About the performance: I worked as a service tech in the days of ATA and I might have some insight. First thing - when measuring performance, always cut 10-15% off the theoretical maximum for overhead. ATA has notoriously bad signal degradation and lots of control overhead. Now, based on the size of the drive that is detected you are definitely running UDMA6 as the protocol (aka ATA133). The transition from UDMA5 to 6 extended the address from 28bit to 48bit thus introducing support for drives larger than 137GB. You can expect that bus to support about 115-120MB/a of throughput in the best case scenario. But the performance can also be degraded with ATA and the host controller and disk controller can negotiate timing separately of the protocol standard, and even though you are running UDMA6, timings could be throttled to UDMA5 or less. Think of it like your CPU throttling when it gets hot, but with UDMA the signal quality is what disctates the switcing speed. So based on your results I think the performance is looking like what what you would expect from a UDMA5 bus speed (ATA100) - around 85MB/a + or - a few MB. This is indicative of the host bus throttling back the switcing speed - UDMA6 switches at a minimum of 15ns (7.5ns on/7.5ns off) - but can throttle back to 20ns (UDMA5 minimum) or even 30ns (UDMA4, worst case scenario). Typical throttle back is one standard lower. This usually happens when 40pin cables are being used instead of 80pin, since the 80pin has another 40 ground wires to attenuate noise as much as possible. Other causes would be an UDMA5 device on the same cable, however since you are plugging the drive what seems to be directly on to the board it could also be just a poorly designed board, crappy host controller, or even a noisy ground on the board itself and this was never resolved in production because noone seriously believed that anyone would actually use UDMA6 or just didn't care of it worked or not, because 🍎.
My immediate thoughts at the end if the video (and never watched vids like this before), think you might know, if you fit nands and a controller that far exceeds the ATA standard, will that fix the very low results on some of the benchmark results? Or is there some other ATA limits or some other more software limits in play?
@@MrSparkefrostie One thing is that this has no RAM cache, as far as I could see. That would help immensely, as can be seen when comparing most commercial drives with and without RAM cache.
@@handlandj well, I don't think so. The nand chips he is using are 29F64B08NCMF2 - each of these chips has about 50-60MB/s of throughput so at the lower end 4 chips in parallel should do +200MB/s. (I am basing these numbers off of the Intel 730 that uses these same chips and that SSD can push +400-500MB/s.) That is more than enough to outperform any DRAM that would be standard on ATA drives back in the day, which was rated for about 133-150MB/s. Adding DRAM here could maybe increase performance at very small IO - like 1k or 4k block performance, but only if the controller knew how to build a queue with pre-fetch - which I doubt a reverse engineered silicon can do. DRAM will not increase the max MB/s throughput since that is purely up to the performance of the back-end chips not the cache. In itelsf this setup should be more than capable to meet or beat the throughput of the ATA bus and in my opinion the bottleneck is the ATA bus itself and the (poorly) negotiated speed between the IC on the disk and the IC on the host bus.
Stuff like this is amazing!! I wish there was something like this for SCSI. The SCSI2SD (and similar) are all great, but a pure simple SCSI SSD would be awesome for older computers.
Man, those boards cleaned up nicely! It was bothering me the whole video that you hadn't cleaned them, but then you used the ultrasonic cleaner! wow! 😅 Anyway, nice work man, that's some next level talent. ✌🏻
@@haydenc2742 oh I know! -- I just didn't know he had one, and was getting antsy, thinking he was going to leave it on there and never clean it, and I was so distressed. -- The moment he whipped out that ultrasonic cleaner, I was like, YUSSS!!! 👊🏻
That thing with the silicon motion programming software, that 3-5 digit password thing is in nearly every version I've used. I don't know if it was lifted off of a floor of a factory, or if somebody just gave the password to somebody, but I do a lot of work in USB flash drive data recovery, and a lot of my time is spent entering 3-5 digit passwords JUST TO CHANGE THE SETTING MENU. Small price for the power the tools have, but definitely a mild irritation.
