TRS-80 Model 4 Repair

The TRS-80 Model 4 was placed out on display.  But apparently, when it was powered on at some point, it put on a smoke show.  So we were told that the computer was working, except that it needs the failing RIFA caps removed.  Sounds simple enough, so we took that on to get the machine back into working order.  How bad could it be, right??  So after removing the remnants of the RIFA caps, we powered the unit on, just to see the situation for ourselves.  It did not throw any more smoke at all.  And it did appear to be trying to display some sensible text on the screen.  But this text was illegible because of a severe smearing effect, blurring toward the right.

Because the screen is pretty much unreadable in this condition, we decided to tackle this blurry situation first.  We found a service manual on-line for the TRS-80 Model 4 (Model 4 Technical Reference Manual).  The version we landed on largely matched our machine, but it was not an exact match.  Good enough to get started.

The video signal originates from the computer side and is sent over to the video display section, of course.  We began by finding this signal and looking at its quality.  It looked good.  The digital waveform is responsible for switching the electron beam on and off as the raster proceeds across the CRT.  Indeed the signal was sharp.  It had a reasonable peak-to-peak voltage and defined edge transitions.  So it look like this issue is on the video display side.  And zooming out, we could clearly see the lines of text as the downward squiggles.

We quickly played with the brightness and focus adjustments -- both user accessible and internal to the monitor.  None produced any improvement.

So we next followed the video signal path through the display's circuitry and inspected it along the way.

For a bit of background, the way the CRT display works is that the electron gun (hot filament) sheds electrons at the rear of the tube.  The electrons then accelerate toward the front of the tube because of the high voltage applied by the big scary anode cap.  As the electrons impact the phosphor screen at the front, the screen lights up.  The image is created by the control of this electron beam.  The coils installed on the neck of the tube are called the deflection yoke.  These coils create magnetic fields which, as the name implies, deflect the electrons as they travel from the rear to the front.  One deflects the beam from the top of the screen to the bottom, then back to the top, again and again.  The other faster one runs the beam from the left to the right side of the screen.  Together, these deflections produced the raster pattern of the beam to fill in the entire screen.  The beam goes from left to right again and again for each pixel line across the screen as it goes from top to bottom for each frame, resembling reading line-by-line over the page of a book.  The display is monochrome, so the complication of color is avoided.  The image is created by varying the intensity of the beam as it rasters.  Where the image is bright, the beam is strong when it is passing over those areas of the picture.  Where the image is dark, the beam is correspondingly weak.  The video signal from the computer that defines the image is locked to the timing of the beam raster left to right and top to bottom.  So the computer is simply sending a digital signal to turn each pixel on or off in time with the beam.

The video signal arrives from the computer to the video board at pin 8 of connector J101.  It then proceeds to the base of transistor, Q302.  The purpose of this transistor is to amplify the signal to the higher voltage used by the CRT tube.  Inside the tube is a grid plate.  This is a metal mesh that sits in the path of the electron beam inside the CRT.  By default, this grid is at high voltage.  That prevents the beam from passing through and hitting the front of the screen, so the screen is dark.  But when the Q302 transistor turns on from the video input signal, the grid voltage goes low.  This allows the beam through, and the beam can then hit the screen and light up that location as the beam rasters.

So which of these parts is the problem?  The circuit has resistors and chokes.  Those aren't too likely to fail.  So, the knee-jerk reaction was to immediately suspect the transistor.  Transistors fail all the time, especially ones that drive somewhat high power and are also several decades old.  But there is a hint in the display output.  The screen, as it rasters from left to right, does turn on correctly.  The dark-to-light transistions are sharp.  It is only the light-to-dark transistions that struggle and show the blurry intensity decay.  Given how the circuit works, that suggests that the issue is the transition going from a low-voltage to a high-voltage state.  The hard part of the transistor's job is really the other transition, from the high to the low voltage state.  So really, what parts are responsible for recovering to its high-voltage state?  That's things like R303 or L301.  Failures of resistors and inductors are more rare, but not unheard of.

