Racal Dana 1996 Frequency counter

[saturation]

Hi Scott,

http://www.tequipment.net/BK1823a.html

http://www.transcat.com/catalog/productdetail.aspx?itemnum=1823A

Wait for live chat to come on line and ask for the best price, its not advertised ... the price I received was before 2012 when everyone was doing holiday sales and end of year sales target attempts.

I don't own either, but the 1856D has less measurement functions, it more a raw counter.  The 3.5 GHz input use an N connector.

The measurement capabilities of these counters are limited to the A channel, B is used for comparisons against A, C is typically just a frequency counter.

I keep meaning to ask, where are you finding the BK counter under $470?

I was also wondering if you had an opinion on the 1856D, their 3.5 GHz counter for about $525?

Scott

[muvideo]

Saturation,
here the results, almost 4 days after trimming,
the OCXO now is slowing constantly about 1x10^-10 par day.
I stopped the logging because the lab temperature is so
low that the results are not that meaningful, but the counter
will be kept working, I will recheck the drift after some monthes
to see if it will be changed. Surely it will slow down, since it
was more stable before the trimming.

The setup:
the system I'm using seem unnecessary complex, the
reason is I have it ready for use for other purposes.
I'm using the counter in TI A->B mode,
Input A DC falling edge and input B DC rising edge COMA
(I.E. connected to input A) so I can measure the width of
a negative pulse.
I have a frontend that takes the 10MHz and GPS PPS and
emits the negative pulse.
The falling edge of this pulse is synchronized to PPS rising
edge, the rising edge is syncronyzed to the second rising
edge of the 10MHz reference that comes immediately after
GPS rising edge.
The result is a pulse that goes from 100nS to 200nS that
represents the phase between the 10MHz and PPS.
If the 10MHz is too slow it's phase will raise from 100 to 200nS
and then restart.
If the 10MHz is too fast the pulse will go from 200nS to 100nS.
The only limit is that the frequency must be enugh close to
have enough points between rollovers, otherwise the data can
be unusable or worse aliased. For example to have at least 10 points
before rollover the 10MHz need to change phase less than 10nS
per second, this translates in one part in 10^-8, i.e. less than 0.1Hz
of error on the 10MHz.

The remaining work is done by a script that unscrambles the data and applies
some processing.
I attach two graphs:
the first is phase data
- top left raw data, pulse width
- top right unscrambled data
- bottom left the same that top right, but with a linear drift removed, not very useful here
- bottom right ambient temperature
second graph:
- top left is the unscrambled data averaged
- top right is the first derivative of phase, i.e. relative error of 10MHz period,
when it's zero the frequency is spot on, now is 2.8x10^-10 i.e. the frequency is
2.8x10^-10 low, i.e. 0.0028Hz low.
- bottom left is second derivative, i.e. relative error variation per second, last
reading is 1.3x10^-15 (x86400=1.1x10^-10 per day)

Fabio.

[robrenz]

Nice work there muvideo!

[saturation]

Thanks Fabio, for the good data and your extensive work; I think you have the best characterized effect of trimming an ovenized clock on the net; others who have posted were not as thorough.

What I was surprised to find is how long the clock will take to settle, so one should not touch adjustments unless one's reference is known stable and reliable as tracking drift will require dedication for some time, weeks at least.

I am on time-nuts too [ readers here should read the archives there for the story].  It reads others have experienced adjustment instabilities but didn't document as well as you, Fabio, kudos!

Saturation,
here the results, almost 4 days after trimming,
the OCXO now is slowing constantly about 1x10^-10 par day.
I stopped the logging because the lab temperature is so
low that the results are not that meaningful, but the counter
will be kept working, I will recheck the drift after some monthes
to see if it will be changed. Surely it will slow down, since it
was more stable before the trimming.

[bingo600]

Same goes from me Fabio.

I have a Racal 1991 here , and a OPT-4E OCXO waiting to be installed in there

I have a PM6680 to test the Racal with , and a Tbolt for 10Mhz.