As near to impossible as it is to find dependable running IDE drives this sounds like a very good solution. My brother is into older systems that utilize IDE and most of his stock of HDD's are failing him. I have some adapters that allow sata drives to work with IDE but I haven't tested them with HDD drives yet. I used a sata dvd drive with the adapter on an older computer and it recognized the drive with no issues.
Impressive, I admire your work from early on, used a lot of your software tools on "unsupported macs" :) Great hardware project, I was just thinking the same for some 1.8" drives, got idea to use in old iPod and nokia N91 besides a laptop/old ultrabook
Bravo my friend! I'm in for at least a few, so no worries there. Thanks for opening the project, as I am getting all kinds of nifty ideas PCB wise, so this should be fun. Most of all, it keeps even more eWaste out of landfills, as ~60-80MB/sec. will more than satisfy a casual online machine. Good stuff all around, good stuff indeed! Also, thanks for your work on the many, MANY iterations of the MacOS Patcher, as it has kept COUNTLESS machines from meeting an unneeded end. Honestly, SO MANY OF MY CLIENTS from a few years back JUST WANTED DARK MODE ON THEIR (UNSUPPORTED?) MACS. Apple nixing support for Mojave almost caused A LOT of eWaste, and A LOT of money wasted on machines they would've gotten no benefit from spending 2K on. 2011 & 2012 Macs were perfectly capable of running Mojave AND BEYOND (with VERY FEW compromises thanks to your hard work, and now others have joined this wonderful community). Just a big, sincere, very grateful thank you from me brother. You're a fu*kin' legend!
Of all the Apple criticism out there, I think this is by far the most valid. EOL hardware being not only unsupported, but blacklisted by future OS releases is just ridiculous. I use a 2012 Mini for a work computer, and it's _fine._ There's no good reason to throw out working hardware just because it's 10 years old.
dosdude you are amazing - I’ve learned so much from your videos! +1 to a 3.5” version - I haven’t had great luck with 2.5 adapters and I don’t see any 3.5” IDE SSD drives out there.
First of all, your work is just amazing. And about the soldering: first I found this "pre-tin then install SMD with hot air" technique a bit weird, since solder paste exists and it's not that expensive. It just seems more work than adding paste with a syringe (not talking about using masks, that costs extra), populating the board and just blasting with hot air. But I got curious when I saw the resistor arrays. My main choice is 0603 with 0402 if really necessary (haven't gone down to BGA yet), but the arrays are so compact and useful, yet annoying to solder without paste mask, I always get a blob or two shorting pins which obviously needs repair. But this technique probably solves the "too much solder on tiny pins" problem! I'll give it a try today.
The Amiga community would love it if you designed one to plug right into the male header on the A600 and A1200. The bus is slow, but latency can help performance. The key is to put the flash chip on the opposite side of the board from the controller to keep the profile low. 32GB is plenty. Also having a female connector on the drive would be nice to plug directly into the motherboard. I am guessing that you could sell quite a few of them if you got those key features engineered.
This is actually something I was thinking about doing, but there is no off-the-shelf NAND controller for SCSI. I'd have to fully implement it myself by way of an FPGA or such.
The cheap ones that are widely available in most places utilize the JMicron JM20330 IDE to SATA bridge IC, which tends to be pretty bad. It has compatibility issues with many machines, and on the machines it does work with, tends to be very slow (I've seen them max out at as low as 33 MB/s on some machines). In the future I plan to design my own IDE to SATA adapters as well, but utilizing a much better IDE to SATA bridge IC. Though for the maximum performance and compatibility, a true IDE SSD like this is the best option.
@@dosdude1 I would be very interested in that... I'd suggest the 88SA8040 and not the 88SA8052. Yes, you lose SATA300, but who cares, IDE is limited to UDMA6 anyways, which is 133MB/sec. And the 88SA8040 is easier to solder.
Nice, I was just going through the process of re-soldering of the damn U8900 chip on my 2012 MBP retina that has the dreaded video cut off problem caused by bad soldering of U8900. Half taken apart yesterday late, continuing today after I finish my usual 1l of coffee. Hopefully I don't kill it.
