Power Supply

LED

LED

LED

LED

On the diagnostic board, the TTL signals on the backplane are buffered by 74LS24x buffer/line driver chips before being used to drive the LEDs and displays.

Due to the way that TTL devices are constructed, a TTL output can typically "sink" more current than it can "source", for example, the National Semiconductor 74LS240/241 datasheet shows that the maximum source current per output is 15mA, whilst the maximum sink current in 24mA. Therefore, when driving loads such as LEDs, the preferred option is to use current sinking.

In simple terms, in current sourcing, the current supplied to the load comes from the chip. In current sinking, the current supplied to the load comes from VCC and is merely switched by the logic chip.

There is a slight issue with this though, intuitively, we think of something as being "ON" when at Logic "1". For a non-inverting buffer, such as the 74LS244 in current sinking mode, turning on the LED requires that the input to the buffer is a Logic "0" to allow the current to flow. Using a 73LS240 (inverting) buffer means that a Logic "1" on the buffer input becomes a Logic "0" on the output, allowing current to flow and the LED to be illuminated.

The simplest method of displaying the status of the Z80 Address and Data bus signals would be to have a single LED per signal, in the same way as for the control signals. Although not really necessary, I decided that I wanted to decode the address line and data line signals and present the actual address and data values on 7-segment displays.

A single 7-segment display can display the data for up to 4 binary inputs using 16 hex characters, 0,1 2, 3, 4, 5, 6, 7, 8, 9, A, b, C, d, E, F.

The binary inputs must be decoded to determine which of the 7 segments should be illuminated to display the character corresponding to the inputs.

Image from: Frederick Blais blog

This was somewhat problematic, whilst Binary Coded Decimal (BCD) to 7-segment display drivers are common, Hex to 7-segment are not. Typically, a ROM or microprocessor based solution is used to translate the binary inputs into the appropriate code to be passed to the 7-segment display.

After much searching, I managed to source this chip, a PDN747, from the US. The cost of the chips, plus shipping & VAT, as well as the cost of the 7-segment displays makes it more expensive and requires a larger footprint on the board, but I think it's a nice feature.

There was a minor problem with the PDN747 though, the device needs a latch signal to clock the BCD inputs to update the displays. The dip switch pack provides a number of latch selection options:

The board features quite a large number of discrete LEDs and 7-segment displays, potentially resulting in a relatively high power demand, for example, 35 single LEDs and 48, 7-segment display elements (6 x 8, including DPs) , each drawing 20mA would give a power requirement of some 1.7A, not including the load from the ICs and pull up/down resistors.

The backplane design includes one pair of pins for 5v power distribution, the DIN connector pins are rated at 2A, allowing up to 4A of 5v power to be supplied to each board - adequate for the worst case load. However, the board interconnections are made using 30 AWG Kynar wire, the type that I am using was advertised as being rated for 0.5A but I believe that this may be an error, a more typical rating for Kynar wire is 0.4A, so consideration needs to be given to how power is distributed around the board. It would have been ideal to have larger capacity conductors on the board for the 5v and ground lines that could form a power bus to which individual terminations could be made - kind of like a ring main, but I didn't think that the size of the board and component density would make that very easy.

Instead, I chose to use a number of radial 5v and ground lines, tied back to single points at the connector pins and distribute the power in sections. Given that it would be possible to overload the conductors, I installed fuses in each of the power lines. Ideally, the fuses should have been sized to be less than the conductor current rating, but I chose to make them the same, i.e., 400mA.

Although not representative of the power consumption, the radial power lines are installed on a board layout basis

The 7-segment displays produce good brightness levels using a current of ~10mA, with all segments and the DPs illuminated, this would require about 480mA, but there will be some degree of diversity depending on the values being displayed at a given time. When displaying values of 0 to F_h the average number of segments illuminated is ~5, so, assuming that the numbers being displayed are random, the current would be ~300mA.

