search

PiSCSI / piscsi

PiSCSI allows a Raspberry Pi to function as emulated SCSI devices (hard disk, CD-ROM, and others) for vintage SCSI-based computers and devices. This is a fork of the RaSCSI project by GIMONS.
https://piscsi.org
BSD 3-Clause "New" or "Revised" License
545 stars 82 forks source link

Alternative to SN74LS641-1DW #591

Closed DottiestDuke47 closed 1 year ago

DottiestDuke47 commented 2 years ago

I ordered some Bare PCBs of the rascsi from JLCPCB, and was checking the price and availibility of the other parts. I found that the SN74LS641-1DW is discontinued and very expensive, so went looking for alternatives. My question is, is the 74HCT245 a viable alternative?

akuker commented 2 years ago

Hi @DottiestDuke47 - There is a discussion about several different transceiver options here: https://github.com/akuker/RASCSI/wiki/Transceiver-Comparison

The 74HCT245 looks like it is designed to sink a maximum of 35mA, while the SCSI spec calls for 48mA.

In practice, the 74HCT245 would probably work. You would be running it outside of its design tolerances, so it will likely shorten the life of the part or you may have unreliable data.

I don't believe the SN74LS641-1 is discontinued. It has been impacted by the chip shortage though. Texas Instruments still has them available: https://www.ti.com/store/ti/en/p/product/?p=SN74LS641-1DW

I purchase these transceivers in bulk. So if you're having trouble finding them, I can share what I have.

DottiestDuke47 commented 2 years ago

Thanks for taking the time check and reply to my question - I really appreciate it. Once I receive the PCBs, I will try to assemble them with the chips I have, if they do not work, is there a way I can get in touch to purchase so chips off you?

akuker commented 2 years ago

Sure. Feel free to email me. My email is the same user name at gmail.com.

akuker commented 2 years ago

Placing this issue "on hold" until @DottiestDuke47 is able to check out their new PCB design.

akuker commented 2 years ago

Merging Issue #577 into this issue.

From @WarlockD

I was looking around for traslator chips and found the one used in the Atomic PI by chance. Its called the NTS0104 from NXP. Apparently its used JUST for stuf like this, but mainly for FPGA's or 1.8v ports.

It autosense input and outputs, supports up to 3.6v on one side and 5.5v on the other. Has up to 50mbps switching and since they are deisgned for thinks like i2c, they work with open collector inputs/outputs. It says it can clamp up to -50 mA so it could be used right on the bus.

Only catch is each on is 4 bits only so need 9 of them, also the hand soder version is the TSSOP14 so you will need a magnifying glass on this one to do it by hand:

Going to buy a few anyway. Want to see if I can make a PicoSCSI now. Sort of in the vein of the BlueSCSI NTS0104-1127256.pdf

akuker commented 2 years ago

Comment from @marciot on Issue #411 :

This is very tangential to the question at hand, but I am also rather curious whether a bidirectional MOSFET based level shifter such as the ones described in this page could work for SCSI:

https://learn.sparkfun.com/tutorials/bi-directional-logic-level-converter-hookup-guide/all

Obviously this would require an entirely new board design, and many, many little SMD components, but it would be amazing to have a sort of "universal" Raspberry Pi adapter that could interface with all sorts of retro things, not just SCSI.

DottiestDuke47 commented 2 years ago

So i received the boards yesterday and populated them with the 74HCT245D transceivers, and to my amazement it worked. I tested with a Macintosh SE for about ~45 minutes, everything functioned fine. I'm not too cure about the long term effect of overrunning this chip, but since this is my first vintage mac, I'll be spending a lot of time exploring using this RaSCSI so I can report if anything out of the ordinary happens.image

DottiestDuke47 commented 2 years ago

Once again thanks a lot for your help @akuker image

DottiestDuke47 commented 2 years ago

Comment from @marciot on Issue #411 :

This is very tangential to the question at hand, but I am also rather curious whether a bidirectional MOSFET based level shifter such as the ones described in this page could work for SCSI:

https://learn.sparkfun.com/tutorials/bi-directional-logic-level-converter-hookup-guide/all

Obviously this would require an entirely new board design, and many, many little SMD components, but it would be amazing to have a sort of "universal" Raspberry Pi adapter that could interface with all sorts of retro things, not just SCSI.