This is so cool to see ! This is the first of you video I see, so i dont know if you already did but it would be nice of you to go over the desing process of this board. It i very interresting to see someone making a SSD at home. it is crazy to that the parts count looks so low for something so complex. It make me when to try it ! Your channel look really interresting. Keep up the good work !
oh man, I've been searching for an IDE SSD for a while. and this actually seems like a better* idea compared to buying one! it's fun! ^^ thanks for making this
Wonderfull homehacking of hw :) I would suggest you get a "bare" stencil from pcbway next time though. Allows you to apply paste to entire board properly instead of tinning pads like you did. This also allows the BGA's to sit a bit taller on the board once installed (correctly) and may improve reliability as well. Not to mention less thermal stress to heat and reflow one side once.
You really shouldn't tin BGA pads before soldering a chipset with solder balls already on it (in most cases). If I could get the stencil without holes for the BGA pads or the pads for the pin headers, that would probably be worth it.
@@dosdude1 The thin layer of paste on the pads melts while chip is on the pads to reflow properly. Using only solderballs does not provide the correct amount of solder.
It seems that the 10 kohms resistor R16 in your design is the Toggle mode resistor. On most SMI SATA boards, this 10k resistor must be populated when using Toggle mode NAND (Toshiba, Samsung and most Sandisk), and should be removed if using ONFI (Intel, Micron) or Legacy mode NAND.
Interesting... Though I’m currently using Intel “ComputeNAND” (as shown). I’ll have to see if I can dig up a data sheet for it, and determine if it is indeed toggle mode. When I initially created and tested this design, I had some Toshiba NANDs I used that would not work properly... I may try those again with R16 removed and see if that makes a difference (I could not find a data sheet for those Toshiba NANDs).
Can you make hard disk with RAM memory chips which have unlimited readwrite resource (like gigabyte i-RAM device, but with modern sata interface)? I agree format disk after each power failure, for some tasks is ok. It is possible? May be you can use a cheap ddr2 chips for that?
I've thought about designing custom memory devices before, but I always figured it would be too difficult. Did you find it to be particularly hard to do?
This one was a bit more difficult, mainly because I couldn't get a datasheet for the SM2236 controller. As such, I had to figure out its pinout by reverse-engineering another board (from a Compact Flash card) that used this chipset. Routing the board was the next difficult part. My goal was to keep the board at 4 layers (two power planes and the top and bottom as signal layers), so I had to spend quite a lot of time doing the routing over and over again until I got it right. All in all, I put in about 3 weeks worth of working hours to complete the whole design.
Wonder if this will be more reliable in detection than IDE to Sata adapters. as i have a 486 that while it works with IDE to CF or IDE to SDCard adapters, the IDE to SATA adapter always fails to detect the drive properly.
It certainly will, though it may be necessary to put much smaller NANDs on it to make, say, a 1GB (or maybe even smaller) drive instead of 256GB to be more compatible with the BIOS of those older systems.
IDE to SATA adapters adhere to ATA2 standard and no longer support /IOCS16 pin/signal. "SATA HARD Disk in 286/386 Mobo: Is it possible?" thread on Vogons has answer and solution. TLDR: old IDE adapters think IDE2SATA is trying to send 8bit data.
@@jjjacer As I said solution is already posted in that thread, 2 chips decoding and asserting /IOCS16 when needed. Lets you use modern SSD is ancient 386/486.
I didn't expect that you can melt solder with stencil, though it'll just solder to it. I also expect that board thickness would have been closer matched to connector pin gap
Yeah, reballing BGAs with the stencil on is how it has to be done, otherwise you’ll get bridging. Unfortunately I couldn’t get an exact thickness match from the PCB fab... The next available thickness size would have been too thick.
