838 - IO Card (Flex Interface card replacement!)

[LASooner]

V2.0 Built in Speech ?

[+dhe]

I'm at the SUATMM point! ?

[InfiniteTape]

21 hours ago, LASooner said:

V2.0 Built in Speech ?

I like the way you think.

[Shift838]

Here is my latest updated.  I am still testing the latest version that PCB version that I just received and assembled but wanted to post an update.

 

I have redesigned the card a bit to use a HD37 shielded cable instead of the ribbon cable (I should have known better!).

 

The ribbon cable worked fine at no more than about 3.5 feet, but after that nada..

 

I have tested the latest version of the card with a 5 foot HD37 cable with no issues so far. I have found one correction I will make on the next run of boards for the side-port card.  I had a pin connected on the 44 pin connector (pin 1 - VCC) that should not have been connected.  

 

What I have done to this version:

 

 

• Use a HD37 connector instead of IDC40
• added 2x 3 pin 5v fan power connectors
• Increased trace widths dramatically to help increase signal strength on both PEB board and side-port card
• This version has been renamed to the 838-IO Plus because:
  • Added expansion port to the side-port piece to be able to add a 32k, SAMS expansion memory or other expansion cards.
    • 32k card is complete and working
      • Next version of the 32k will have an on/off switch.
    • SAMS needs to be designed - current side-port version of the SAMS will not work and I believe it may be because there are no buffers.  Currently redesigning.
    • There are plans for another expansion card that will also connect to the card to further move another card from the PEB to the side-port.
  • Added expansion memory LED to side-port for activity/use.

 

This design has increased the height of the board to be a bit higher once in a case (yet to be designed) than the 99/4A console.  In order to get the trace widths where I wanted them this had to be done.

 

Here are some photos.  I also have a through hole version of the side-port card but the through-hole version cost more to build based on the components.

 

Note on the side-port card there looks to be a bridge connection on the 44-pin edge card connector, but it's not.  it was lint from a rag when I cleaned the alcohol spray off.  Did not see it until after I already had the photo on the PC.

 

Once I get it passed for my testing then I plan on building one and sending it to another TI'er for testing before I release it into the wild and start taking pre-orders.

 

 

 

 

[GDMike]

It really looks beautiful

[+dhe]

Common TI, you couldn't have found it in your heart, to give us 8 more wire for a 16bit data bus? ?

Thanks for the update, nice looking card, I'd be happy to pay more for gold plating and machine screw sockets.

 

[MikeV]

On 12/19/2021 at 3:30 PM, Shift838 said:

What I have done to this version:

• Use a HD37 connector instead of IDC40
• added 2x 3 pin 5v fan power connectors
• Increased trace widths dramatically to help increase signal strength on both PEB board and side-port card
• This version has been renamed to the 838-IO Plus because:
  • Added expansion port to the side-port piece to be able to add a 32k, SAMS expansion memory or other expansion cards.
    • 32k card is complete and working
      • Next version of the 32k will have an on/off switch.
    • SAMS needs to be designed - current side-port version of the SAMS will not work and I believe it may be because there are no buffers.  Currently redesigning.
    • There are plans for another expansion card that will also connect to the card to further move another card from the PEB to the side-port.
  • Added expansion memory LED to side-port for activity/use.

I would definitely be interested in one of these, especially with the SAMS feature. What other cards can be moved to the side-port? Thanks.

[Shift838]

12 hours ago, MikeV said:

I would definitely be interested in one of these, especially with the SAMS feature. What other cards can be moved to the side-port? Thanks.

Currently the side port cards can be like the 32k memory and the SAMS sideport.  I do have another card on paper that I need to prove out.  But I have a couple of other projects ahead of it.

 

 

[Shift838]

So during my vacation I decided to work on getting this card finalized and ready for production and pre-orders.  I think over the last year this has been the busiest I have been with TI hardware projects!  And I still have more on paper I need to work on.

 

A couple of problems I was having with the card:

 

1. When testing with an IDE or SCSI controller I would periodically get errors when copying files from the TIPI over to either of them or even from one partition to another.
   • I would get errors from DM2K of could not read from source or could not write to the destination or something along those lines.
2. The other error I would get would be even though the IDE card passed the test, loaded the DSR and then trying to partition a CF card DU2K could not find the IDE card.
   • I also noticed when this happened that if I did a CALL IDEST in basic then even though the CF Card was already partition, it would not show any setup.

 

Well that got me thinking that the cable I purchased even though it was shielded and all that the cable was just not up to par and I was losing signal somewhere.

