HP5245L Freq. Counter Repair

[gmilliorn]

I'm trying to repair an old HP5245L (Nixie) freq. counter.  The counter
part works fine, the oscillator does not.  The A26 card provides the
reference clock at pin 1-5, it is 0V.

Here's the schematic:

IMG is not working for me - here's a link:

The voltages are all pretty close to what HP noted:
        C         B       E
Q1   -13.4   -8.0   -7.2
Q2   -10.6   -3.3   -2.6
Q3   -10.5   -2.6   -1.9
Q4   -6.5     -2.3   -1.4

Q1 was furthest off, and since the AGC feedback is AC coupled I keep
thinking it must be around Q1.  I replaced it with a 2n2907 which seems
pretty close (detailed specs on the 1853-0009 transistor seem sparse
other than PNP + 30V).  No fix, though.


I removed the (A26) card and powered it with a bench supply and
a 1MHz xtal across J2-1 to J2-4, with no change, so I'm pretty
sure it is localized to this card.

L1 is not open and pulled C6 and it measured .064 uF.  I have to pull
the parts to check them so I haven't done everything.

If anyone has ideas what to check I'd sure appreciate it!
Gary
 

[mij59]

Hi,

How did you measure the output pin 1-5 ?

[gmilliorn]

With a scope and 10X probe.  Do you think I'm affecting it, somehow?

The HP manual shows waveforms, but unfortunately not for the oscillator
side of the card.  The pin 1-5 output of the card should be about the same
as the external input option, which is 2V pk-pk max.

I'm not even sure of the oscillator type; the L-C arrangement doesn't
resemble what I can find online, though I guess Colpitts is most similar.

-Gary

[mij59]

With a scope and 10X probe.  Do you think I'm affecting it, somehow?

The HP manual shows waveforms, but unfortunately not for the oscillator
side of the card.  The pin 1-5 output of the card should be about the same
as the external input option, which is 2V pk-pk max.

I'm not even sure of the oscillator type; the L-C arrangement doesn't
resemble what I can find online, though I guess Colpitts is most similar.

-Gary

I don't think probing the output would cause the oscillator to stop.
If you connect a xtal you also need connect the capacitors as shown in the schematic.

Check the connections between the A26 and A24 board.
Check board A24 for bad solder joints.
Trimmers can lose capacitance due to oxidation, check C1, C2, C3.



 

[gmilliorn]

"Trimmers can lose capacitance due to oxidation, check C1, C2, C3"

I didn't know that, thanks - I'll look into that.  C2/C3 are accessible, C1
is inside the oven which I was hoping I don't have to open that up.

Building up a proper set of caps on the crystal is a good idea, thanks, I
shouldn't have tried to skip that (thought I would at least see a burst).

(And 0.017uF ? Odd, I don't think I've ever seen that value before).

-Gary

[Rasz]

Here's the schematic:

IMG is not working for me - here's a link:

just upload it here as an attachment
imgur will delete this picture in few months and we will end up with a post with dead link.

[gmilliorn]

Thanks, here's that image using attach.

I tried hooking up the multiple capacitors on the crystal I'm connecting, it still
doesn't work.

Interestingly, when I short from the AGC node (CR1-CR2 common) to the base
of Q2, I get an oscillation: about 57kHz 2V pk-pk.  That seems to suggest the
Q2-Q3-Q4 path is at least OK, and that it is Q1 or thereabouts.

I suppose it is possible the 2N2907 is not a good substitute for the HP 1853-0009.
Sphere has them for $7 ea. which I was trying to avoid.

[mij59]

Thanks, here's that image using attach.

I tried hooking up the multiple capacitors on the crystal I'm connecting, it still
doesn't work.

Interestingly, when I short from the AGC node (CR1-CR2 common) to the base
of Q2, I get an oscillation: about 57kHz 2V pk-pk.  That seems to suggest the
Q2-Q3-Q4 path is at least OK, and that it is Q1 or thereabouts.

I suppose it is possible the 2N2907 is not a good substitute for the HP 1853-0009.
Sphere has them for $7 ea. which I was trying to avoid.

Did you check the reaming capacitors round Q1, CR1 and CR2?
Maybe you could use  transistors like BFxxx

[gmilliorn]

I guess at this point I need to start pulling off all the parts and checking them one
by one.
I have an old capacitance meter, it can't measure them in-circuit (if that's even possible).

It might also be the 2n2907's, the only cross references to HP 1853-0009 I could
find are:
1853-0009 = TI SM5727
TI SM9477
Mot SS2111
Fairchild S15097

But all those part numbers are odd, probably either custom for HP or the gov.

[mij59]

I guess at this point I need to start pulling off all the parts and checking them one
by one.
I have an old capacitance meter, it can't measure them in-circuit (if that's even possible).