Awesome design!!!!! I have a few original XBOX's that I modded and put 80gb spinning hd in them for the games...this would be INCREDIBLE for upgrading them to SSD! So epic!!!! My cousin gave me a powermac G4, I wonder if this would work with that
This is a brilliant project and i'm just now scratching the surface of this sort of thing. I can't help but wonder if/why they didn't include a solder paste stencil? I'd think with that much to solder it would have been more time saving to do it that way? Either way this is exciting! Just found your channel and insta-subbed! 👍
They didn't include a stencil because I didn't request one... I don't see any benefit to using solder paste when assembling by hand. Now if I were using a pick-and-place machine, then yes, I would definitely need the stencil and solder paste.
You could solder with just solder paste and hot air, without so much flux and a soldering iron. It is necessary to solder on a surface that does not remove heat well, like cardboard, wood, matte rough ceramic tiles. But in no case not on metal, the metal quickly takes heat from the board, then the board requires more heating and there may be areas that need to be heated much more.
Makes me think of a few things, and I tend to have a futuristic point of view. 1. From a perspective of someone who doesn't make his own chips, it seems like a lot of people could do it with the adequate toolkit. 2. If it is possible, then what are the limits to the type of components? Considering powerful LLMs emerging, programming is even more automated, etc., couldn't you hypothetically make wacky components like a 3.5" SSD with 8Tb of memory, etc. (I'd love to get some realistic considerations on these types of ideas and what type of limitations could arise in the process, generally curious what's theoretically possible)
I don't know much about electronics of this complexity, but would there be any mileage in creating sockets for the Nands, or even an adapter for a daughter board? The flexibility of changing the 'platters' out sounds handy! Of course, you'd have to ensure that your nands were identical ... Hmm, food for thought!
I mentioned this on Action Retro's video about the device, but I wonder if this would work for Playstation 2 consoles? Right now people will take a SATA SSD and use a converter to make it IDE. It might also work in an original XBOX, but that console locks/unlocks the drive using commands, so it might be iffy. Any plans to make a full-size IDE connector version so a converter would not be needed? Also, do trim commands work over IDE? If so, would they work with this?
I have no interest in obsolete retro tech especially macs but it's very interesting to see how the manufacture process looks like regarding what happens after the PCB is soldered.
IDE is too slow to make any use of it, which is why this controller (and all other IDE NAND controllers I’ve found) don’t have any support for DRAM cache.
Only performance improvement i could see would probably be seek times as there is no head seek, but as dosdude replied IDE is too slow to take full advantage of an SSD. What i see is this is probably a more reliable alternative to old IDE drives or SD card adapters.
@@jjjacer IDE is plenty fast enough to make use of a better storage medium . Especially if the 20 -60 gig hard drive you started with runs around 1-3 mb/sec . The interfaces had a max theoretical through put of ATA 100/133 mbs per second(probably shared amongst all I/O). There are new ssd's that still don't run that fast yet. Turning a seek time into a zero also saves time EVERY operation. It has increased stability long term , will be more responsive and at some level will be faster .
@@tobiwonkanogy2975 True although the IDE device id be throwing one into is in a 486 so probably plain IDE or EIDE 8mb-16mbps which even most later mechanicals i have can do But as all my mechanicals are getting on in age, ssd's look like a better choice for me even though the speed might not be much since they are able to be made new and im not dealing with used drives.
Great project! Serious question though - where do you source your IC components? Can't seem to find the controller or memory chips on any of the standard places (mouser/digi-key/ebay)
Capacity might be a little too much for a problem that I am still working on. I have this old RocTec RocHard-RH800c harddrive-ram sidecar for Amiga500. It has this tricky and strange ROM that only allow max 256 megabyte drives.
Would love to install this into my Mac Mini G4 1.5 and all my Powerbooks. I got these M.2 to 44pin enclosers and had a bunch of 128gb laying around from recycling depots. as for speed wise?