Meanwhile, Will, my partner in all of this, had done some research and offered some interesting information from the internet.  Apparently, this effect that we are seeing is a common problem -- to the point that it has a name -- "comet trails."  People on-line say the choke L301 fails in the circuit.  That makes sense with our understanding of the problem, because that would indeed make the signal unable to pull up to the high voltage and properly shut the beam off.  If the choke L301 were broken (high resistance or open circuit somehow), then in fact we would anticipate an effect like these "comet trails."

So Will removed the choke L301.  Unfortunately, in doing so, it was damaged, and we could not corroborate that it did have a component failure.

But the bigger problem is how to replace it.  The RICM museum/lab space does have a lot of various components.  But sadly, there was no convenient selection of replacement inductors.  We saw from the schematic that this has a 22 uH inductance.  So that's roughly what we are looking for.

What we do have, though, is an old TV.  Looking like that TV will need to make a sacrifice for the greater good.  We looked around its main board and saw a dipped inductor.  Its inductance value was indicated by the cryptic pattern of colored dots on its body.  And this was orange white black, corresponding to 39 uH.  Though we were looking for a 22 uH, this would be suitable enough for the job of just limiting current during the transitions of the video signal.  So Will did an excellent job soldering that in as a replacement.

And with that, we turned it on.  And it worked!  We now have a nice image.  Of course, the brightness needed adjusting to turn down the gray background / raster line.  And the image itself was cockeyed, requiring that we rotate the deflection yoke on the tube to correct its angle.  And you'd think that would be the end of the repair journey, right?  Right??

So we were off to celebrate (or actually probably to clean up).  But after letting it run for a bit, we came back to notice that the screen was blank.  Reboots, power cycles did nothing.  We also noticed that, unlike before, now the disk drive light did not come on during boot.  What now?

First thing, we looked at the video signal, which we were already very familiar with.  Indeed, we saw no activity there.  So it appeared that this was not a problem with the video monitor or our recent fix.  It was displaying nothing because there genuinely was nothing to display.

There was some sort of computer problem now.  So on to diagnosing the digital.  We look at pin 6 of the Z80 processor and indeed did see a healthy clock signal.

The chip has power and is commanded out of reset too.  But we look at the address lines to see what the Z80 is thinking.  And each one is a flatline -- always low.  The Z80 isn't accessing memory like it should.  Pretty suspicious.  So we replace it.  And indeed, with the new Z80 we now see normal activity.

Although there was Z80 activity now, we still aren't back to a working state.  We see activity now, but is it actually trying to do anything sensible?  Well, some time in the start up process, it should be initializing the MC6845 video controller.  So we decided to look at its chip select line.  And indeed did see a series of accesses, which were presumably writes to set up its configuration.  This bodes well for the processor doing the right thing.

We inspected various other things as well, like the /DIRCNT, the RAM data pins, other chip select lines and other internal signals.  No huge red flags.  At this point, we definitely see that the computer it trying to do something.  It just isn't getting very far.

There are often problems with RAM in this era of machine.  On an educated but blind guess, we snooped at the data line for each 4164 RAM chip.  Although the data is a chaotic scramble, each looked like it had valid signaling.  We also looked at the address lines and control signals of the RAM bank as well.  These also looked alright.  This unit also had an second bank of 4164 RAM chips, upgrading it from 64k to 128k.  All RAM was socketed, so next we also just plucked all the RAM chips out and tested them individually.  That was fruitful, in that we did find a bad RAM chip.  But the bad one was part of the upgrade bank, so probably was not contributing to the failure to boot.  We changed it out anyway, presumably avoiding some future head scratcher.

There is a pair of disk drives on the machine.  They had seemed to work initially when we started this repair.  But now that we had basic functionality confirmed by the test software, we started eliminating other possibly faulty hardware.  So the ribbon cables to the disk drive controller card ware removed.  We also removed the test ROM and restored the original boot ROM chip into U68.

And with that, the machine booted properly.  This was rewarding because the overall outline of this text looks like the initial blurry text that we had first been seeing.

Now a moral quandary arose.  Do we keep going at this and repair this apparently failed disk controller?  Maybe another day.  But we decided that, since this machine was never used with disks anyway, it was good enough to button back up for BASIC, as it always is while on display.  The ribbon cable would just stay out, and the drive power was removed too.

Our final detail was to include some notes for the next poor schlep (who might possibly be us).

And that was it -- getting the TRS-80 Model 4 awake again for display duty.