After all the help I received on my SG503 analogue repair recently, I thought I would write up an account of some digital faultfinding and repair on my second unit, in case it might be useful to someone.
For this unit, the levelled oscillator output was OK across the entire frequency range. The fault was with the counter, which was fine up to 0.999 MHz, but then went flickery at all higher frequencies (a high speed flicker on all the digits, making them unreadable, not the slower blinking to indicate unlevelled output).
I suspected a fault in the divide-by-eight prescaler, which is supposed to become enabled only at frequencies above 0.999 MHz, in other words on the first ‘overflow’ of the counter which should also then drive the selection of the next faster master clock which controls counter gating, display blanking and much else. But probing around with an oscilloscope was very confusing – the circuit was not in a stable state and I was seeing a lot of unexplained activity on the SET pins of the prescaler D-types, so it was difficult to tell what, if anything, was preventing the prescaler from working. I suspected the thing was in some sort of endless cycle of overflowing, then (correctly) notching the range up, then for some reason underflowing and notching down again, etc.
This was the first of two general challenges I found when faultfinding the counter – because the whole thing essentially consists of a big control loop whereby changes to the input frequency drive the auto ranging up and down. Once it gets into this sort of fault condition cycle, because each individual control signal looks correct in isolation it’s very difficult to trace a root cause without ‘breaking the loop’. And it’s not immediately obvious where to do that while still being able to home in on the fault. So I wasted a lot of time tracing through the complicated auto ranging logic only to find it all seemed to be performing perfectly.
The other challenge is that all the control signals (counter gate, display enable, clear counters, autorange state machine etc) change at least 1000 times more slowly than the oscillator signal being measured, so it’s difficult to see everything together on an oscilloscope. And the counter clear pulse is a mere 100ns wide.
Anyway, the mighty DSOX3014A came to the rescue again, and in the process I learned a lot more about how to use it. I set things up to use the digital (MSO) inputs for the TTL logic, including a full decode of the most significant counter digit, and analogue inputs for the faster signals around the ECL prescaler. Using ’single shot’ and pattern triggering looking for counter gate enable high and a 9 -> 0 on the most significant digit of the counter (ie the overflow condition), I managed to capture the overflow point when I turned the variable frequency from 0.999 MHz to 1.000 MHz. I could immediately see that the signal input to the counter gate then stuck high at the point that the prescaler became enabled – in other words, the prescaler wasn’t working, as suspected. With the prescaler effectively dividing by infinity instead of eight, the autoranging logic then detected a much lower frequency (DC!) at the counter input and perfectly correctly stepped back down again. Thereby disabling the prescaler, re-establishing the 1.000 Mhz input to the gate, overflowing, etc.
From there it was fairly simple to step back through the three ECL D-types of the prescaler, to find that although the input signal and its DC bias to the first stage (U390) looked good, clocking Q-bar back around to the D input wasn’t resulting in a change of state as it should. The final piece of luck was that U390, like the transistors, has those tiny individual spring sockets soldered into the PCB holes (did Tektronix know it might be a failure point?). So the very easy swapping of U390 with my other unit made the counter spring back to life perfectly from 50Hz to 250MHz, thus confirming that the U390 was dead. So another is now on order – fortunately I managed to find the fairly obscure and very obsolete MC1670L in stock in the UK.
In hindsight of course I should have gone straight to the prescaler, but I hope this account is useful for the record anyway.
Home
Help
Search
Login
Register
