One more time about Zorro bus termination

[tnt23]

Last night I was fiddling with the riser board in my A3000, and have discovered that on its Zorro slots, there are mere 4.6 volts instead of 5. Perhaps cleaning the mobo slot contacts would help, or maybe adding a separate +5v path from the PSU to the riser could do, but here's another oddity I quickly was distracted by.

It had always puzzled me how hot the A3000 guts became after half an hour work. Yes, this is good old 74F series, fast and dissipating lot of heat, but the riser board gets hot, too, with all boards installed in it.

I know most bus lines here have thevenin termination (see attached picture from http://amigadev.elowar.com/read/ADCD.../node0298.html). A pair of 220/330 Ohm resistors, pulling line to +5 and common ground. Must consume ~100mA per line, or roughly 4A ~10mA per line, or roughly 0,4A for all terminated lines, and all that has to be dissipated. Oh well.

(I have read somewhere that the dual termination resistor packs could have been soldered upside down, giving 330/220 Ohm instead and resulting in 2v on the inactive line. I can observe less than 2v in my case, but that also can be a result of overall +5v rail drop. Have to check this)

On the host side of things, there is a number of 74F646 to drive the bus. I was under the impression that they should be quite powerful to drive the terminated lines from their ~3v potential to as close to +5v rail as possible for TTL. Not as close as if it were CMOS but still way over 4 volts.

However, even with no cards in the riser, I can see the line driven hardly over 3 volts. I will post some shots later. This does not seem right to me, but of course I can learn a lot about logic levels and bus termination.

All of which make me quite puzzled. From what I know about TTL logic levels, one'd better start at 2.5v or higher, and the higher the better. Even with this termination which came from SCSI world, shouldn't them high levels be closer to 5v?

Or is this something specific to my own setup? Bad mobo 74F drivers maybe? I have a spare riser and it behaves just the same.

-----------------
I know there was a thread by chiark (http://eab.abime.net/showthread.php?t=55793), but it was on 1000 bus so I did not dare resurrecting it BTW active termination mentioned there is of some interest I think.

[rockape]

Hi tnt23,

You mentioned in your post about

"I have read somewhere that the dual termination resistor packs could have been soldered upside down, giving 330/220 Ohm instead and resulting in 2v on the inactive line. I can observe less than 2v in my case, but that also can be a result of overall +5v rail drop. Have to check this"

I'm no expert but have found this site to be helpfull.

See http://wonkity.com/~wblock/SCSI/SCSIExamples.html

Quote

"Specific Troublesome Devices"

"C. Termination. Various A3000s seem to have come with no terminators, soldered-in SIPs, or even sockets. Check the motherboard controller termination, and follow the guidelines laid out in the Termination section above.

D. Many A3000s had a manufacturing flaw which resulted in terminator power not being supplied at the external SCSI connector. The easiest way to test this is with an external terminator with an LED indicator. Otherwise, you'll need to check pin 25 of the A3000 DB25 SCSI connector for +5V (the shield around the connector provides an easy ground test point).

Before testing, you should disconnect all other SCSI devices, both internal and external, because one of them may also supply termpower and confuse the results. If no voltage is supplied on pin 25, diode D800 (or D801, this may vary depending on motherboard revision) is reversed inside the A3000. Unsolder and replace it (this should be a 1N34 type, although a 1N5817 should work and might be more suitable).

The motherboard silk screen is likely to be wrong as well, so ignore it."


I hope this helps,


Regards, Michael

aka rockape

[tnt23]

Michael, thanks for the link. To be honest, it is regarding SCSI, and I seem to have trouble understanding Zorro bus physical aspects.

However, you made a great point regarding SCSI: I can take a look at it and see what is going on there, and probably how I could use SCSI active terminators to play around

[Vot]

[tnt23]

74F646 bus drivers are TTL though. I will see if I have any 74F chips in the junkbox to build a test bench.

[SpeedGeek]

@thread

tnt23 is asking for information on Zorro3 bus termination not SCSI bus termination!

@tnt23

I had the same problem with my A3000. The 220/330 Ohm resistor packs were installed backwards. You have to unsolder them and it makes sense to replace them with 330/440 Ohm resistor packs (see the Zorro3 Spec. addendum).

This will reduce the riser/daughter board power consumption a bit and improve performance and reliability (especially when the bus load increases with many Zorro Boards installed).

Unfortunately, the A3000 PSU connector can cause a voltage loss also (with no easy fixes). I cleaned the pins and contacts, crimped the contacts (for a temporary fix) and finally added an extra wire from the PSU to the Riser/Daughter Board.

It probably isn't worth the effort to replace the 74F series logic, but you can easily replace the 4 mobo PALs with GALs to reduce power consumption a bit more.

