The Unidisk 3.5 is an 800K floppy drive for Apple II computers, using the Smartport protocol to communicate with the Apple II. The BMOW Floppy Emu disk emulator is also capable of emulating a Smartport disk, so in theory it should be possible to plug the Emu into the Unidisk’s daisy chain port, and use them both together. Unfortunately it doesn’t work, for reasons that weren’t clear until recently. The good news is I now understand what’s going on, but the bad news is it’s probably impossible to make it work without hardware modifications.
Smartport is a bus-based protocol. Each disk is assigned a unique address at startup, and it should only respond to commands for that address. The original Floppy Emu firmware for Smartport was intelligent enough to do that, but it contained some implicit assumptions that were wrong in a daisy chain situation with multiple Smartport drives present. Fixing those was the first task. For example, it would ACK the receipt of any Smartport command, even if it didn’t actually respond to it because it was for the Unidisk. It would also enable its output on the READ and HANDSHAKE lines whenever any Smartport drive was enabled, interfering with the Unidisk.
Address Assignment
After resolving those problems, daisy chaining still didn’t work. The logic analyzer showed that the Floppy Emu was never even assigned a Smartport address. Here’s the telltale trace:
The first set of squiggles there on WRDATA (channel 08) is the computer assigning address 1 to the Unidisk with an init command. The following squiggles on RDDATA (channel 07) are the Unidisk’s reply, which is “OK, and there are no more Smartport devices after me”. The next command on WRDATA is a request to read sector 0 from the Unidisk, so Floppy Emu was completely ignored.
Determining what those squiggles mean is a tedious process. I have to zoom in until I can see each positive and negative transition of WRDATA. Every 4 microseconds there will either be a transition (a logical 1 bit) or there won’t be any transition (a logical 0 bit). I have to write down the bit sequences, frame them properly into bytes, and then consult the Smartport spec to make sense of it all. Maybe someday I’ll write an automated tool to do all this, which would make the debugging process dramatically faster. For now I’m happy simply to graph all the signals, because there was a time when I didn’t have even that much.
So why doesn’t the Floppy Emu get assigned a Smartport address? If I were designing the Smartport protocol, I would probably have it send as many init commands as necessary to give addresses to all the drives. Just keep sending init commands, incrementing the address each time, until all drives have received an address and no more init responses are received. But Apple chose a different solution, where each Smartport device is expected to know definitively whether or not there are more Smartport devices behind it in the daisy chain.
Input Becomes Output
Apple used a sneaky trick to accomplish this. On the DB-19 connector, pin 16 is normally an input to the disk called HDSEL, which is used to control non-Unidisk 3.5 inch drives. But on the Unidisk 3.5 (and presumably other Smartport devices) pin 16 of the male connector is tied internally to ground. On the Unidisk’s daisy chain output connector, pin 16 has a 2Kohm pull-up resistor to 5V. Internal logic senses whether pin 16 on the daisy chain connector is low (another Unidisk or Smartport device is daisy chained, and its internal ground connection pulled pin 16 low) or high (no Smartport device).
Turning a disk input into a direct ground connection is dangerous. It means that if the computer tries to drive a high value on HDSEL, and a Unidisk 3.5 is connected, it will create a power to ground short and likely fry the disk controller. This will happen for certain if a Unidisk 3.5 is connected to a Macintosh. The Apple IIc and the Liron disk controller don’t connect anything to HDSEL, so they’re safe. The Apple IIgs does make use of HDSEL, but its schematics reveal a 470 ohm inline resistor to protect against a power to ground short. I’m not sure about other disk controllers like the Apple 5.25 controller or the Duo Disk controller. The Disk II controller has an incompatible 20-pin connector, but if you used the Floppy Emu’s adapter cable to connect a Unidisk 3.5 to a Disk II controller, it would directly connect +5V to ground. Ouch!
This kind of I/O switcheroo seems like a very bad idea to me. Ideally, you could plug any kind of 19-pin Apple drive connector into any kind of 19-pin controller, and the worst that would happen is it wouldn’t work. But Apple created a situation where you can actually destroy your equipment by doing this. It’s not the first time, either. Pin 4 was similarly repurposed, from a ground connection on Unidisk 5.25 and Unidisk 3.5, to a drive input signal on the Apple 3.5 drive. And pin 10 is a drive input for Macintosh and Lisa, but an output for Apple II drives.
An Unintended Voltage Divider
The Floppy Emu’s CPLD can be reconfigured to treat pin 16 as an output when in Smartport mode, with an output value of zero, to simulate a ground connection. Setting aside the potential for damage this presents to a Macintosh connection, it should get the Unidisk to recognize there’s another Smartport device on its daisy chain connector. Unfortunately it doesn’t work. Ironically it’s the CPLD protection resistors that were added in Floppy Emu Model B that cause the problem, by creating an unintentional voltage divider with the Unidisk’s pull-up resistor on pin 16.
All of the Model B’s CPLD inputs have a 1K series resistor to help protect against voltage spikes and static. This is fine when the inputs are actually inputs:
The CPLD input buffer draws only a few microamps of current. From V = IR, we can calculate that the voltage drop across the resistor will be a few microamps times 1K, or a few millivolts total. If the computer drives a 5V input signal, the CPLD will see something like 4.99 volts, which is fine.
Things are quite different when the input becomes an output, and that output has a relatively strong pull-up resistor:
Now there’s a path through the two resistors, from 5V to ground. From the voltage divider rule, we can calculate that the voltage at the point between the two resistors will be 1.66 volts. (I measured it at 1.55 volts experimentally.) That’s far too high to be recognized as a logical zero value; 0.8 volts is the maximum valid zero voltage for 5V TTL logic. So the Unidisk doesn’t think there’s a Smartport device on its daisy chain connector, and the Floppy Emu never gets a Smartport address.
I was able to get daisy chaining working by adding a small value resistor between HDSEL and ground, external to the Floppy Emu. But that’s not much help to anybody, and it also prevents the Floppy Emu from working correctly in 3.5 inch disk emulation mode.
Solution?
So what’s the answer here? I’m afraid there probably isn’t one, and Unidisk 3.5 daisy chaining just won’t work, wah-wah and sad trombone. But maybe a reader will have a clever suggestion.
Changing the Floppy Emu’s protection resistors to something less than 1K could help. My math says a resistor of 381 ohms or less would put the pin 16 voltage at a valid logical zero for 5V TTL. By combining an old Floppy Emu Model A (no resistors) with some manually-wired external resistors, I was able to directly confirm that 1K ohm protection resistors don’t work for Unidisk daisy chaining, but 330 ohm resistors do work. But dropping from 1K to 330 ohm would be a significant reduction in the amount of protection for the CPLD. I’m also reluctant to make any changes to the Floppy Emu hardware design, which has become like a supertanker that’s difficult to change course. Any changes now would cost lots of time and money, and wouldn’t help owners of existing hardware anyway.
Another possibility is some kind of external adapter, with a physical switch for shorting pin 16 to ground. That would work, but the time needed to design, build, and stock such an adapter would be too high relative to the importance of Unidisk 3.5 daisy chaining. It’s unlikely that many people would be interested in buying such an adapter.