Due to space constraints, the board uses stacked LEDs that allow 4 LEDs to be installed in a very small space, there was a trade-off between the space saving and the current consumption of the LEDs. The current required to produce an acceptable level of brightness is higher than the current required with a "normal" LED, I found that 16mA produced an acceptable brightness for each of the colours being used. The power requirement of the LED modules will be 35 x 16mA, i.e., 560mA, but again, some degree of diversity will apply; the LEDs will only be on when the associated signal is active, looking at the signals involved, an on time of 50% would seem to be a reasonable assumption, leading an estimated current requirement of ~280mA. (When the bus is not being driven, the logic inputs will float and likely drift high. The LEDs turned on by a logic "1" will illuminate, but as this is only 50% of the available LEDs, this is not likely to be a problem either.)

If the current drawn by the indicators proves to be higher than the fuse rating, it would be relatively easy to rewire a couple of the 7-segment displays to take power from one of the lightly loaded lines feeding the ICs, but I will leave things as they are unless it proves to be a problem.

This is the "as-built" design of the diagnostic board, as a result of the 7-segment displays and associated logic, the board is now much more complex than I originally envisaged. (v1.03)

The schematic is drawn with KiCad, but the prototype (1-off ?) board is constructed on a eurocard sized prototyping board.

At this stage, the board was not going to be a PCB, it was constructed on prototyping board and the "tracks" are wire links.

Nevertheless, the KiCad layout and routing tool (PCBnew) was very useful in optimising the placement of the components. It was also useful to see the required wiring paths when moving components around to see if I could optimise the wiring too.  This is an early version of the layout before I settled on the final design shown below.

The layout for the diagnostic board, as you can see, the components all fit on the Eurocard sized board (160mm x 100mm), but there is not a lot of free space and wiring it up was quite a challenge!

The other challenge was how to make sure that all of the indicators and switches were accessible from the end of the board, particularly if the computer ever makes it into an enclosure.

7-segment displays have a much larger footprint than equivalent single LEDs, but I was able to minimise that by using 0.3" displays and making them visible from the end of the board by using right angled sockets.

The sockets are Series 800 Vertisockets, manufactured by Aeries Electronics, these are quite expensive new (~£10). I was able to source some "pulls" from Israel for ~£1.50 - fitted with red displays.

There are also around 36 single LEDs needed for the Z80 control bus signals, MTXPlus backplane signals and power indicators. To optimise the available space on the front of the board, I used 2 x 3mm quad-level LEDs

There are also a number of switches that need to be accessible from the front of the board.

They are mounted directly below the display sockets. This was possible since positioning the sockets to align the displays with the quad level LEDs left three unused columns of holes below the socket.

A PCBnew 3D view of the components placed in their locations during the design phase.

Although I don't have 3D models for the 7-segment display vertical sockets or the stacked LED modules that I used, the model did give a good indication of how the finished board would look.

This image is of the "as-built" board which includes the various minor modifications made during the build but does show that the finished article looks very much like it was supposed to.

This one also includes the decoupling capacitors installed inside the sockets of the ICs - the model is a bit misleading in that respect - it looks like they are sat on top of the ICs - obviously, this isn't the case.

The diagnostic board starting to take shape, the sockets closest to the DIN 41612 connector are for 74LS240/244 buffers and the adjacent sockets are for the PDN747, 7-segment display drivers.

You should be able to make out the decoupling capacitors installed inside the DIP sockets to save a little board space. The layout layout is a little different to the original layout drawing - I did some optimisation as I placed the components - the schematic drawing has been updated to "as-built".

A view of the solder side, with only the power and ground connections to the DIP sockets and decoupling capacitors made so far.

Some progress - the Address lines (A0-A15) wired up to the LS244 buffers and onward to the 7-segment display drivers.

The incremental difference in the photos does not do justice to the effort required to get to this stage! I am reasonably confident that the wiring to this point is OK - at least, checking with a meter, it agrees with the schematic.

Another update - the blue and green bundles are the output signals from the LS240/244 buffers that will control the red and green LEDs in the quad level LED modules.

At around this point, the hex-to-7-segment display drivers that I had bought finally arrived from the US and I was able to test one. Things did not go quite as well as hoped (you can find the details on the notes page) and I need to make a small modification to the wiring of the PDN747 sockets and add a little logic to control the update of the 7-segment displays.