Great video indeed , I can see quite big potential in this ; I'm thinking how many old gen console enthusiasts waiting to put an ide ssd into an old Original xbox
There are 2 ways commercial systems do surface mount small parts, one is similar to your method called Pick n Place. However, I think you would be better off with a glue line style method where you glue all the parts down with an adhesive and then apply the solder. Admittedly they are generally applying the solder via dipping the board into a puddle of hot flowing solder with a part in a fixture to protect things that shouldn't get solder on them. I think you would just be better off gluing the parts down first in their proper places and then soldering them afterwards by hand. The industrial soldering specification is IPC-601. I learned under IPC-601-D, but I'm not sure of the current revision.
[Computer Programmer] As a data recovery engineer, I use SATA ATA/ATAPI 1TX3-TB5D-R1RC4P2P2 driver to create and install an OS to your computer because it requires 8GB of minimum space to install it directly to your PC SSDs have 5 chips and in total of 256GB of memory to install the MacOS Tiger and copy it to the newly assembled 2 and a half inch drive as a MacOS Tiger Post-Installer
If you ever get the itch for another project, pcmcia linear flash memory, (so I can install different os to my omnibook) ram upgrades for old devices, (I need one for my omnibook 425) pcmcia I always wondered if a combined ata storate and ems memory, could be made. So you could use it with an hp lx200 series or similarly spec'd old palmtops/laptops with limited memory, could be neet, maybe even have room for things like a uart/esp32 and a yamaha sound chip or something "extra extra" (lol not asking for much, I can dream can't I?!) it'd be neat to get some new design for something like the IBM5140 convertible, I wonderd if it were possible to perhaps stuff more memory into it it, and do some form of EMS mapping for it, (although for that machine it would likely be hard to do, since so much of the machine is inside of asics) I started a project quite a while ago attempting to use the internal modem pins to add a hard drive to it, but there's no way I can find to put any extra rom's into it, like xtide, easily. I know so many of each of us nerds desires are one off and my ideas, might not be of any interest to you, but If any of that excites you, or you have similar projects, that cross with my desires I'd be happy to sponsor some of the costs. You do some nice work friend.
What are the dimensions of the pcboard? I'm keen for an SSD for an MP3 player I have (IAudio X5), which has a Toshiba styled hard drive. I'm excited to see if this would fit :)
why not use a stencil for solder paste for your passives and such? You can just set the parts in the paste and hot air them. Pretty easy and clean. That's my preferred method since when I'm not doing small prototype things, I'm stenciling the whole board and running it through my pick&place so I'm just used to the process. If I don't feel like warming up my conveyor whole reflow oven for one or two boards I just hot air them.
As a data hoarder, I would love to be able to custom assemble SATA class or NVME class SSDs from bulk parts. It would be great to be able to revive hardware RAID by having recoverability built into the drive via Raid5-like structure, with replaceable NAND and ARM controllers, so you can just replace whichever chip is going bad and rebuild on the new chip entirely on an internal high speed bus.
Hey Dos, loved the video. Just a quick question, what is the benefit of doing it this way as opposed to buying a $15 IDE - MSata adapter, then running an MSata SSD?
question about voltage regulator most of 3 pin linear voltage regulators can be adjusted to some point by installing diode or resistor in the ground path of the regulator, why would you want to cascade 2 regulators when all you need to do is either put a jumper or a pasive component on the regulator path ro ground? the board could be much simpler and if you give 1,5v offset to ground to a 1,8v regulator it would magically become 3,3v regulator anyway
@@dosdude1 oh then that is the explanation :] i didn't realise you need both rails controller takes 3,3 and nand take 1,8v or is it like dual voltage cpu with core taking one voltage and io the other
GREAT video! Honestly as a mechanical engineer maybe I didn't understand even 30% of what you did, but I enjoyed the process itself! One thing caught my attention, the programming of the SSD itself, I have a problem with the Samsung 850 Evo 500GB SSD that is giving me headaches A friend of mine brought it to me so that I, as someone who "knows computers" can try to unlock it, the laptop he had is broken and ended up in the recycling bin and problem is that he doesn't remember setting any code to it. The symptoms are as follows: when I connect it, it won't turn on - it asks for a code, when I connect it via USB as an external device - it turns on, but Windows doesn't see it and won't initialize it. Is there a program to at least erase it all with that damn code, there is nothing important on it anyway except old games. I hope he can at least use it for a few more years. We tried through Samsung and we didn't get any useful information "you locked the device and it works as such" Does anyone know a way, we read the forums and nothing worked and I honestly thought I'd throw it away until I came across your video Greetings from Serbia
If you can find the specific MPTool for the controller it has, then yes, you can remove the lock. Be aware, though, that doing so will also delete all data on the drive.
