Behold the Tektronix 7D01, first released in 1977, shown here with it’s companion display formatter. As a side note, I really need a better lens. This pin-cushioning is just too much. 105mm or 200mm prime macro (or micro as Nikon calls them) is high on the list.
This wasn’t Tektronix’s first bench-top logic device, but it was the first to have a tiny bit of smarts to it.
- In 1975 they released the 821 4 bit word recognizer released in 1975. Yep, a 4 whole bits.
- In 1976 they released the LA501 as a part of the TM500 series instruments. It could store up to 16 channels x 256 bits, and relied on an external display. No word recognizer, no cursor, no read-out, just a simple timing diagram.
Hewlett Packard had a jump on the bench-top logic analysis tools, starting in 1973 with the 5000A logic state analyzer. It’s no WOPR, but a respectable display of blinking lights nonetheless.
By 1975, they had released several bench-top products. The HP Memory project has a great page on the topic.
Back to the 7D01. I’d picked it up a few months ago, and separately picked up one of the two 8 channel probes. The probe connections are to the left of the word recognizer. Note most logic analyzers are useless without their probes, and the probes are usually specific to one product or product family. You can pick up this and other 80’s era logic gear for under $100, but be prepared to spend a comparable amount on a set of probes.
So I started mucking about with it, generating test sequences from an Arduino and learning out how to operate the thing, and saw a variety of intermittent failures:
Here’s a quick video of one failure mode, this time with the display formatter removed. Notice the characters look different, more on this later.
The thing is, I am now in possession of no less then three of these units, as I was given two ‘for parts’ when I picked up the 7603 & 7D20 a few weeks back. And of course, these units both worked flawlessly. I contemplated just calling my original unit the ‘parts’ spare, but that seemed like a cop-out, so I went forward attempting to repair my original, knowing I had an ample supply of parts, should they be needed. Besides, that’s the real fun anyway.
I studied the manual during my daily subway commute the week prior to attempting to diagnose and repair. Putting the display formatter issues aside, I suspected the counter on the cursor board was responsible for the racing numbers in the video, and that something was wonky with the flag signal which could explain the display blanking.
When confronted with a multitude of seemingly unpredictable problems such as these, the best approach is often ‘take it apart and put it back together’, making sure to take pictures along the way to help in the reassembling effort. This isn’t just the physical equivalent of ‘have you tried turning it on and off again?’; Close visual inspection can reveal failed components, and re-seating 35+ year old connections can often fix intermittent problems. Besides, without plug-in extenders which would let me probe around on the live-plugin running outside the mainframe, there really wasn’t much else I could do.
As you can see, it wasn’t a total teardown. I left the output board and word recognizer in place, as I could now reach and re-seat all of their connections and chips.
I didn’t individually pull and re-seat every chip, but I did give them all a firm press, and found that many had worked themselves loose. All that SRAM and discrete logic really cooks, so it’s not surprising to see failure modes change as the unit warms up and pins move slightly as they heat up and expand.
It took another hour or two to get the whole thing back together. While I took a bunch of photos (more then shown here), I still had a few stray wires that I needed to hunt down in the manual to figure out where they reconnected. It helps that the wire harnesses are pre-bent and just the right length, so often times it’s obvious where to connect; however that also means there’s little to no slack, so there’s really only one way to route each wire. I had a few runs that I had to re-snake through after I realized I’d taken the wrong path.
After one more visual check, I slid the plugin back into the 7603 mainframe, held my breath, and pulled the power button (it’s a strange power button).
Here it is a little while later, after I reattached the Display Formatter (DF01).
Notice the difference in how the characters are drawn. The details of both are worthy of their own posts, but short answer:
The mainframe’s built-in character generator uses a column & row format to address each character, but the characters themselves are drawn as a series of vectors.
The character generator in the DF01 (and DF02, and 7D20) addresses each whole row of characters as a scanned, raster canvas. Interestingly, when a ‘pixel’ in a character is to be drawn, not only does it unblank the beam, but it also makes the horizontal scan pause for a moment to attain sufficient brightness on that spot.
In the above image, the top and bottom rows are handled as a raster canvas by the DF01, while the timing diagram is swept in the traditional method; one horizontal sweep for each channel. The DF01 & 7D01 are each responsible for their own portion of the canvas. In either case, the 7D01 handles all of the X, Y & Z signals to the mainframe; there is no direct connection from the DF01.
Note that in this mode, the everything on the display is being generated as a raster from the DF01.