• Updated 2023-07-12: Hello, Guest! Welcome back, and be sure to check out this follow-up post about our outage a week or so ago.

Converting Acard AEC-7722 to AEC-7726

Dave L

Member
BTW, how many folks want one of these? In the original thread I jokingly suggested selling them for $125, but I'm thinking $60 plus shipping is probably reasonable.

Count me in. I really appreciate the work you are doing.
 

chelseayr

Well-known member
sorry to ask a silly question out of curiousity but I presume that it requires a master-set single drive on the ide side right? (like no dual drives setup or even a single slave-set one)

still thinking of being interested in buying one of targ's rework soon enough as anything but the atapi-only bridges on ebay are just too freaking much $ in my own opinion :-s
 

trag

Well-known member
sorry to ask a silly question out of curiousity but I presume that it requires a master-set single drive on the ide side right? (like no dual drives setup or even a single slave-set one)

still thinking of being interested in buying one of targ's rework soon enough as anything but the atapi-only bridges on ebay are just too freaking much $ in my own opinion :-s

Yes, the adapters support a single drive set to Master.

Reading the little bit of documentation available on the ARC-765 chip, there is a version which supports two devices, Master and Slave, but I've never seen that version in the wild.

Apparently, they have a single die (the little silicon wafer square) and depending on whether they package it in a 208 pin package or the 144 pin package, it supports Master & Slave or just Master. At least that's the huge stew I'm making out of the tiny potato of documentation I can find.

No progress recently. Still need to swap out test machines. I will get back to it soon. I'm actually hung up because each time I think about it, I think I really ought to set up a couple high-resolution monitors on the same desk to test Radeon 7000 cards with and I can't figure out how to make it all fit, and then I don't do anything. I ought to just set up the computer and get to work on the non-Radeon things.

Better can be the enemy of good enough.
 

trag

Well-known member
Got the Beige back up last night. I pulled the sound card and blew out the connector, removed the modem, and reinstalled the sound card. That seems to have jostled something back into working. Also, the XLR8 Mach Speed Control got through with a message about auto-detecting the cache speed.

I suspect that what may have happened is when I zapped PRAM, I killed the XLR8 cache settings and it was the thing hanging the computer up just before extensions loaded. It's the first one to load and if it didn't show it's icon until after it's auto-speed detection, that would have the result I was seeing.

I have a 500MHz NewerTech ZIF in there running at 533 (8X on 66 MHz) and the XLR8 extension wants to run the cache at 267, which seems to be a little too fast. It's rated for 250MHz.

Anyway, I have the cache manually set to 213 or something like that now, which seems stable.

Also installing anything in the 2nd PCI slot seems to be an issue. It seems to press against the Round IDE cables on the card in PCI slot 1. I have a vague memory that this was a problem long ago and I avoided it by installing a really low profile gigabit ethernet card in slot 2. To low to interfere with the cables from slot 1. This may have been an issue when I was trying to test 6880M cards.

Did a little more testing. Here's what I found:

Rev. 1.8 card with TSOP Flash is rock solid. Runs great. Completes Atto ExpressTools benchmark every time with no issues.

Rev. 1.8 card with PLCC socket is also solid. See note below.

One Rev. 1.5 board is PLCC socket is solid as above. See note below.

One Rev. 1.5 board that appears to work, but then fails the ExpressTools benchmark part way through. I spent a lot of time installing the PLCC chip just right and trying different Flash chips in the socket. Either the socket is broken or not quite soldered down properly. I'll need to rework this one.

One Rev. 1.8 still waiting to have it's TSOP flash soldered down after flashing.

Note: I took the bottom out of the PLCC socket so I could solder the pins. If one presses the Flash chip all the way into the PLCC socket, it doesn't make good contact. The flash chip needs to be just flush with the top of the socket. Unfortunately, I was overly efficient at house keeping and through away the socket bottoms that I clipped out. Otherwise I could tack them down in the bottom of the socket after the pins are soldered.

So the issue with boards on which I've installed a PLCC socket is don't press the chip in too far. In the future I won't have this problem. Right now I just have the three examples.

I ordered some more 7722s but wanted to finish my learning experiences before starting to convert the additional units. Things like, don't throw away that bottom plate from the PLCC socket.... They sent me 100% rev. 1.5. At this point I'm kind of liking the Rev. 1.8 more. I wish they had sent me those. I was worried about those tiny pins on the TSOP chip, but they're actually easier to solder/desolder than the PLCC stuff.
 

trag

Well-known member
Finally got around to the Benchmarks @Unknown_K asked for a while back. Actually did them Sunday, but it turns out the forum software doesn't like PICT files, so I had to convert them.