[tnt23]

@SpeedGeek thanks for showing me to the addendum. I should have studied it better How do I tell if my resistor arrays have been incorrectly soldered? They bear no markings, at least I cannot see any; and measuring them is somewhat useless. I get something like 134 Ohms on both resistors in each of array's pair. Bourns datasheet recommends using 'guarded ohmmeter' to measure array resistors' values correctly.

Ok, here's a picture showing one bus read cycle with /FCS (yellow) and AD20 (cyan).

Address is being driven on the bus just before the /FCS asserts. AD20 line can be seen raising to 3v, then it tristates and rolls down to approx. 1.8v. There is no slave to drive data on the bus, so AD20 keeps floating till the end of the cycle. I have no idea why it gets driven high again along with deasserting /FCS, but that does not matter.

[bubbob42]

Also keep in mind that the old PSU might need a rework, since the capacitors are getting old, too. A voltage drop is a typical sign for this.

[tnt23]

Originally Posted by bubbob42

Also keep in mind that the old PSU might need a rework, since the capacitors are getting old, too. A voltage drop is a typical sign for this.

Peeking inside the PSU is something I have been avoiding from when I bought this A3000 in 2007. However it does develop 5.1v measured on HDD connector, this makes me think main voltage drop occurs down the path 'mobo connector - expansion slot'.

[Stedy]

Hi,

It looks like the terminators in your A3000 are backward. The bus should 'park' at 3V when not in use, if it parks around 1.8V with a 4.6V supply, you will increase power in the logic slightly and are dangerously close to TTL switching thresholds.

When using Thevenin termination, you want to 'park' the bus away from TTL or CMOS switching thresholds, 2.0V Vih and 0.8V Vil for TTL or 3.33V for Vih and 1.67V vil for 5V CMOS. This provides margin for component tolerances, aging and higher current draw.

Measuring the resistors packs in circuit will be tricky as you have multiple packs in parallel. You may be down to 10's of ohms.

If you want to replace the resistor packs, look for 5V VME terminators, 330/470 ohm packs.

Clean all your connectors with contact cleaner and if necessary, try a separate, parallel, power feed. Have you measured the ripple on the +5V power supply?

[bubbob42]

Originally Posted by tnt23

Peeking inside the PSU is something I have been avoiding from when I bought this A3000 in 2007. However it does develop 5.1v measured on HDD connector, this makes me think main voltage drop occurs down the path 'mobo connector - expansion slot'.

Hrm, I've seen some A4K PSUs with +5V at the drive connectors and lots less at the ICs at the clock area. But motherboard connector should be cleaned with isopropanol anyway...like Stedy said.

[tnt23]

@Steady

I will also check the ripple on +5V rail and clean the connectors. I have tried a separate power feed to the riser, it added 0.3v but I did not like the idea of the loop it might create, and that the rest of the system might have been feeding via the riser.

@bubbob42

All right, time to crack open the PSU then. Will have to clean it anyway

Last night I took a closer look on my resistor packs, and in fact they do have markings. It is on the other side which was tricky to read. And if I read it correctly, pin 1 is where silk on the board says it should be. However, pin 1 measures +5v (or less in my case), while the schematics says pin 1 should be ground. And vice versa for pin 10 (shows ground on the board, should be +5v).

So it appears that the silk on the board is wrong, and resistors have been soldered according to it incorrectly. I will re-solder them.

[hooverphonique]

Originally Posted by tnt23

So it appears that the silk on the board is wrong, and resistors have been soldered according to it incorrectly. I will re-solder them.

indeed it points in that direction - and it will not be the last time C= gets the silk screen wrong, either

[tnt23]

@hooverphonique

Anyway, I have ordered new Bourns dual term 330/470 packs. They will take some time to arrive, so I de-soldered old 220/330 packs. They look rather rusty, I had tough time to cover their legs with solder as it just would not stick. And tada! I like this pic.



Yet I must say I don't understand why the '1' level is just slightly higher than the bus idle voltage.

[tnt23]

And here's the result of replacing 220/330 packs with 330/470.

[pandy71]

Perhaps stupid question but what is wrong with active termination?

http://www.linear.com/docs/4222
https://www.fairchildsemi.com/applic.../AN/AN-610.pdf
http://www.retrotechnology.com/herbs_stuff/s_term.html

[tnt23]

Originally Posted by pandy71

Perhaps stupid question but what is wrong with active termination?

http://www.linear.com/docs/4222
https://www.fairchildsemi.com/applic.../AN/AN-610.pdf
http://www.retrotechnology.com/herbs_stuff/s_term.html

Nothing is wrong with active termination, perhaps except that it is not used in stock Zorro daughterboards?

[pandy71]

Originally Posted by tnt23

Nothing is wrong with active termination, perhaps except that it is not used in stock Zorro daughterboards?

And?

[tnt23]

Originally Posted by pandy71

And?

And so much for active termination

[pandy71]

Originally Posted by tnt23

And so much for active termination

My point is that only backplane is terminated not cards - and this is quite strange...