 

Well that was it!

 

FYI when I make the cables they will be between 4 and 5 feet.  Mine I am currently testing with is 4.5 feet.

 

I purchased a new SCSI cable and soldered the DB37 ends on it and grounded the shielding and WINNER WINNER CHICKEN DINNER!

 

I can now partition, load, execute and copy.  no errors anywhere.

 

I have copied about 240mb of programs all through DM2K from my TiPi directly to my IDE card.  I then copied it all from the IDE1 to IDE2 with no issues.

 

I have tested with the TI 32k card, SAMS PEB card, 838 Sideport 32k card and the sideport SAMS card with no issues.

 

There are still a few things I need to do before I start taking pre-orders.

 

1. Create a 3D case for this one!

2. move the 3 pin pc fan headers to be horizontal headers and relocate them to a place where they do not interfere with each other.

3. Make a short video of it working!

4. Work out a final price that will include an entire unit including cable.

 

Picture of the PEB and sideport card are in Post #54 above:

 

 

 

 

 

 

[+9640News]

Do you know if one card is plugged into another card in a separate PEBox, you can access those cards as well?  Thinking here about a Geneve and  being able to access more devices. 

[Shift838]

30 minutes ago, 9640News said:

Do you know if one card is plugged into another card in a separate PEBox, you can access those cards as well?  Thinking here about a Geneve and  being able to access more devices. 

it works with the same theory of operation as the original flex interface card.  you would need a splitter just like the original.  But I do need to to test that type of setup.

 

Guess I need to find me a PEB splitter card.  

Edited March 25, 2022 by Shift838

[wierd_w]

If you dont mind sharing a STP file of the PEB card's PCB (with some general keepout clearances), I can probably modify the existing data I have for my generic PEB shell, to produce a printable housing for your interface card.

 

Or, I suppose you could just lift the STP files for that housing, and modify it yourself.  It might save you some time.

 

 

Given that you are using energy efficient power regulators, PLA should be an acceptable material.  The big shells print quite nicely when PLA is used.  It's the "tolerates heat" materials that warp so damn bad. 

[Shift838]

3 minutes ago, wierd_w said:

If you dont mind sharing a STP file of the PEB card's PCB (with some general keepout clearances), I can probably modify the existing data I have for my generic PEB shell, to produce a printable housing for your interface card.

 

Or, I suppose you could just lift the STP files for that housing, and modify it yourself.  It might save you some time.

 

 

Given that you are using energy efficient power regulators, PLA should be an acceptable material.  The big shells print quite nicely when PLA is used.  It's the "tolerates heat" materials that warp so damn bad. 

The main housing I was talking about is the one for the card that will plug on vertically to the 99/4A side port.  The PEB card is currently designed to be long enough to go into the card guides on the peb cage.  So no housing needed for that one.

[wierd_w]

Ahh.  Well, I have the generic shoe style for the beige speech also.  I can make a "narrow" version of that quite easily.

 

(nevermind me, I tend to prattle. I just really like this project, and want to be helpful any way I can. I lack the skills to contribute other ways. Enclosures I can totally do though)

Edited March 25, 2022 by wierd_w

[+9640News]

5 hours ago, Shift838 said:

it works with the same theory of operation as the original flex interface card.  you would need a splitter just like the original.  But I do need to to test that type of setup.

 

Guess I need to find me a PEB splitter card.  

Wait a minute.  Wasn't thinking of adding a PEBox splitter card.  Rather, plug one card into one PEBox, and then another card into the other PEBox, and use connectors on both ends of your cable to connect the two cards together.  If it works, then I think you are only offering 2 cards+1 cable for a Geneve setup.  That is if it works.  I know it has been mentioned elsewhere there is a limit to the number of cards that can be in a single PEBox.  I just did not know if you bridged two PEBoxes if this would work or not.

 

Beery

 

[+Ksarul]

The PEB Splitter board is the point that joins two flex cable cards together @9640News. The ones in circulation now also allow connection to the standard TI sideport connector. Building a larger PEB backplane for use with a Geneve would be a bit more difficult because it doesn't use the flex cable interface, although it might be possible to set up a segmented backplane to allow the insertion of a set of drivers for the second half of the backplane between two groups of seven slots.

[+dhe]

 

I purchased a new SCSI cable and soldered the DB37 ends on it and grounded the shielding and WINNER WINNER CHICKEN DINNER!

 

>>So does that mean you are going to need to make every cable? and there are no quality cables out there? Because that would sub-optimal.

