vrs wrote: The bus driver/receiver voltage spec (absolute maximum voltage of 2.5V) of these parts is NOT suitable for UNIBUS/QBUS. These parts are designed for low voltage swing busses and may work in the short term on UNIBUS, but will eventually fail, as the UNIBUS termination voltage (static high level) is 3.5V, 1V over the abs max rating. I've looked at this problem a bit and the most suitable modern day (ie, still in active production) part I've found is the Am26S10C. TI now produces this older AMD part, and it is available in both 16DIP and 16SOIC. It is a quad part not unlike the DS8641. You can get them at Digikey for about US$1.35 each. Even in Pb-free packages. The 26S10 has 100ma OC drivers, and the bus receiver input threshold is in the 1.75-2.25V range, very close to UNIBUS specification. The only issue I see in using this part is that the specs of the output rise/fall time are sub 10ns transitions. This may cause problems with reflections, especially on UNIBUS systems with longer cables. Don

From: "Wai-Sun Chia" <waisun.chia at gmail.com> (I think you meant the AM26S10C.) It would, if there were any such parts! That was why I was excited about the sn75als057 (which apparently won't work). Vince

This is my main concern as well; the Am26S10C may be just a little to 'hot' for some UNIBUS configurations. My suspicion is it will work just fine for most configurations, but some (lightly loaded using longer UNIBUS cables) may have reflections/ringing problems that will manifest themselves as 'flakiness'. The Vih(min) of the 26S10 is 2.25V, similar to that of the DS8837/8 parts with Vih(typ) of 2.5V. It is still about 0.5V higher than the 'gold standard' DS8641 tho. If you look in the UnibusTroubleshooting guide on bitsavers, they have several charts of allowed AC loads vs static high level. Allowing for 200ua per load leakage, the minimum static high level for a 20load segment works about to be about 3.0V (down from 3.4V which is the termination voltage). The issue will be that due to imperfect termination, negative ringing on the high level now has to be less than 0.75V (3.0V - 2.25V) with the 26S10 versus 1.3V (3.0V - 1.70V) with the 8641. So the 26S10 receivers will be more sensitive to ringing on L-H transitions, possibly causing false transitions on the output. This is probably not a concern for data/address lines given adequate deskew delay, but more problematic for such lines as MSYN/SSYN/etc that are actually clock edges. I don't have a good solution to this other than don't configure heavily loaded / very long UNIBUS segments if you plan to use the 26S10. I don't see another part that is usable however (other than salvaging old 8641s).

On 2/6/06, Don North <ak6dn at mindspring.com> wrote: For Unibus projects, I happen to have tubes of NOS 8641s left over COMBOARDs with several 8641s each, if one wants to pull them from a 6-layer board). We used 8641s and DC013s for our Unibus interface layer (and real DEC DC005s and DC010s for our Qbus boards). I don't think I could supply all the Unibus hobbyists in the world, but I have enough chips for a limited run of new boards. Still, I think it'd be nice to find something modern that works for bus drivers and bus receivers. -ethan

I looked at the 7438 for drivers and the TL3116 comparator for the receiver for a while. I found that comparators were pretty spendy if you get them fast enough. I also had some trouble finding a quad part for an "ultra-fast" comparator. The worst problem then was with the real-estate. You really don't want a zillion 8 pin parts on the boards :-). (There were some inexpensive possibilities that weren't rail-to-rail, and so would need to run off the +/-15V or something.) I think we may be forced down that path when parts like the Am26S10 are no longer readily available. Vince

Nice :-)! Agreed. Among the off-the-shelf interface parts, I couldn't find anything better. (Wish the 75als057 had higher voltage tolerances!) Vince

Me either. I do think I could make a Q-bus receiver from a linear IC, like an "ultra-fast" comparator, pretty easily. Vince

I have done extensive research on this issue. If you happen to take a look of the schematic of the mscp scsi project, you know my choice. It is the plain TTL 7406. I can defend this choice for Qbus. It satisfies the majority of the requirements, and has work around of the rest. 1. availability, price 2. leak current satisfies requirement (Yes, I measured. 0.25mA specification is for 30V, not 5V) 3. rise/fall time satisfies requirement (This is the most important factor) 4. drain current work around. use one for single terminated QBUS, parallel two for double terminated QBUS. (when in parallel, leak current doubles; but the QBUS tolerance doubles too.) vax, 9000

From: "Don North" <ak6dn at mindspring.com> factor) Isn't the Iil value a couple of orders of magnitude out of specification (for a receiver) too? Vince

On 2/6/06, Don North <ak6dn at mindspring.com> wrote: I use ALTERA EPM1270 CPLD to receive QBUS directly. It has 4V absolute max input level but I can enable the internal diode to limit the possible voltage surge (in PCI pin compatible mode). I set the input level to 3.3V LVTTL for a time and it worked OK too. vax, 9000

From: "9000 VAX" <vax9000 at gmail.com> Ouch. I'd at least use the 74HCT14, to reduce the bus loading. That way, the problems will show up with your device, rather than some other :-). But you still won't get much noise immunity from the 0.6V Vil. Vince