In the following pictures you will also notice that the position of the single step circuit components and the orientation of the LED current limiting resistors is different to the schematic - I made a number of changes when I added the latch logic for the PDN747s.

The rest of the sockets mounted; the top 3 are for the single step instruction circuit and the one below them is a 74LS04 for the 7-segment latch logic.

The  5 x 10K resistors are pull-ups for the single step circuit and the 48 x 330R resistors are for the 7-segment displays (including the decimal points). That "only" leaves another 36 resistors, the display elements and switches to be added!

The left hand side of the board is now almost fully wired.

There is an interesting dichotomy when doing the wiring, making the wires as short as possible makes the board look neater, but would be problematic should any rework be necessary. I tried to leave a little slack on most of the connections, but the need for rework should be minimised as I am testing the connections at convenient stages.

Getting there!

- The obvious difference from the last photo are the additional column of resistors for the 4 level LED modules, the 4 x 10K pull-downs and diodes for the fuse monitor circuit as well as the capacitors for the single instruction step circuit.

Just for completeness . . . . . . the solder side of the board.

Still looking a bit of a rat's nest, but there is some structure to it - honest! All that remains is to mount the displays and switches, then connect them up to the unterminated wires and resistors.

The last of the components mounted, just a hundred or so connections left to solder now!

The black modules at the left hand side are the quad level LEDs and the odd looking right angled sockets are for the 7-segment displays. You can also see the toggle switches mounted on the underside of the board.

To allow the switches to be mounted in the holes below the display sockets, the switch contact legs and securing posts were trimmed so that they were flush with the top side of the board. This will result in the loss of some mechanical strength, but I don't think that is significant here.

The completed board with all components, apart from the single instruction step ICs, installed. The board is being powered by a temporary supply which provides only 5VDC.

The address and data bytes are all reading "FF" as the bus signals are floating and currently "high", i.e., each bit is currently interpreted as a "1".

The board has been tested and appears to work as required (after fixing a problem that I had created when choosing the value used for the chip enable pin pull-down resistor - more details on the notes page).

The red LEDs are all off (the active low signals have floated high) and the green LEDs (apart from the two spares tied to 0v) are all on (high).

The "final" state of the solder side of the board, still not looking too neat!

There are a number of spare wires that are not fully terminated and the slack left in other wires means that it does not look very tidy, but it's the best that I could do. I will probably spend a little more time tying up the wires to make them a bit more secure though.

"Crib Card" for my benefit . . . . .

Showing the position and function of the various LEDs and switches, it's a poor quality drawing, being done in KiCad during the schematic design phase, but it does the job.

I'll probably do a decent drawing at some point.

With the single step circuit components fitted, and after correcting a silly wiring error, the single instruction step circuit appears to be working.

This photo was taken with an MTX512 switched to single step mode, at this point, the WAIT, MREQ, RD and INT signals are all low.

Having looked at the MTX ROM disassembly, Martin was able to advise that the CPU was stopped in the middle of the keyboard scanning routine, 36DF happens to be IN 6 and it’s pointing at the 6.

A slightly better photo, with similar results

Again, Martin has advised that the display corresponds to the start of the keyboard reading routine, 3622 3E is the first byte of LD A,251

(I really do need to learn to navigate my way around the ROM listing)

Martin built one to help him diagnose a problem with an ailing MTX and I thought that it was a good idea to build one too. The board only patches the signals from the MTX cartridge port to the MTXPlus backplane layout. Using the board on a good MTX will help me get an idea of what the bus signals should look like for an operating Z80 before I get to testing the MTXPlus CPU board.

Slightly less impressive than the diagnostic board - the MTX adapter.

The two jumpers allow me to isolate the +5V and +12V lines on the MTX bus from the diagnostic board.

The wiring side of the board - as you can see, the board just cross wires the terminals from the MTX cartridge port to the corresponding pins on the MTXPlus bus.

A rather blurry photo of the board running on my original MTX512.

I hadn't fitted the single instruction step components at this point (they were on order), so the 7-segment displays are showing flickering "8"s as all of the segments are lit displaying values at high speed. Similarly, the LEDs are also flickering, but the indications are consistent with what I expected, but are more easily seen if I halt the CPU.