[Shift838]

17 hours ago, dhe said:

 

I purchased a new SCSI cable and soldered the DB37 ends on it and grounded the shielding and WINNER WINNER CHICKEN DINNER!

 

>>So does that mean you are going to need to make every cable? and there are no quality cables out there? Because that would sub-optimal.

I have tried multiple pre-made cables out there and none seem to work.  but I am still looking.  If I find one more then I will try that, otherwise, yes all cables will have to be hand made.

[RXB]

On 12/19/2021 at 12:30 PM, Shift838 said:

Here is my latest updated.  I am still testing the latest version that PCB version that I just received and assembled but wanted to post an update.

 

I have redesigned the card a bit to use a HD37 shielded cable instead of the ribbon cable (I should have known better!).

 

The ribbon cable worked fine at no more than about 3.5 feet, but after that nada..

 

I have tested the latest version of the card with a 5 foot HD37 cable with no issues so far. I have found one correction I will make on the next run of boards for the side-port card.  I had a pin connected on the 44 pin connector (pin 1 - VCC) that should not have been connected.  

 

What I have done to this version:

 

 

• Use a HD37 connector instead of IDC40
• added 2x 3 pin 5v fan power connectors
• Increased trace widths dramatically to help increase signal strength on both PEB board and side-port card
• This version has been renamed to the 838-IO Plus because:
  • Added expansion port to the side-port piece to be able to add a 32k, SAMS expansion memory or other expansion cards.
    • 32k card is complete and working
      • Next version of the 32k will have an on/off switch.
    • SAMS needs to be designed - current side-port version of the SAMS will not work and I believe it may be because there are no buffers.  Currently redesigning.
    • There are plans for another expansion card that will also connect to the card to further move another card from the PEB to the side-port.
  • Added expansion memory LED to side-port for activity/use.

 

This design has increased the height of the board to be a bit higher once in a case (yet to be designed) than the 99/4A console.  In order to get the trace widths where I wanted them this had to be done.

 

Here are some photos.  I also have a through hole version of the side-port card but the through-hole version cost more to build based on the components.

 

Note on the side-port card there looks to be a bridge connection on the 44-pin edge card connector, but it's not.  it was lint from a rag when I cleaned the alcohol spray off.  Did not see it until after I already had the photo on the PC.

 

Once I get it passed for my testing then I plan on building one and sending it to another TI'er for testing before I release it into the wild and start taking pre-orders.

 

 

 

 

So as I do not have a Side Cart SAMS I would be forced to buy one?

What is the point of having this if my old cards will not work?

[Shift838]

2 hours ago, RXB said:

So as I do not have a Side Cart SAMS I would be forced to buy one?

What is the point of having this if my old cards will not work?

no.  this will work with the sideport SAMS, my specific 32k module or a internal original PEB 32k made by the various companies or any other PEB based memory card for that matter. 

 

 

[wierd_w]

It may sound stupid, but the obvious solution to "need a special cable", is to just connect the shield (metal outer part of the port, usually. The mounting tab/screw is usually electrically connected with the shielding. A bodge wire on one of the screws holding the port onto the PCB, routed to the card ground on both ends of the cable, should suffice to not need to hand make cables. Revising the cards to just have a trace for that instead, would remove all handwork. Test with a bodge wire and see.) to the ground plane of your interface cards, using your PCBs.

 

Because there may be reasons to isolate circuit grounds, cable manufacturers usually do not join the shield ground with data wire grounds inside their cables. (It may be desirable to send voltage spikes for direct power circuits to a ground that is decoupled from signal-level grounds, to avoid damaging sensitive components that cannot have reverse voltage bias applied, for instance.)

 

Without your cable's shield being grounded to a beefy earth ground, it at best functions as a floating ground. This is not sufficient to fully sink out all radio frequency resonation in the cable, and you have the induced signal noise/corruption problem. The inverse is also true-- if the data ground is not suitable, then data levels do not have an agreed upon ground. See also, the noise problem with the TI's tape input circuit, which expects the data recorder to bridge the grounds-- and the resulting ground hum that happens if they are not.

 

After joining those grounds together in your cards, just assert to users that they MUST buy shielded cable, and that unshielded cable is not advisable/supportable.

Edited March 28, 2022 by wierd_w

[Shift838]

22 hours ago, wierd_w said:

It may sound stupid, but the obvious solution to "need a special cable", is to just connect the shield (metal outer part of the port, usually. The mounting tab/screw is usually electrically connected with the shielding. A bodge wire on one of the screws holding the port onto the PCB, routed to the card ground on both ends of the cable, should suffice to not need to hand make cables. Revising the cards to just have a trace for that instead, would remove all handwork. Test with a bodge wire and see.) to the ground plane of your interface cards, using your PCBs.