With regards to this, how would you control the actual direction of data transfer? On the original raSCSI project, there was a method where you could directly attach a scsi cable to the pi, but the logic levels could fry it, this leads me to belive that we dont need control of A-B B-A but just hi and low level data transfer.

akuker commented 2 years ago

So i received the boards yesterday and populated them with the 74HCT245D transceivers, and to my amazement it worked. I tested with a Macintosh SE for about ~45 minutes, everything functioned fine.

Thank you for reporting back @DottiestDuke47 - Do the chips get very hot while you're using them?

DottiestDuke47 commented 2 years ago

They did get warm but I wouldn't say very hot.

thbreidenbach commented 2 years ago

Would the SN74ABT245B be a suitable alternative? At first glimpse it seems okay and better than the hct drivers

akuker commented 2 years ago

I spent some time working on this page tonight: https://github.com/akuker/RASCSI/wiki/Transceiver-Comparison

SN74ABT245B does look like a potential alternative!

I'm going to give this one a try: SN74F245DW

thbreidenbach commented 2 years ago

The abt drivers are among the fastest I know. I wonder what actually limits the speed of the rascsi implementation. But alas the abt drivers are available and since I have a few here I’ll give it a go and report back.

Thanks for the great work!!

Yours Thorsten

Am 12.02.2022 um 06:10 schrieb akuker @.***>:

I spent some time working on this page tonight: https://github.com/akuker/RASCSI/wiki/Transceiver-Comparison https://github.com/akuker/RASCSI/wiki/Transceiver-Comparison SN74ABT245B does look like a potential alternative!

I'm going to give this one a try: SN74F245DW https://www.ti.com/lit/ds/symlink/sn74f245.pdf? — Reply to this email directly, view it on GitHub https://github.com/akuker/RASCSI/issues/591#issuecomment-1036996990, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADEUE5VEASEXXUIV4HCHYODU2XTUPANCNFSM5LEEIPFA. You are receiving this because you commented.

akuker commented 2 years ago

I believe the main limitation with RaSCSI is the speed of the GPIO access. The non-determinism of the Linux kernel doesn't help either. In this way, something like BlueSCSI or SCSI2SD have an advantage. They don't have the overhead of an operating system.

But, at the same time, they don't have the BENEFIT of an operating system either. Without an OS, things like the Daynaport emulation and the web interface would be much more difficult to implement.

thbreidenbach commented 2 years ago

I tried the Baremetal implementation on a raspi 3A+ in a fullspec RaSCSI which did 1.3MB/sec on an Amiga 500 with Oktagon 508 with Oktapussy ROM and a reasonable fast Accelerator (35Mhz 68ec020).. I might swap in an @.*** to see if the Computer limits the speed.

The Zorro 2 Interface of the Amiga should offer something around 3.5MB/sec throughput. A Zorro 3 SCSI Card will be available soon since the A4091 Fast-SCSI 2 Controller is in the process of being reverse engineered. This Card should offer Speeds beyond 16MB/sec.

Am 12.02.2022 um 17:42 schrieb akuker @.***>:

I believe the main limitation with RaSCSI is the speed of the GPIO access. The non-determinism of the Linux kernel doesn't help either. In this way, something like BlueSCSI or SCSI2SD have an advantage. They don't have the overhead of an operating system.

But, at the same time, they don't have the BENEFIT of an operating system either. Without an OS, things like the Daynaport emulation and the web interface would be much more difficult to implement.

— Reply to this email directly, view it on GitHub https://github.com/akuker/RASCSI/issues/591#issuecomment-1037296894, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADEUE5RPPQ3U4A64HNNU5E3U22EX5ANCNFSM5LEEIPFA. You are receiving this because you commented.

WarlockD commented 2 years ago

Ugh I got to check my notifications more often, Its been a fun few months:P Anyway the NTS0104 seem to work fine using the pi, but I didn't have time to do any kind of board mock up, but the showed a nice edge on both the 5v and 3 volt size. That being said, it was a month or so ago and need to rebuild my experiment.

Honestly, the hard IRQ kill that RaSCSI does is probery the best idea. Its not clean by any means, but there isn't alot of ways you can turn off the scheduler in linux without replacing it completely. The best way would be to make this a kernel module. This way it can dirrectly fiddle with the kernel during transfers.

