operation.

Also, there is an electrolytic cap on the PSU section, just next to the small green edge connector that I have found is really needing replacement in 2 out of 2 units I've poked at so far.