 

Because there may be reasons to isolate circuit grounds, cable manufacturers usually do not join the shield ground with data wire grounds inside their cables. (It may be desirable to send voltage spikes for direct power circuits to a ground that is decoupled from signal-level grounds, to avoid damaging sensitive components that cannot have reverse voltage bias applied, for instance.)

 

Without your cable's shield being grounded to a beefy earth ground, it at best functions as a floating ground. This is not sufficient to fully sink out all radio frequency resonation in the cable, and you have the induced signal noise/corruption problem. The inverse is also true-- if the data ground is not suitable, then data levels do not have an agreed upon ground. See also, the noise problem with the TI's tape input circuit, which expects the data recorder to bridge the grounds-- and the resulting ground hum that happens if they are not.

 

After joining those grounds together in your cards, just assert to users that they MUST buy shielded cable, and that unshielded cable is not advisable/supportable.

My mounting holes for my DB37 already are connected to ground via a trace on the PCB.  No wire needed.  

 

I have done exactly what you are suggesting but the molded cables just simply do not work.  I think it may be because they are using a thinner gauge of wire than the twisted pair scsi cable.

 

All the premade cables I have purchased are shielded as well.  I believe I have found a place that will make me the cables and not too bad of a price.  I have ordered a test cable from them to verify it will work.  If so then I will offer two options when ordering.  buy a supplied cable from me or have one made to the specifications by the company that I am using.

 

I have been quoted for a cable of $32.00 + shipping for a 5' cable, this cable has 24gauge twisted pair and shielded.

 

 

[Shift838]

I received the cable and this new cable seems to work.  Need to run it through some more testing, but initial test are good.

 

I can detect the extra RAM, load, save and execute files from floppy and IDE as well as TIPI.  I have copied a huge file list of close to 230 megabytes from the TIPI to the IDE with no issues.

 

@wierd_w is currently working on a 3D case for me to test, so it should be soon i start the preorders.

 

Stay tuned...

 

 

[wierd_w]

Yup yup.

 

I am having to make a new enclosure from scratch, due to the dimensions of this set of cards. (the shoe-shaped enclosure I modeled previously just wont work. The cards will never fit inside.)

 

So far, I am working on the notion that I can use interlocking plastic tabs.  The boards will each get knurls/standoffs to mount to, on their respective halves of the enclosure. The boards will mate using the 44pin header, and the enclosure halves will snap together at the same time.  I have not yet modeled the tabs in Enclosure Side B. (just the receptacles in Enclosure Side A)

 

Not finished yet, but something kinda like this:

 

 

 

The window on the outside facing wall might get removed, if Shift838 says to omit it.  I am putting it there in case he decides to put a short card-edge riser, so that downstream devices could be connected.

 

The "Enable 32k" switch has a little recessed well you will have to put a long, stiff object into to reach the tab of the switch.   The reason for that, is the apparently tall height of a discrete transistor in there.

 

 

 

This arrangement may or may not work.  I believe I have given sufficient clearance to slide the main flexcard board in through the window in the back, where it will then "fall down" onto the screw-mount pads.

I will test print a set of enclosure halves out of PLA once I have tabs modeled in, to see how well they stick together.  (without a flexcard and 32k board, I cannot do a full test fitting. Hopefully these STP files are accurate.)

 

 

[Shift838]

I have run the hell out of this new cable.  it's 5 foot long and I have had absolutely no issues with it so far.  I have ran various copies from the TIPI to the IDE.  From IDE1 to IDE2, floppy to tipi, floppy to IDE, floppy to tipi, etc.  all seems rock solid.  Using a twisted pair cable works wonders.  the DB37 connector makes the board footprint a bit bigger than the 40 pin IDC connector, but it's worth it to get the stability from the twisted pairs.  Of course the twisted pairs have to be paired to certain connections on the cable to get it to work this way.

 

All in all, i'm very pleased.  I have tested it on a stock TI system console with no upgrades and another with the F18A, USB keyboard, Reset switch with no issues.

 

As soon as we have a solid case pre-orders will go up.  not sure how much yet, depends on how long the case takes to print and what type of deal I can work for the cables.  Watch for a poll i will post to try to gauge how many will be interested.  I do know the cable itself will run me about $36.00 each to have produced.

 

there will be options on the preorder to make your own cable and I will provide all the documentation to do so and what cable to buy.  The standard off the shelf cables will not work.