@@dosdude1 First of all, thanks for the quick response, as for the rest name saving of the data - it is completely irrelevant, there was the old installation of windows, drivers and games, and maybe it would be better for him to wipe all data clean to get rid of old viruses. As for the rest, I'm in absolutely unknown waters here, I studied machines not software and electronics, I don't know what or where to look. Maybe you can help me, where should I look It is a 500GB Samsung 850 Evo P/N MZ7LN500 Model MZ-75E500 MSIP-REM-SEC-MZ-75E500 Do you need PSID: S/N or WWN numbers?
As a data recovery engineer I can tell you this:
We could use more guys like you!!
Nice work on the SM2236!
About the performance: I worked as a service tech in the days of ATA and I might have some insight. First thing - when measuring performance, always cut 10-15% off the theoretical maximum for overhead. ATA has notoriously bad signal degradation and lots of control overhead. Now, based on the size of the drive that is detected you are definitely running UDMA6 as the protocol (aka ATA133). The transition from UDMA5 to 6 extended the address from 28bit to 48bit thus introducing support for drives larger than 137GB. You can expect that bus to support about 115-120MB/a of throughput in the best case scenario. But the performance can also be degraded with ATA and the host controller and disk controller can negotiate timing separately of the protocol standard, and even though you are running UDMA6, timings could be throttled to UDMA5 or less. Think of it like your CPU throttling when it gets hot, but with UDMA the signal quality is what disctates the switcing speed. So based on your results I think the performance is looking like what what you would expect from a UDMA5 bus speed (ATA100) - around 85MB/a + or - a few MB. This is indicative of the host bus throttling back the switcing speed - UDMA6 switches at a minimum of 15ns (7.5ns on/7.5ns off) - but can throttle back to 20ns (UDMA5 minimum) or even 30ns (UDMA4, worst case scenario). Typical throttle back is one standard lower. This usually happens when 40pin cables are being used instead of 80pin, since the 80pin has another 40 ground wires to attenuate noise as much as possible. Other causes would be an UDMA5 device on the same cable, however since you are plugging the drive what seems to be directly on to the board it could also be just a poorly designed board, crappy host controller, or even a noisy ground on the board itself and this was never resolved in production because noone seriously believed that anyone would actually use UDMA6 or just didn't care of it worked or not, because 🍎.
thank god we moved away from pata
My immediate thoughts at the end if the video (and never watched vids like this before), think you might know, if you fit nands and a controller that far exceeds the ATA standard, will that fix the very low results on some of the benchmark results? Or is there some other ATA limits or some other more software limits in play?
@@MrSparkefrostie One thing is that this has no RAM cache, as far as I could see. That would help immensely, as can be seen when comparing most commercial drives with and without RAM cache.
@@handlandj got it, dramless SSD, now it makes sense yo me thank you
@@handlandj well, I don't think so. The nand chips he is using are 29F64B08NCMF2 - each of these chips has about 50-60MB/s of throughput so at the lower end 4 chips in parallel should do +200MB/s. (I am basing these numbers off of the Intel 730 that uses these same chips and that SSD can push +400-500MB/s.) That is more than enough to outperform any DRAM that would be standard on ATA drives back in the day, which was rated for about 133-150MB/s. Adding DRAM here could maybe increase performance at very small IO - like 1k or 4k block performance, but only if the controller knew how to build a queue with pre-fetch - which I doubt a reverse engineered silicon can do. DRAM will not increase the max MB/s throughput since that is purely up to the performance of the back-end chips not the cache. In itelsf this setup should be more than capable to meet or beat the throughput of the ATA bus and in my opinion the bottleneck is the ATA bus itself and the (poorly) negotiated speed between the IC on the disk and the IC on the host bus.