This is a converted 7722 connected to a 400GB Seagate PATA drive and an Adaptec 29160 SCSI card. All tests were run on a Beige G3 with a NewerTech G3/500/1MB running at 533MHz.

ST3400832_AEC7226_29160.jpg

Here's the same hard drive connected directly to an ATA-133 card (AEC-6880M).

ST3400832_AEC6880M.jpg

Same arrangement with a Seagate 700 GB PATA drive. 7726 and then 6880M

ST3750640A_AEC7726_29160.jpg



ST3750640A_AEC6880M.jpg

I wish I had a really fast Ultra 160 drive so we could test the bottlenecking of the 29160 card. I have a few 10K drives in the attic, but hte attic ladder isn't safe right now. So I just ran a benchmark with this old U2W (80MB/s Max interface) Western Digital 18GB drive:

WDE18300_29160.jpg

Based on the comparison, the READ speeds with the 7726 are not being bottlenecked by the hard drive. But they might be bottlenecked by the SCSI card. Without a faster SCSI drive to test with , I just can't tell.

Also, not sure and too lazy at the moment to check the Acard documentation, but does the 7726 support U2W, U160, U320? What's the maximum. These results are acting like U2W (80MB/S) is the maximum protocol being achieved.


BTW, I ran a quick test on one of the converted cards and it appears that once the 7722 is converted to 7726 with the 1.73Q firmware, it no longer supports optical drives. So, if you think you might want some SCSI-IDE adapters for optical drives, get the 7722s on Ebay while they're cheap.
 

trag

Well-known member
I was away from this for about two weeks. Sometimes work (my job) gets busy.

For anyone who is eager for me to send them one, I apologize for the delay.

After my test results two posts ago I just wasn't satisfied with the reliability I'm seeing. Also, while taking benchmarks (previous post) I noticed that sometimes (or some cards) would not always complete the benchmark.

So without a play by play, I did some more testing and soldering tonight. I have started using the Benchmark Utility in the ExpressPro Tools software as a reliability test.

I removed all the sockets I was using and soldered down the flash chips, thinking that was the/an issue, but it didn't really change anything.

Only two out of the five units I have converted reliably pass that reliability test.

Two of the others mount hard drives just fine, and look good -- until you try to run that benchmark, at which point the benchmark freezes during the Read (one card) or the write (the other card).

One card is just completely refusing to boot any more and locks up the computer during the initial SCSI interrogation. I'm beginning to wonder if I installed the Flash chip into its socket backwards at some point and damaged something on the board.

So the two that won't finish that benchmark -- They'll finish the less intense version of it just fine. It's when I set all the parameters to the maximums that they have trouble. And they get quite warm while executing the long benchmark. They're cool before the benchmark starts, so they only generate a bit of heat when they're working hard.

Anyone else notice that all the 7726s you see have a heat sink glued to the main chip?

The TL: DR is I don't want to send these out and have them not work reliably. So again, I apologize for the delay for additional testing, but I think it's important to solve these reliability issues before declaring the project a success.

Of course, if you're only even using the thing on a slowish narrow bus, I bet the heat would never become an issue.

Oh, one final question. Anyone want to recommend a good, affordable adhesive heat sink? These chips are 3/4" (~19 mm) square.
 

feipoa

Member
I have also been working on this. You probably came across some of my old posts on eevblog while researching this. I have recently revisted this work and today updated my eevblog post to reflect recent findings. Because some people don't like cross-forum double posting, I will only paste a link to my latest summary, https://www.eevblog.com/forum/micro...ry-one-byte-(all-file)/msg3840641/#msg3840641 And provide a brief conclusion,

CD-ROMs on AEC-7722IR v1.8 PCB work with v1.77B BIOS and a PLCC32 socket adaption. But the PLCC should be rotated 180 degrees compared to v1.5 and 3.0 PCB revisions. CD-ROMs work fine on 8-bit (usually 50-pin) and 16-bit (usually 68-pin) SCSI buses. Hard drives and CF cards work on the AEC-7722IR with 7726 BIOS versions 1.69H and 1.73Q, but only on 16-bit SCSI buses. This is in contrast to the AEC-7720U, which uses a different ACARD chipset that can accomodate IDE hard drives and CF cards on 8-bit 50-pin SCSI buses. Most CF cards I tried work, but again, you need to be using 16-bit, 68-pin cables and host controllers. Most CD-ROMs work, but not all. The LG CD-ROM on SATA-IDE-ACARD adapter I tried worked, but slower than a native IDE CD-ROM.