To be honest though, I kind of had to switch to using an ncr53c80. I am trying to emulate a old tape drive and I HAVE to handle reconnection correctly and as a plus it handles the REQ/ACK easy using the dma pins.

Honestly I kind of wish those old 5V tolernt FPGA's were still around:P

WarlockD commented 2 years ago

Oh found another. Seems like it supports auto direction detection, 8-bit port. NTS0308E. Push pull is 20mbps but open drain is only 2mbps. They aren't to bad at 1k@ US$0.54. Humm, might be fine, max properation delay is 20ns humm, I think the 2 mbps is only in regards to lower votage trasistions and might be fine at 3.3v. Ah well, I am going to buy a bunch a do some tests.

As for fast scsi-2, its possible, but may not be? These older controllers usually fudge quite a bit on sync and give way more leway than a more modern controller. If you kill the irq to the kernel so only your function is running, I think the delay is around 100ns so it MAY be possible as it only matters on the data transmission.

OH now THIS is intresting. NVT2010PW is a TSSOP-24, but its a 10bit auto detecting votage between 1-5V. goes up to 33mhz. Going to get a few of these as well.

akuker commented 2 years ago

NVT2010PW does look interesting....

akuker commented 2 years ago

More data!!!!

I did some testing with the SN74F245, SN74ALS245A-1 and SN74LS641-1 (as a baseline).

Device Max Read Max Write Cost each
SN74LS641-1 (baseline) 950 KB/sec 506KB/sec $3.66
SN74F245 1245 KB/sec 574 KB/sec $0.93
SN74ALS245A-1 1193 KB/sec 571 KB/sec $1.49

Here are the waveforms of the D0 data line during a read operation: SN74F245 Max Read: 1245 KB/sec Max Write: 574 KB/sec sn74f245_io_low SN74ALS245A-1 Max Read: 1193 KB/sec Max Write: 571 KB/sec sn74als245a-1_io_low SN74LS641-1 Max Read: 950 KB/sec Max Write: 506 KB/sec sn74ls641-1_io_low

Tests were done with a Raspberry Pi 4, 8GB RAM, running a 22.03 development build. RaSCSI version 2.4a and Digilent Analog Discovery 2 for the O-Scope. The Test PC was a Macintosh SE/30 using SCSI Director 4.0.

akuker commented 2 years ago

From this data, I'm really leaning towards the SN74F245. The cost can not be beat. Especially since its available from JLCPCB as an assembly option. I'm still going to do more testing, and I'm open to other opinions!

akuker commented 2 years ago

FYI: I ordered a FLIR camera so that I can take some temperature measurements between the SN74F245 and SN74LS641-1. This has been my one remaining concern with the SN74F245. I'll report back when I have more data!

uweseimet commented 2 years ago

Any news on this?

akuker commented 2 years ago

Yes, the new boards are sitting on my desk. I just need to finish testing them.

uweseimet commented 2 years ago

Good news!

thbreidenbach commented 2 years ago

I have ordered some older production date 74abt245 (100 I believe) Would you like to check those too? They’re quite readily available (at least from me..)

Thorsten

Von meinem iPhone gesendet

Am 01.04.2022 um 03:13 schrieb akuker @.***>:

 FYI: I ordered a FLIR camera so that I can take some temperature measurements between the SN74F245 and SN74LS641-1. This has been my one remaining concern with the SN74F245. I'll report back when I have more data!

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.

jameswood commented 1 year ago

FYI: I ordered a FLIR camera so that I can take some temperature measurements between the SN74F245 and SN74LS641-1. This has been my one remaining concern with the SN74F245. I'll report back when I have more data!

How did you get on with this? I'm using the SN74F245DW and they run hot! Did you find the same? (I may also have some other problem, of course)

@akuker is the recommendation still to use the SN74LS641-1DW? Thank you!

akuker commented 1 year ago

I'm closing this ticket, since I'm no longer able to spend time on it. The current recommendation is still to use the SN74LS641-1. The heat issues go away when supplying the SN74LS641-1 with a VCC of 3.3v instead of 5v. However, this is running the chip outside of its advertised limits. I haven't had any issues with it though.

So, in summary, SN74LS641-1DW is still the recommendation.