It might also be the 2n2907's, the only cross references to HP 1853-0009 I could
find are:
1853-0009 = TI SM5727
TI SM9477
Mot SS2111
Fairchild S15097

But all those part numbers are odd, probably either custom for HP or the gov.

In-circuit testing can be tricky, in this case,  remove the components to test them.

I found a cross reference list , sadly the 1853 0009 is not on it, but the 1853 0008 and 1853 0007 are.
It's a bit of a gamble, maybe the 2N3251 is a replacement for the 1853 0009.

 
 

[gmilliorn]

I've started pulling caps and checking them; a few are slightly off.  Perhaps I should just
replace all the discretes as that is cheap.  The inductors are not open, and I don't have an
LCR meter or a sine wave oscillator so I can't think of a way to check those out.

That leaves Q1 yet again: since it seems like Q2-Q4 are working, I might move Q3 to Q1 and
put the 2N2907 there.

To try another transistor, I assume the parameter that is next most important (after 30V
range) is high gain.   I really wish HP had shown a few waveforms on the left side.

[mij59]

I've started pulling caps and checking them; a few are slightly off.  Perhaps I should just
replace all the discretes as that is cheap.  The inductors are not open, and I don't have an
LCR meter or a sine wave oscillator so I can't think of a way to check those out.

That leaves Q1 yet again: since it seems like Q2-Q4 are working, I might move Q3 to Q1 and
put the 2N2907 there.

To try another transistor, I assume the parameter that is next most important (after 30V
range) is high gain.   I really wish HP had shown a few waveforms on the left side.

If the components test good there's no need to replace them.
The output level of Q1 could be small, at what sensitivity setting of the oscilloscope did you test Q1 ?

[gmilliorn]

Removed at least one lead and measured a lot of parts, and made the following
changes:
  * Swapped Q1 and Q2
  * Replaced R8 (was 4600)
  * Swapped C6 for C3, C6 was 0.056uF and I had no spares.

Now I have some progress: 1MHz 1V pk-pk.

This is with a bench supply but attached to the HP5245's crystal circuit.  Not all is
good, though:

* It takes about 16s to start up warm, about 2X cold.  First it is just DC then some
    impulse noise and finally a sine wave ramp up in voltage.

* When I re-install the card in the chassis, it stops working.  The supply from
   the '5245 is spot on at -15V, so it's something else.

This seems to confirm that Q1 was bad and that the 2N2907 is not a good
substitute.  I'm going to look for a higher-gain 30V PNP, otherwise I might
have to but a few real ones at $7 ea.

The only oscillator connections to the backplane are power/ground and osc
output, so either the output connection is shorted, or being placed in a shielded
environment keeps it from starting up.

[gmilliorn]

  Still could not get it working while installed in the system, so I guessed that maybe the alligator
clip I used to bridge the two independent grounds (J1-1 and J1-3 are separate grounds for the
oscillator and feedback portions) was adding inductance.

So I soldered a direct bridge between the two and sure enough, it no longer worked.

So, I removed the bridge and ... now it no longer works.  At all.   Argh.

I'm about to the point where I
A. remove and replace everything, or
B. remove and install a more modern oscillator.  Something like
    any of these:
http://cds.linear.com/docs/en/application-note/an12fa.pdf
   as the oven control circuitry is still working (measures correctly, at least)
that would seem to be the hard part.

[mij59]

  Still could not get it working while installed in the system, so I guessed that maybe the alligator
clip I used to bridge the two independent grounds (J1-1 and J1-3 are separate grounds for the
oscillator and feedback portions) was adding inductance.

So I soldered a direct bridge between the two and sure enough, it no longer worked.

So, I removed the bridge and ... now it no longer works.  At all.   Argh.

I'm about to the point where I
A. remove and replace everything, or
B. remove and install a more modern oscillator.  Something like
    any of these:
http://cds.linear.com/docs/en/application-note/an12fa.pdf
   as the oven control circuitry is still working (measures correctly, at least)
that would seem to be the hard part.

It would be wiser to replace Q1 and go from there.
If you want to replace the oscillator use a ocxo.

An other option
C. buy an other counter.

[gmilliorn]

Oh, I could buy another counter but this was supposed to be "fun".  I've worked
with nixie's before (clocks, etc). so I assumed it was cheap on ebay because of
problems in the display section.  I can supply a external 1MHz source at the back
and it works, so it is all in the oscillator section.

I even created an LTspice model of the circuit in an attempt to understand how it
works, so if nothing else, I learned that oscillators rely on thermal noise to start
up so spice simulations of oscillators are difficult-to-impossible.

I think replacing Q1 is indeed the next step, thanks.

[gmilliorn]

Well, I broke down and ordered the 1853-0009 transistors from sphere.ca.
I guess I'll see if there is anything magic about them.
In for a penny, in for a pound...

