Strange product design decision... Or just screw-up?

[Leo Bodnar]

I have been doing some TLC to a beautiful Anritsu network/spectrum analyser MS420A. Replaced tarnished BNC connectors, cleaned up, aligned CRT, etc.  It's a great instrument: hand-assembled Japanese quality from mid-80s, 74LS series logic throughout, no ASICs, thick service manual with schematics diagram, repair, alignment and take-apart instructions, etc.



However I have service manual for the second revision (MS420B) which seemed exactly the same apart from the fact that I could not find the master reference 10MHz clock module. 
On MS420B it is in a big massive box in the middle of the rear panel to the left of the fan exhausts:



Then I have taken out CRT driver board for cleaning and - boom! Here it is, just between two CRT coil driver heatsinks - complete with botched re-wiring job. 





Who in their right mind would place master reference clock of the instrument that supposed to have 5x10^-8 stability on a hot noisy CRT driver board?!  Not that there is no other space in there - the thing weighs 35kg but there is lots of empty space still.  It's otherwise beautifully engineered - RF sections are sitting inside milled out cast aluminium slabs, all analogue modules are inside formed aluminium boxes, etc

I reckon the designers forgot about the master reference clock module and when the panic hit the only available space was on the lowly CRT board so they have hacked in some space there and just crammed the poor fella in.  They were in such a rush that still screwed it up and needed to do manual re-wiring on production PCBs. 

I had a chuckle! 

Cheers
Leo

[Kiriakos-GR]

I am aware of the Japanese Anritsu  brand,  never owned or use any of their gear.
It is terribly expensive in Europe.

Even so , the golden rule still is, that if it works , do not touch it.

I have in my hands one Sony CRT monitor that worth ed 1000 Euro , before 10 years. 
Now its dead , all the parts inside looks like new , the PCB soldering looks to includes silver.
I can not find what go wrong with it ..
And even so, I know that even if I would had any success about repairing it ,
the tube has lost their freshness , and it is just a bit blurry.

It looks great inside , PCB and parts , nice as spectacle to be presented in a technological museum. 

But I would prefer to had in it a botched re-wiring job, and to was alive too ..   

[tekfan]

Very strange place to put a reference oscillator. Probably the worst too. All the heat those power reisistors, transistors and a lot of magnetic flux from the flyback transformer won't help the crystal stability. Well atleast they corrected the problem in the B version.

It's otherwise beautifully engineered - RF sections are sitting inside milled out cast aluminium slabs, all analogue modules are inside formed aluminium boxes, etc

Are there any more pictures of the machine? I'm sure a lot of people would like to see the insides of such an advanced piece of test equipment.

[Leo Bodnar]

Are there any more pictures of the machine? I'm sure a lot of people would like to see the insides of such an advanced piece of test equipment.

I will be taking it apart again to replace some more connectors and to create some custom hardware. More pictures then! I want to replace some boards with modern hardware.  This is 80's kid heaven!
The whole thing is based around 5-6 8 bit micros.  I am also thinking about replacing CRT circuit and tube with TFT VGA screen because the machine has VIDEO output that mirrors the screen contents.

One of the CPU boards.  There are 5 in total.  CPUs are R6502 (main) and INS8070 that is used for multiplication/division only.  The main CPU places two 16 bit integers into a specific shared RAM location, triggers I/O port, INS8070 takes over, multiplies (or divides) them and triggers another I/O port line.  37usec per multiplication and 42usec per division.  PIC24F can do it in less then 1usec! Main CPU then reads the result from shared RAM:


Frequency synthesizer for 56...86MHz range with 0.01Hz resolution built entirely from 74LS logic.  Wow.



One of the boxes open for BNC replacement.  35 years of heavy use...



New silver plated BNCs fitted...

[Mechatrommer]

no doubt that is a marvel, even the osc at the wrong place.

[tekfan]

I want to replace some boards with modern hardware.  This is 80's kid heaven!
The whole thing is based around 5-6 8 bit micros.  I am also thinking about replacing CRT circuit and tube with TFT VGA screen because the machine has VIDEO output that mirrors the screen contents.

