First post, by mkarcher
The availability of 50-pin SCSI hard drives, especially working ones, is poor the last years. On the other hand, SCSI drives with 80-pin SCA connectors are cheap and common. In the end, it's all SCSI, so why not just use SCA drives in retro PCs? I bought a couple of narrow + wide to SCA adapters off ebay. They look like this:
They map the pins between the 50-pin and the 68-pin SCSI connector as one would expect. They do not short the ground pins to each other, so they should be good enough for both single-ended and differential SCSI. Let's not talk about HVD ("high voltage differential", usually just called "differential SCSI" back in the days before LVD), as there is are no HVD SCA devices, as far as I know. So, don't make the mistake of looking at HVD pinouts when you try to understand the connections on this adapter. LVD and HVD use different pinouts, with LVD being very similar to single-ended if you ground all negative data pins. This adapter routes 75 percent of the traces on the front side, and the remaining 25 percent on the back side. The back side just contain the traces that need to be routed through the pins of the wide SCSI connectors to the respective last pin row. The layout of this adapter looks quite nicely and shows how the different connectors relate to each other. Wide SCSI just having some extra pins at each end, and being very similar to narrow SCSI in the central part. SCA being very similar to wide SCSI, but without the TERMPWR and GND connectors in the center, but having a lot of power pins on the outer pins instead.
These adapters worked quite well with my 1542CF, but failed to work properly on Disruptor's 19160. Luckily, Disruptor has active terminators with LEDs indicating the bus operation mode (LVD or SE). With an SE-only hard disk connected to that adapter, the terminator still showed "LVD". This is wrong. Running the bus at LVD with a SE drive connected explains why the SCSI bus scan in SCSISelect hung. Connecting a non-SCA SE drive to the same bus made the bus run in SE mode, and also made the SCA drive work. The same is also happens with LVD SCA drives: They only work when you connect an SE device to the same bus.
This description likely points already to one primary suspect: The DIFFSENS line on the SCSI bus. It works like this: single-ended devices are supposed to connect DIFFSENS to ground. LVD devices do not pull the line to any specific voltage. LVD terminators try to pull DIFFSENS to 1.3V. If they manage to do that, they know that no SE device is on the bus. In that case, the terminator operates in LVD mode an all connected LVD-capable device sense the 1.3V level on DIFFSENS and enable LVD operation. On the other hand, if the DIFFSENS pin is not driven to 1.3V, the bus is operating in single-ended mode, and LVD-capable devices operate in single-ended mode. In case DIFFSENS is not reaching an LVD-capable drive, that drive doesn't detect 1.3V and will operate in SE mode. Also, a SE drive connected to the bus with the DIFFSENS line not connected is unable to force the bus to single-ended operation. So this matches the observed behaviour: Devices only work if the bus is forced into SE mode by some other device. And now take a look at the pin-outs of SCA and classic SCSI cables: DIFFSENS is near the end on SCA (pin 46, the 6th pin on the second row of pins), but close to the center on wide (DIFFSENS on 16 of 34 pins on the first row of contacts) and narrow classic SCSI (DIFFSENS on the 11th of 25 contacts on the first row). A trace connecting DIFFSENS would need to snake around some connectors or routed using vias to cross other traces, but no such trace is visible. A multimeter confirms: No continuity on DIFFSENS.
This helps:
Luckily, the DIFFSENS pin is at the outermost row of pins on the SCA connector, making it easy to add a bodge wire to that pin on the SCA component side, so soldering this added wire is an easy fix. It has been tested to make both SE drives and LVD drives, even LVD drives jumpered to "force SE" work perfectly with that adapter.