Switchable SCSI termination

[mikro]

Hi,

since there is a lot of hardware experts, perhaps someone can give an advice. There is a few SCSI internal boards for the Falcon like this one or this one.

The trouble with them is nicely explained by Daniel here: http://atari-forum.com/viewtopic.php?f= ... 38#p383152. Basically if you have an internal SCSI device in Falcon (as replacement for the resistor arrays), you have to:

• Terminate the external SCSI connector
• Terminate the device connected to the internal SCSI connector

By no means you can't (as far as I understand it) put a terminator in-between the device and the host controller. Correct?

Well, my question is - what if that terminator (resistor arrays) is not powered? So if I take the device out, I would just switch on +5V to those resistor arrays, making them work as a SCSI terminator again. And if I put the device in, I would just cut +5V, making them "dormant". Is this a plausible way of thinking or should I just remove the resistor array completely if I put the device in?

[exxos]

Termination AFAIK is done between signal and 0V not 5V.. Though there are a few ways this can be done.

A summary..

It is there to stop oscillations/reflections on the signal which causes noise/under & overshoot etc. So termination puts a load on that signal to prevent it. However this increases the current though the signal which actually contributes to the problem and increases RF noise etc. If you have 47R resistor on 5V then you pull 106mA max on that pin.. In terms of that SDMA termination resistor on the Falcon.. its one reason I advise against it.. 106mA is a lot of current for the combel to drive. Also I found it clamps to 0.5volts so 47R is just all round bad idea IMO.

Placing "in line" resistors is the "preferred" method in a lot of designs. You will see for example 22R,68R in series with MADx lines on the ST ram circuit. I put resistors in series with all clock lines because of such oscillations.

I have used SCSI stuff in the past for years, I never used the termination block. I'm not saying its not needed, and may well be needed in some situations. But in general "in line" resistors should be investigated first.. generally these are only needed when long tracks or cables are involved anyway.

There are huge technical documents on top manufactures sites which go into great detail about all this. But it would be better to just scope out the signals and build in "fixes" if & when needed. I don't think we are in the age of hooking up miles of cables and multiple scsi drives anymore anyway.

[PaulJ]

@mikro , SCSI II is pretty easy to use. It must be terminated on each end of the bus which shouldn't exceed 3 meters. The falcon is terminated internally so you don't need to worry about that end. So for example if you have three additional SCSI devices on the bus the last one on the bus should be terminated. The items in the middle of the bus should not be terminated. Each device must have a unique address. Been using SCSI for a lot of years and its very reliable and easy to setup if you follow the simple rules. Most devices have inline plugin resister packs and must be be disabled via jump or unplugged on the devices that are not a endpoint.

Falcon termination.

scsi.png (169.91 KiB) Viewed 3951 times

[mikro]

The falcon is terminated internally so you don't need to worry about that end.

Actually, I do. The modification I'm about to make removing that end (the 3 resistor packs) so the game is a bit changed now. I want to connect a SCSI device internally (which is itself terminated) with an option to remove it if/when I like. So ideally with just a simple switch instead of adding/removing the resistor pack.

[PaulJ]

The falcon is terminated internally so you don't need to worry about that end.

Actually, I do. The modification I'm about to make removing that end (the 3 resistor packs) so the game is a bit changed now. I want to connect a SCSI device internally (which is itself terminated) with an option to remove it if/when I like. So ideally with just a simple switch instead of adding/removing the resistor pack.

Don't know what sort of device your adding but a lot of times there's a jumper to enable the device termination so you could run those pins to a switch to enable/disable the termination if that jumper exists.

[mikro]

Don't know what sort of device your adding but a lot of times there's a jumper to enable the device termination so you could run those pins to a switch to enable/disable the termination if that jumper exists.

SCSI CosmosEx but it applies also to internal SCSI HDD. The problem isn't the device, that is properly terminated. The problem are those resistor packs, which we are replacing with an internal SCSI board. If the device is connected, they shouldn't be active, if the device isn't connected, they should be.

So my question is whether this is possible to achieve with only (not) powering the resistor packs or I really have to put in / take out the resistor packs as needed.