If you'd post the schematic of the phasedetector it would be nice

/Bingo

[muvideo]

Thank you robrenz,
saturation are you one of the participants of this topic over there (time nuts)?
Bingo, the 6680 is a very capable timer, 250ps single shot,
should be nice for interval measurements, you are well equipped for time
measurements with that and the Thunderbolt.
The schematic is the logic part of this circuit, the output is OUTD pin:



Measuring this OCXO for the first time I was feeling the real accuracy
and stability of an average Rb oscillator. So I tried to make a a comparison:
I used the same setup with an Rb, it's period error was around -1x10^-11,
the period error variation is less than 10^-16 per second -> less
than 1x10^-12 per day, but buried in the noise, so these data are
not really accurate. (See second attachment below).
The funny thing it that the Rb was powered form cold immediately
before the start of the graph, the frequency error could be trimmed
out by the serial interface interface, but my Rb has no
thermostabilization yet, so nothing to gain.
The abrupt frequency variation at 90000s reflects an internal
DDS frequency variation, probably due to some temperature
compensation, I'm still figuring out this aspect.

That's all, for now

[bingo600]

Fabio thankyou for the detailed explanation.

I'll see if i can hack a phasedetector together

/Bingo

[muvideo]

Bingo, probably you know better than me, so what I write here
is not for you but for other readers, maybe can be userful...

That circuit is only a TAC (Time to Analogue Converter)
experiment. By the way, on time nuts, in a recent
discussion i was warned to use a different approach,
there is a very nice design example over there by
Dr Bruce Griffiths.

To make phase measurements a decent counter
is enough, just use two channels in start-stop mode,
and collect results.
Start with the PPS and stop with the reference.
Any decent counter like the Racal Dana or the
philips/fluke PM6680/81 or many HP 53xx can
cope with these intervals with nS or less resolution.

Fabio.

[poodyp]

Here's the schematic if anyone needs it: http://www.schematicsunlimited.com/?z=racal_dana

It seems there's like 2 copies of the manual out there, both the same scans, but one's completely out of order. Also I have no idea how part of the page gets cut off and put on the opposite side of the page. If I could find a paper manual I'd scan it myself if they weren't almost as much as what I paid for the 1996.

[saturation]

Download open source PDF architect and you can rearrange the pages, if you have time.  These manuals aren't the best but its better than nothing, and these are widely available.  I just rearrange pages in my head, its faster.

http://www.pdfarchitect.org/

Here's the schematic if anyone needs it: http://www.schematicsunlimited.com/?z=racal_dana

It seems there's like 2 copies of the manual out there, both the same scans, but one's completely out of order. Also I have no idea how part of the page gets cut off and put on the opposite side of the page. If I could find a paper manual I'd scan it myself if they weren't almost as much as what I paid for the 1996.

[saturation]

Thanks for these Fabio, I'm not a heavy participant on time nuts as I lack time to maintain my own atomic clock, but I'm very interested in the topic; its eventually how I settled on the 1996 versus any older HP counter, at least to 1.5 GHz.  I need more volt nuts, so I spent more time posting there. 

From what you've already posted, attempting to recalibrate the 199x is going to be time consuming, and one needs already a stable reference.  As of this writing, at least the internal 1996 reference is rock stable and its measurements highly repeatable,  so I can make relative measurements with high confidence and absolute ones may only need a simple algebraic correction to offset any error from my units calibration status.

Thank you robrenz,
saturation are you one of the participants of this topic over there (time nuts)?
Bingo, the 6680 is a very capable timer, 250ps single shot,
should be nice for interval measurements, you are well equipped for time
measurements with that and the Thunderbolt.
The schematic is the logic part of this circuit, the output is OUTD pin:

That's all, for now

[PappyTG]

I've was wondering if anyone has run into error code 63 when doing the self test on your 1996.  I've had the unit I currently have for serveral years now and it's been working great other then failing this self test.  The manual indicates that the error involves the input circuit but I haven't been able to trace it out on the schematic.  One thing I did notice is that if I have the unit in standby overnight and power it up in the morning it pass the self check but after about 15 minutes I get the Error 63 again. Appreciate any input.  Plus I've been checking it daily this against my GPS and Argo Systems AS210 and I'm usually running 10.00000003 MHz ±2 counts

[saturation]

Hi,

A common issue with system errors after warm up in old gear can be traced to the power supply, its the most likely item to go from age compared to all the subsystems [ other than the EEPROM].  It may pay to check the filter caps or the ripple on the main rails.  It can save a lot of troubleshooting legwork.