[gmilliorn]

After trying to drive the crystal oven externally, and failing, it seemed as though it
must be the crystal that had failed.  Not really a component I thought was going to
be the issue.  Replacement ovens are really expensive.

Fortunately, the OXCO is capacitively coupled to the drive card, so all I had
to do was inject a reference clock.  I had a Millirem OXCO, but 5 MHz, so
after powering and dividing it, it works great. 

It might even be better, though I have one nagging doubt:  does the divider
introduce any uncertainty?  It's all about the switching thresholds in the end.  Ah
well, either way it is better than what I had before (nothing).

[mij59]

Thanks for the update 

The level conversion from the 5 MHz sine wave  to square wave for the input of the divider is the only critical part, would you post the schematic ?

To get similar specs as the original ocxo you need something like the MTI 260 5MHz.

[gmilliorn]

You're right about the accuracy, I though the '5245 was 1 part in 10^8 but it's 1 part in 10^9, much better.

I can't seem to find my schematics, so I scribbled this in paint.net:

I used the schmitt to try to avoid having the sine wave slowly transition across the TTL switching threshold.

[flanneltuba]

For the sake of future travelers who find themselves looking to cross list the HP transistor 1853-0009, I happen to have an HP 5232A Counter which uses these parts on its A9 1MHz crystal oscillator board. On my particular A9 card rather than the parts being marked with the elusive 1853-0009 HP part number, these transistors are marked RCA 2N384.

The parts appear to be original from the factory, with no signs of having been replaced -- the solder is unblemished in any way that might indicate having ever been reworked.

I have also designed a modern replacement Nixie decoder+driver board for the aging custom HP neon optical decoder boards, found in this generation of HP counters. Let me know if you're interested, I can post my design here. I'll post a pic for show and tell anyway.

[factory]

HP also used the 4514 (1820-1683) decoder IC for the display boards in the early 80's. They used individual transistors instead of those ULN2003 arrays. Details are in the really crap quality manual scanned by Agilent some years ago.

David

[flanneltuba]

Indeed, I got a conceptual nudge from a photo of one of the later variations of the display card I spotted on ebay. The photo was too grainy to see the actual p/n of the 24 pin dip IC they'd employed, but from the look of the rest of the BCD counter logic on the card, it looked like the decoder and driver where clearly a retro fit onto the original counting circuit. I begged Google for a schematic or reference to the newer displays, but came up empty and set forth to just design something myself. So, ruling out the idea that they used any custom rolled ICs here, I concluded they must be using some kind of standard decoder/demux. From there it was a short leap to start playing with the 4514. So, perhaps to say I designed the circuit rings a little phony, but reverse-guessing the rest of the circuit without a schematic and only the grainy pic to go off of, at least qualifies me for honorable mention at the next gathering of the Technology Plagiarists Anonymous meeting. In any case, I now have a full set of clearly lit and responsive nixie displays, as opposed to the drunkenly foggy mess they were prior. Success.

- Scott

[Paul B]

hi I would love to get the CCT diagram for the replacement to the neon and photo resistiv display driver

ps there is a fix that i have used on a number of these displays when thy will not tiger / count properly

Paul

[factory]

Indeed, I got a conceptual nudge from a photo of one of the later variations of the display card I spotted on ebay. The photo was too grainy to see the actual p/n of the 24 pin dip IC they'd employed, but from the look of the rest of the BCD counter logic on the card, it looked like the decoder and driver where clearly a retro fit onto the original counting circuit. I begged Google for a schematic or reference to the newer displays, but came up empty and set forth to just design something myself. So, ruling out the idea that they used any custom rolled ICs here, I concluded they must be using some kind of standard decoder/demux. From there it was a short leap to start playing with the 4514. So, perhaps to say I designed the circuit rings a little phony, but reverse-guessing the rest of the circuit without a schematic and only the grainy pic to go off of, at least qualifies me for honorable mention at the next gathering of the Technology Plagiarists Anonymous meeting. In any case, I now have a full set of clearly lit and responsive nixie displays, as opposed to the drunkenly foggy mess they were prior. Success.

- Scott

Diagrams for the later display boards & mods for the 10544A type crystal oven, are in the manual on the Keysight webpage. Would be nice to find an original of the late manaul, as the scan quality is complete & utter shite. 
https://www.keysight.com/gb/en/assets/9018-02735/user-manuals/9018-02735.pdf

I took the other option with mine & bought some spare boards from Sphere to rob the decoder plates & neon assemblies to repair mine, had two with corroded traces on the decoder plate.
Note: with these that there are at least two different decoder plates, for the various BCD output code options.

P.S. It's a good idea to add a separate fuse for the small transformer for the crystal oven, that way if the oven heater shorts out, it hopefully won't burn the transformer up.

David