[PaulJ]

I thought the SCSI CosmosEx was an external device but for conversation sake lets try this. Leave the falcon termination in place, plug a SCSI termination into the falcon SCSI port on the back of the falcon, remove termination from the CosmosEx and connect it internally to the SCSI bus and all should be fine with or without the CosmosEx. The SCSI interface is designed to drive two terminations. Never less, never more. Whatever you connect in-between those two terminations is fine as long as they have their terminations disabled or off.

[mikro]

Leave the falcon termination in place, plug a SCSI termination into the falcon SCSI port on the back of the falcon, remove termination from the CosmosEx and connect it internally to the SCSI bus and all should be fine with or without the CosmosEx. The SCSI interface is designed to drive two terminations. Never less, never more. Whatever you connect in-between those two terminations is fine as long as they have their terminations disabled or off.

CE 3.0 is specifically built to fit into the Falcon.

So you are saying that (external terminator) -> (external SCSI port) -> (host controller) -> (our SCSI board with resistor packs which is placed on P1/P2/P3 pins) -> (non-terminated CE) is totally legal and supported?

[PaulJ]

Leave the falcon termination in place, plug a SCSI termination into the falcon SCSI port on the back of the falcon, remove termination from the CosmosEx and connect it internally to the SCSI bus and all should be fine with or without the CosmosEx. The SCSI interface is designed to drive two terminations. Never less, never more. Whatever you connect in-between those two terminations is fine as long as they have their terminations disabled or off.

CE 3.0 is specifically built to fit into the Falcon.

So you are saying that (external terminator) -> (external SCSI port) -> (host controller) -> (our SCSI board with resistor packs which is placed on P1/P2/P3 pins) -> (non-terminated CE) is totally legal and supported?

Yes, normally in a larger SCSI system there are numerous devices some of which may or may not be powered up. As long as both ends are terminated all is well. I personally wouldn't screw with the falcon termination since it is the real end. I would run the elf unterminated. You can try differient configurations just try to key both end terminated and make sure your using a SCSI storage medium you can afford to get corrupted. If your only using the elf just connect it and use its termination and for get the termination on the SCSI port. Now if you try and add another SCSI device you can't use that configuration because your terminations would no longer be at the ends.

[ari.tao]

Gotta CE3 me too, probably for Falcon.
So, having kinda same problem now.
When another SCSI device externally is plugged on, then that end of the SCSI chain is undaubted and must be terminated there, ie. externally.
My problem is, when CE3 is installed internally in Falcon, what is the other end of the chain? Is it the Falcon board as suggested by it´s terminators here:

So you are saying that (external terminator) -> (external SCSI port) -> (host controller) -> (our SCSI board with resistor packs which is placed on P1/P2/P3 pins) -> (non-terminated CE) is totally legal and supported?

or is it the CE? ie.
Ext.term. <-> ext.dev. <-> ext.port. <-> F30 board without term. <-> CE with term.
How to determine?
May be either possibility works, because as You certainly know, in every SCSI chain with at least 3 members, necessarily the middle one(s) are little T-branches. So the first suggestion just appears to have it´s T a little bit extended - afaik 10cm are allowed (and in praxis may be 15cm are tolerated). Should be decided by experiment?

In my TT I was facing nearly the same Problem as I have a Yamaha SCSI2IDE installed internally and wanted to run the TT sometimes with an external SCSI device plugged on too. So I unplugged the TT´s internal terminator arrays and jumpered the Yamaha to terminate SCSI. It´s working well for years now. If I remove the external SCSI device, I can plugon a terminator plug, but it appeared not to be necessary - seems to work well without a second termination. I think, that´s because the internal SCSI cable is rather short. So my hope is, this might appear in the Falcon similar. But then, if at any day the CE is removed, then again the resistor arrays for termination should be reinstalled. Better make them pluggable.
I don´t think, just to switch off the vcc for the terminator resistor arrays is a good idea: Then obviously the resistors between the datalines are still there, so if any of the datalines is active, it brings noise to others.
Does anybody of You know, how the term.jumpers on SCSI devices function? May be they have active termination?