I've was wondering if anyone has run into error code 63 when doing the self test on your 1996.  I've had the unit I currently have for serveral years now and it's been working great other then failing this self test.  The manual indicates that the error involves the input circuit but I haven't been able to trace it out on the schematic.  One thing I did notice is that if I have the unit in standby overnight and power it up in the morning it pass the self check but after about 15 minutes I get the Error 63 again. Appreciate any input.  Plus I've been checking it daily this against my GPS and Argo Systems AS210 and I'm usually running 10.00000003 MHz ±2 counts

[slburris]

Well, it turns out my 1992 is deaf on the channel C input.

So I went ahead and picked up the BK precision counter and
it seems to work like a charm.  It's nice to have 1 working
fully functional counter.

It's got a few things to complain about compared to the 1992.
Doing high precision measurements takes a gate time of 10sec.
The 1992 is much faster, presumably because it uses a different
measurement technique?   The internal oscillator seems more
drifty than my 1992, which has the ovenized standard.
You can't get more than 8 digits without using an external standard,
presumably due to the same drifty internal oscillator.

I'm using it now in external standard mode, connecting it to my
Trimble Thunderbolt.

The 1992 is still in service for frequencies less than 160Mhz, but
nominally goes on my list of things to repair, which is quite long now:

Racal Dana 1992
Racal Dana 1998
EIP 545A 18Ghz frequency counter
Heathkit GC-1000 clock
Tektronix 2246A scope
Tektronix 2465 scope
Heathkit H8 computer

and a few other odds and ends.

Scott

[edpalmer42]

Well, it turns out my 1992 is deaf on the channel C input.

<snip>

The 1992 is still in service for frequencies less than 160Mhz, but
nominally goes on my list of things to repair, which is quite long now:

Racal Dana 1992
Racal Dana 1998
EIP 545A 18Ghz frequency counter
Heathkit GC-1000 clock
Tektronix 2246A scope
Tektronix 2465 scope
Heathkit H8 computer

and a few other odds and ends.

Scott

In case you haven't thought of it, a quick thing to do for your deaf 1992 is to check the fuse that's hidden inside the connector.

Also, your list of things to fix seems rather short.  You're not trying very hard, are you?   

Ed

[grumpydoc]

..list of things to repair...
Racal Dana 1992
Racal Dana 1998

Take the 1.3Ghz prescaler out of the 1998 and put it in the 1992 (assuming that it's not why the 1998 needs repairing).

I'm pretty sure that they are the same board.

Service manuals readily available for both on the 'net

[saturation]

My 1996 on arrival had trouble with Channel C.  I reinserted the board, it read erratically but worked.  I then connected the 10 MHz out of the Racal into C to check the calibration, it read the signal and worked fine every since. 

[hwengineer]

Great forum. Found while researching some new equipment. I've had a 1996 for several years. I worked last time I powered it up, maybe 9 months ago. When I turned it on earlier this week to test a new pulse generator I got pretty much nada. No error codes. Some evidence a POST was start by a number of LEDs coming on for a half second and then stalling with only the GATE and both TRIG LEDs on. I see some evidence of charring around negative voltage regulator U49 but I don't think it is recent. +5 and +15V testpoints show the correct voltage.

I'm currently without a scope until January. A component failure I can maybe fix. One fear is the EPROMs have gone senile. Looking for any hints before I start digging deep. Thanks.

[saturation]

Assuming the PSU is fine, a can't hurt test is reseating all socketed boards and ICs, taking due static precautions.

Great forum. Found while researching some new equipment. I've had a 1996 for several years. I worked last time I powered it up, maybe 9 months ago. When I turned it on earlier this week to test a new pulse generator I got pretty much nada. No error codes. Some evidence a POST was start by a number of LEDs coming on for a half second and then stalling with only the GATE and both TRIG LEDs on. I see some evidence of charring around negative voltage regulator U49 but I don't think it is recent. +5 and +15V testpoints show the correct voltage.

I'm currently without a scope until January. A component failure I can maybe fix. One fear is the EPROMs have gone senile. Looking for any hints before I start digging deep. Thanks.