The ACARD 765E chip should work up to Ultra160 speeds, according to the manual. Some benchmarks I did with a PCI Ultra160 host controller and an Ultra320 HDD vs. Maxtor IDE on ACARD are 78 MB/s and 65 MB/s, respectively. While one can use IDE HDDs on the 7722 w/7726 BIOS and on 8-bit 50-pin buses, the benchmarks are VERY slow - 275 Kbytes/sec - and generally not usable.

Have you had any luck using the 7722 w1.73Q or 1.69H BIOS on an 8-bit, 50-pin bus? If so, how did you accomplish this so I can replicate your work?
 

trag

Well-known member
The ACARD 765E chip should work up to Ultra160 speeds, according to the manual. Some benchmarks I did with a PCI Ultra160 host controller and an Ultra320 HDD vs. Maxtor IDE on ACARD are 78 MB/s and 65 MB/s, respectively. While one can use IDE HDDs on the 7722 w/7726 BIOS and on 8-bit 50-pin buses, the benchmarks are VERY slow - 275 Kbytes/sec - and generally not usable.

Have you had any luck using the 7722 w1.73Q or 1.69H BIOS on an 8-bit, 50-pin bus? If so, how did you accomplish this so I can replicate your work?

I have not tried adapting the 7722 or the 7726 down to 50-pin, 8 bit SCSI. I will try it now that I have seen your results. That is concerning.

I have some really good 50 - 68 pin SCSI adapters with high-byte termination, and I can't find them, darn it. Might have to get @max1zzz to do another run of his -- although I think those were SCA to 50 pin. I think the ones I can't find were the last ones OWC had for sale.

Anyway, one thought. Did you install the 'SE' jumper on the 7726 when testing with 50-pin, 8 bit cable?

The SE jumper is for Single-ended operation, as opposed to LVD (low voltage differential) operation. Unless you have a U2W SCSI bus that is only 8 bits wide, any 8 bit SCSI bus you have (UW, SCSI2 or earlier) is going to be Single Ended -- that is, not differential.


Regarding EEVBLOG posting: I haven't read your update, but I did read the older material. I didn't know if you were still looking at it (posts are 2 years old?) so I didn't reply.

The Rev. 1.8 with the TSOP Flash chip uses a 16 bit wide flash, but only the lower 8 bits of every 16 bit "word" is used. That is why the firmware on the Flash chip has alternating 'FF' bytes. 'FF' is blank on flash chips.

If one has a regular Firmware file upgrade for the 7722 or 7726 and wants to update the Rev. 1.8 board's TSOP flash chip, one must insert blank bytes in the code to make it work. But that is only required with the 16 bit wide TSOP flash.

When you're programming an 8-bit PLCC 32 flash you can just load the firmware without modification.
 

max1zzz

Well-known member
I have some really good 50 - 68 pin SCSI adapters with high-byte termination, and I can't find them, darn it. Might have to get @max1zzz to do another run of his -- although I think those were SCA to 50 pin. I think the ones I can't find were the last ones OWC had for sale.
Yep my adapters are for the 80pin SCA type drives, however the design could be fairly easily adapted to 69 pin drives if there was enough damand for a 68pin adapter with high byte termination
 

Unknown_K

Well-known member
I remember getting a good 50/68 pin SCA adapter from compgeeks a long time ago and they sold little terminators for the adapter so LVD drives could be terminated on the 50 pin bus. No idea what machine it is in.
 

feipoa

Member
trag, yes I tried the SE jumper. I noted this in my last eevblog post. I decided to update the eevblog post because I have been getting private messages about that thread. I suggest reading that final eevblog post, otherwise I just as well copy/paste the whole eevblog post here. Yup, I'm aware of the 16-bit code issue with the TSOP-48 and dummy 'FF' byte entries. In my opinion, it best just to remove the TSOPs and use PLCC32 EEPROMs in sockets.

Now, something contrary to my findings, is that another member (from the Vogons forum) is also working on this, and he was able to get normal 50-pin, 8-bit SCSI speeds using HDDs on his setup. Apparently, he was using some scanner SCSI card, while I am using mainstream Adaptec cards. It suggests there may be a strong dependence on host controllers. I'll need to investigate this further. I have all 3 PCB versions of the AEC-7722's, but still need to do the PLCC socket adaption on v1.5 and v3.0.
 

trag

Well-known member
I remember getting a good 50/68 pin SCA adapter from compgeeks a long time ago and they sold little terminators for the adapter so LVD drives could be terminated on the 50 pin bus. No idea what machine it is in.

