Group Actions
Engagement
  • Author Author: michaelkellett
  • Date Created: Date Created
  • Views 793 views
  • Likes 12 likes
  • Comments 8 comments
Related
Recommended

Two rather different communications testers.

michaelkellett

I spotted 2 Hewlett Packard 3784A Digital Transmission Analyzers on Ebay recently. I thought they might be interesting for the boxes but when I looked more closely I realised that they were a bit more interesting because they should be able to talk to an instrument I designed in 1986/6 for BT - the Tester 167C.

I acquired the 167C a couple of years ago and it sits on a shelf to remind me of when I used to design this kind of thing. BT (in those days) were still specifying their own custom built test equipment and designed several instruments for them. The 167C was one of the last we did and it was powered by a Motorola 6809 processor with all the code written in assembler. It had a CRT display and operated as low speed (up to 2Mbit/s) telecoms signal generator with a very unusual analogue narrow band measuring system which was used for testing individual signal booster/repeaters arranged in a string. We probably made between 150 and 250 of them and I think they were sold for about £350. It might have been a little more but certainly no more than £500.

The HP 3784A is a more general purpose instrument, and capable of working at higher frequencies (up to 34Mbit/s) and it was marketed world wide, although designed in Scotland I think. It was based on an Intel 8088 processor and offered a very expensive option for jitter measurement.

The 167C was released in 1987 and the 3784 in 1989. In 1990 the 3784A appears in the HP catalogue at $10200 (+ $5800 for the jitter option).

I couldn't turn down the chance to get my 167C something to talk to and I was able to buy both 3784A instruments at a very reasonable price.

Bearing in mind that all three instruments are now about 35 years old it's impressive that they all work - and the 3784A can lock on to message from the 167C. One of the 167C features is that it can inject code or bit errors into a data pattern at any rate from 1 in 10E3 to 1 in 10E9.

In the test shown in the pictures the 167C is generating a pattern of 1010 encoded in HDB3 at 2048 bit/s and injecting bit errors at the rate of 1 in 10E6.

Both instruments are based on a micro backed up with large numbers of 74HC logic chips, some static RAM and some nice analogue stuff. (The 167C has the analogue slective meter and the 3784A the jitter measurement. (Sadly my 3784As don't have jitter measurement installed.)

Although the 3784A has a nicer box than the 167C I like to think that we offered a much nicer user interface - and  a much more competitive price. But in the long run HP got all the glory Relaxed

image

image

image

MK

  • Nice find.

    It will be interesting to see if they still work and what you do with them.

  • in reply to michaelkellett

    Impressive! yes - I was thinking about HP at 34Mb/s (it seems that tolerance is getting narrower at every mux level). There was also different number of stuff bits in the frame (to accommodate for tributary frequency variations) at different levels if I remember correctly...  

  • in reply to JWx

    The 167C only operates at 2.048 Mbit/s which is defined in (amongst other docs) G.732:

    image

    So my recollection that we did nothing that special with our clock is probably correct, and the signals are within spec after 38 years !

    G.732 is delightfully vague for an official international spec, I liked this:

    image

    It features a double fudge, "should be" rather than the modern "shall be" and  "a few seconds". Those were the days !

    MK

  • in reply to michaelkellett

    Very nice! It always amazes me how very advanced equipment can be bought for relatively low money after being phased-out...

    BTW - I was being under impression that PDH multiplexers were using some expensive clocking sources but it seems it is more like 20ppm for E3 (for some arbitrary mux I have found specs of).

    Must check where rubidium clocks were used - maybe in switches synchronized from PRC in pre-GPS era?   

  • in reply to JWx

    The full service manual for the 3784A is available on line, complete with schematics and parts lists.

    In the operating manual they spec the clock like this:

    image

    The 167C schematics and spec are no longer available Frowning2

    But I have a screwdriver !

    image

    167C with the covers off showing the top board which is mainly power supply stuff. The transformer on the PCB is a custom made dual bobbin 1:1 transformer with very low inter-winding capacitance. Its driven by a low frequency oscillator to generate a low noise isolated power supply for the analogue level measuring function. Yow can just see the selective filter with its huge ferrite inductor in a shiny metal hum shield at the bottom of the picture.

    image

    167C processor board - the 8MHz clock only works the the processor (at 1MHz) but memory access was shared with the display so there were two memory accesses per 1us clock cycle.

    image

    Digital signal generating board with its own 12.288 MHz clock. The two 8kx8 static RAMS are used to store test patterns which are calculated by the micro.

    image

    The Racal frequency meter is using an external GPS disciplined clock (good for about 1ppb). This is about a minute after switching on the 167C. The pattern is 1010 so after HDB3 encoding you get 1/4 of the clock frequency. The 167C is -19ppm.

    image

    After about 15 minutes warming (and it doesn't have the cover on !) and the 167C is -22ppm.

    image

    The 3748A does a bit better on frequency accuracy, -1.25ppm

    So the conclusion here is that HP used  a (much) better clock oscillator than we did. I don't remember the 167C spec for sure but I think it was 20 or 25ppm. The 167C was not intended by BT to be used as universal test signal generator but as a portable trouble shooting tool - so precision frequency testing was outside its scope.

    MK