Be careful though. you wouldn't want to end up with just a single board doing everything. LOL

Are the empty RAM sockets for memory expantion?

[Leo Bodnar]

No, I just want to make it a little bit faster on UI side and replace a CRT.

Are the empty RAM sockets for memory expantion?

They are for extra RAM that can be used for BASIC programming.  This is pretty much stand-alone computer with full ASCII keyboard and graphics display.

[saturation]

Nice job, thanks for sharing.  is it working to specification?  

[qno]

Maybe it is the best place.
Reference oscillators run at 60 to 80°C.

You do not want that in your highly sensitive (maybe temperature sensitive) equipment.
Or maybe it did not meet the required specification when the oscillator oven was inside.

Maybe they needed the distortion for the oscillator to double as a marker.

As far as I know Anritsu they know what they are doing.

[Alex]

I am aware of low noise designs next to switch mode power supplies on the same board with -120 dB crosstalk.
10Mhz is a relatively low frequency for the trace lengths shown and the oscillator is partly electrostatically shielded. There might be some circuitry around it to restore signal integrity.

Having said that, it iclearly is an afterthought.

[seattle]

This is not surprising.

Even in the early 90's cellphones were using temperature compensated reference oscillators. A tiny processor measured the temperature inside the oscillator package, and applied a voltage to a varactor attached to the leg of the crystal and with that small variable capacitance could pull the oscillator to well within spec from -30'C to +85'C (which is where phones were tested back then...not anymore). The processor would even account for aging--crystals tend to change frequency over time in a predictable direction. And of course, AFC would always pull you in to perfect.

If the TCXO was spec'd to run over a wide temp range, the sticking it next to something hot is fine.

[Leo Bodnar]

Maybe it is the best place.
Reference oscillators run at 60 to 80°C.

You do not want that in your highly sensitive (maybe temperature sensitive) equipment.
Or maybe it did not meet the required specification when the oscillator oven was inside.

Maybe they needed the distortion for the oscillator to double as a marker.

As far as I know Anritsu they know what they are doing.

LOL, I hope you are being sarcastic!  

Surely the perfect location for the OCXO is supplemented by unsurpassed quality power and output pigtails soldered directly to scratched out copper pads on the CRT driver board and sporting horrendously looking  MOLEX connector for what supposed to be a clean 10MHz clock. All other RF connections are made with coax SMA connectors.

Also to adjust the reference clock (which is the most basic instrument calibration) you have to take the CRT board out, hang it loose on all the wires and navigate your way with a torch light and screwdriver hoping to miss 16kV thingy.  The CRT driver board is mounted with components looking inside the instrument.

Revised design in subsequent model:


First try...

[tekfan]

Also to adjust the reference clock (which is the most basic instrument calibration) you have to take the CRT board out, hang it loose on all the wires and navigate your way with a torch light and screwdriver hoping to miss 16kV thingy.  The CRT driver board mounted with components looking inside the instrument.

LOL
Well, atleast the pcb is on the side of the machine and not burried somewhere inside. But when I think of it, that qould probably be a far better position for the TCXO.

Is it a TCXO or OCXO?
Probably too small to be an OCXO.

[Leo Bodnar]

Is it a TCXO or OCXO? Probably too small to be an OCXO.

I believe it is an OCXO as it was made by NDK who continue making them in similar packages.
Some of them are as small as 21 x 13mm http://www.ndk.com/en/products/search/ocxo/1188023_1509.html

[Leo Bodnar]

Nice job, thanks for sharing.  is it working to specification?  

Yes, as far as I can tell. I am still halfway through its functions.  One thing I have noticed is very hot output video signal. It is supposed to be standard video level and even the output connector on the panel is marked as VIDEO / 1Vpp.  However it shoots to full 5V at white level and according to schematics it is by design.  Another cock-up. Luckily it is a simple resistor DAC with two levels of white so I am going to replace them to tune it down to standard 1Vpp swing.