Are those the folks that changed their name to just geeks.com later. And then closed up about five or ten years ago? I got lots of great stuff from them back in the day. Sigh.

Yep my adapters are for the 80pin SCA type drives, however the design could be fairly easily adapted to 69 pin drives if there was enough damand for a 68pin adapter with high byte termination

One wonderful virtue of your adapters is that one knows they're terminated. I hate shopping for SCSI adapters because, who can tell?

I guess one should assume they contain no termination, unless they specifically mention it. But I also remember seeing some SCA-50 pin adapters that claimed they were terminated and had no circuitry on them at all.
 

weedeewee

Member
trag,

I agree about the high byte termination being needed to get the 7722 working on a narrow scsi bus.
Most cheapo adapters do not have this, neither does the single adapter I have.

Every time I tried with a 50p male to 68p male adapter, the scsi adapter only gave timeouts, indicating, to me, a bus issue.
only time I didn't get a timeout was with a buslogic adapter, then a device was detected, yet the name presented was just garbled data.

.

WRT the 16 bit tsop firmware issue. Just make yourself a file where every byte is 0xFF, then use romwalk to combine the two files together. the same way a Hi/Lo bios dump would be.

.

Hello feipoa !

.

the adapters I've used and functioned ...
LSI 53C1030 U320 scsi controller
adaptec 2940UW, 2940AU, AIC-7856T
buslogic flashpoint LT BA81c15
HP scanjet 5p (53c416) sym20403


adapter that didn't work for me, some HP netraid controller, got stuck on domain validation. might have worked in SE mode, can't recall.

To get the narrow bus working, on the 2940AU&UW, I used a 50p female to 68p female adapter, connected to a 68p cable terminated on both ends with the 7722 connected to one of the other connecters. That setup seems to work ok.

on the sym20403 card, I used a similar setup, though just added a few extension cables since the scsi card only has an external 50p scsi connector.

Tested with an old WD2000JB drive, which works fine
a JM2033J based msata to IDE convertor, which also works fine
and an almost directly connected CF card adapter with a sandisk ultra II 4GB cf card, which didn't work.

the 7722, in open air, reaches a temperature of about 52°C when just doing a linear read of a whole disk.
average operating temperature seems to be around 40°C

transfer speeds for
the ISA card were over 2MB/s, tested copying a 1G file to an SSD ,
the 2940AU around 13MB/s, tested using checkit in dos ,
the 2940UW around 18MB/s, tested using checkit in dos ,
the 53c1030 with checkit in dos around 31MB/s, linear read in linux, dd iflag=direct... , around 96MB/s .

The ISA, 2940AU&UW tests were done on an abit be6-II board with a p2-350 cpu,
the U320 test on a tyan S2895, can't recall the exact cpu spec.

.

Anyway, if you have a good link to purchase some of those 50p male to 68p male with high byte termination adapters which aren't priced over the top, please let me know.

Cheers !

edit: all the 7722 boards I have are rev.3 and the firmware I used is the latest available for the 7726Q
 
Last edited:

feipoa

Member
hello weedeewee!

I like your simple idea for making those 16-bit TSOP firmwares. Fortunately, I've gone all 8-bit though.

I tested my v. 1.5, 1.8, and 3.0 AEC-7722 units on a different computer today. Attached are photos of the 7722's with PLCC32 sockets installed. I used a variety of different host controllers: AHA-2930CU, 2940UW, 2940U2W, and ASC-39320A. I will present my results to you, then ask some questions. I am using an Asus TUSL2-C motherboard with VIA Nehemiah 1.2 GHz CPU.


I can confirm that v1.5 of the PCB doesn't like EEPROMs other than W29EE512.

AHA-2930CU
50-pin connector - using 50pin female to 68pin female adapter w/68-pin LVD/SE terminated cable.
No issue. Maxtor IDE HDD booted fine.
Buffered linear benchmark - 16400 KByte/s
Photo of 2930CU and 50-68 female adapter included.

AHA-2940UW
68-pin ultra connector: IDE HDD booted fine. Bench = 29125 KByte/s
50-pin connector - using 50pin female to 68pin female adapter - IDE HDD is detected, but won't boot. Why?

AHA-2940U2W
68-pin LVD connector: IDE HDD booted fine. Bench = 46065 KByte/s
68-pin ultra connector: IDE HDD booted fine. Identified as Fast/Ultra SE. Bench = 29127 KByte/s
50-pin connector - using 50pin female to 68pin female adapter - IDE HDD is detected, but won't boot. Why? The 68-pin and 50-pin connector on this card look to be wired together on the PCB.

ASC-39320A
68-pin LVD connector: IDE HDD booted fine. Bench = 34688 KByte/s. Notice how it is slower than the 2940U2W? I ensured the Adaptec BIOS was set to 320 MB/s speeds. Hmm...

Unfortunately, the motherboard this time around doesn't have ISA, so I cannot confirm my previous findings with an AHA-1520B card as tested in the PR440FX.

weedeewee: You mention that you were able to get the narrow bus working on your 2940UW. Did you try booting from the connected HDD? If yes, and it worked for you, what is your host controllers termination set at? I left mine on auto. I tried with and without SE jumper on the ACARD. Do you have the SE jumper installed?

52 degrees celsus doesn't sound too hot. I had planned to add a 27mm heatsink with 25mm fan, but perhaps only the heatsink is really required. When you measured 52 C, was it with an Ultra160 or Ultra320 controller? And what was the transfer speed being used while you measured the temperature?
 

Attachments

  • 50-to-68_pin_adapter_for_ACARD.JPG
    50-to-68_pin_adapter_for_ACARD.JPG
    2 MB · Views: 25
  • ACARD_AEC-7722_rev_1.5_3.0_1.8.JPG
    ACARD_AEC-7722_rev_1.5_3.0_1.8.JPG
    2.6 MB · Views: 27

weedeewee

Member
weedeewee: You mention that you were able to get the narrow bus working on your 2940UW. Did you try booting from the connected HDD? If yes, and it worked for you, what is your host controllers termination set at? I left mine on auto. I tried with and without SE jumper on the ACARD. Do you have the SE jumper installed?

52 degrees celsus doesn't sound too hot. I had planned to add a 27mm heatsink with 25mm fan, but perhaps only the heatsink is really required. When you measured 52 C, was it with an Ultra160 or Ultra320 controller? And what was the transfer speed being used while you measured the temperature?
There's a clue the scsi bios gives when detecting the device whilst connected to the narrow bus on the 2940UW.
It mentions 40.0 as speed for the device connected to the 2940UW while only connected on a narrow bus.
This is wrong. Now why exactly this happens? I do not know.
The solution though, is simple.
. Disable wide negotiation for that device.
tested with WD2000JB drive & Sintech SD-IDE adapter with 32G µSD card. max speeds around 17MB/s

the temperature of 52°C, without case, was measured during a linear read of the disk using a U320 adapter and the speed during that read was around 80MB/s max of about 95MB/s.

no SE jumpers set on the 7722, and termination on scsi controllers is set to automatic.
 

feipoa

Member
Disable wide negotiation for that device.

Perfect, thanks!

I forgot to mention - for those who are going to modify their AEC-7722 v1.8 board for use with an 8-bit PLCC, you will need to rotate the PLCC 180 degrees compared to v1.5 and v3.0 of the PCB. This should be evident from the photo I included above!
 

trag

Well-known member
WRT the 16 bit tsop firmware issue. Just make yourself a file where every byte is 0xFF, then use romwalk to combine the two files together. the same way a Hi/Lo bios dump would be.

Welcome, weedeewee. Thank you for the suggestion and all the other interesting information. I actually just wrote a little PERL script some time ago to input the BIOS/Firmware file one byte at a time and output a new file interleaved with 'FF'. Worked great, but there are many ways to solve that challenge.

I like having the 16 bit compatible firmware file because I find the TSOP chips are actually a little easier to desolder/solder than the PLCC chips. Unfortunately, in the new batch of boards that I ordered, they sent me 100% rev. 1.5. Sigh.

I agree that chips other than the WE29EE512 don't seem to work.
 

Unknown_K

Well-known member
Are those the folks that changed their name to just geeks.com later. And then closed up about five or ten years ago? I got lots of great stuff from them back in the day. Sigh.
They used to give out bumper stickers with orders and for a time I recall getting jelly beans in the box of stuff I got from them. New items that were pretty much closeout stuff but cheap. Think I got my NIB PPro overdrives there when I setup my Intel Pr440fx Win2k server back in the day.

Long gone since nobody could really compete with Newegg and Amazon at the time.
 
Top