This guy is on another level of hobbyist. Awesome.
This is truly amazing. I'm so envious with people with electronic design skills! Well done and thank you for this impressive video.
Stuff like this is amazing!! I wish there was something like this for SCSI. The SCSI2SD (and similar) are all great, but a pure simple SCSI SSD would be awesome for older computers.
dude you're amazing, your skill level is evident throuout the video. congratulations on your project!!!
Man, those boards cleaned up nicely! It was bothering me the whole video that you hadn't cleaned them, but then you used the ultrasonic cleaner! wow! 😅
Anyway, nice work man, that's some next level talent. ✌🏻
Believe it or not...a large sonic cleaner can easily clean up entire motherboards in the same fashion
@@haydenc2742 oh I know! -- I just didn't know he had one, and was getting antsy, thinking he was going to leave it on there and never clean it, and I was so distressed. -- The moment he whipped out that ultrasonic cleaner, I was like, YUSSS!!! 👊🏻
That thing with the silicon motion programming software, that 3-5 digit password thing is in nearly every version I've used. I don't know if it was lifted off of a floor of a factory, or if somebody just gave the password to somebody, but I do a lot of work in USB flash drive data recovery, and a lot of my time is spent entering 3-5 digit passwords JUST TO CHANGE THE SETTING MENU.
Small price for the power the tools have, but definitely a mild irritation.
As near to impossible as it is to find dependable running IDE drives this sounds like a very good solution. My brother is into older systems that utilize IDE and most of his stock of HDD's are failing him. I have some adapters that allow sata drives to work with IDE but I haven't tested them with HDD drives yet. I used a sata dvd drive with the adapter on an older computer and it recognized the drive with no issues.
Amazing work! Proper job and a quality bit of kit! Hopefully they'll be available to buy in the UK!
This is AMAZING!
Impressive, I admire your work from early on, used a lot of your software tools on "unsupported macs" :) Great hardware project, I was just thinking the same for some 1.8" drives, got idea to use in old iPod and nokia N91 besides a laptop/old ultrabook
Yep, I already made a 1.8" ZIF version of this drive.
@@dosdude1did you made video about it?
Wow ! Very impressed with the re-balling work.
You are a very clever Man. Thank you for all of your hard work.
Colin you are amazing. Those drives should speed up those old G4 macs.
Wow so tedious yet so satisfying. You sir are a soldering rock star. Thank you!
Bravo my friend! I'm in for at least a few, so no worries there. Thanks for opening the project, as I am getting all kinds of nifty ideas PCB wise, so this should be fun. Most of all, it keeps even more eWaste out of landfills, as ~60-80MB/sec. will more than satisfy a casual online machine. Good stuff all around, good stuff indeed!
Also, thanks for your work on the many, MANY iterations of the MacOS Patcher, as it has kept COUNTLESS machines from meeting an unneeded end. Honestly, SO MANY OF MY CLIENTS from a few years back JUST WANTED DARK MODE ON THEIR (UNSUPPORTED?) MACS. Apple nixing support for Mojave almost caused A LOT of eWaste, and A LOT of money wasted on machines they would've gotten no benefit from spending 2K on. 2011 & 2012 Macs were perfectly capable of running Mojave AND BEYOND (with VERY FEW compromises thanks to your hard work, and now others have joined this wonderful community).
Just a big, sincere, very grateful thank you from me brother. You're a fu*kin' legend!
Of all the Apple criticism out there, I think this is by far the most valid. EOL hardware being not only unsupported, but blacklisted by future OS releases is just ridiculous. I use a 2012 Mini for a work computer, and it's _fine._ There's no good reason to throw out working hardware just because it's 10 years old.
Absolutely amazing - I just upgraded someone's AppleTV1,1 to use a 1 TB SATA SSD (with IDE adapter), this looks even better tho!
dosdude you are amazing - I’ve learned so much from your videos!
+1 to a 3.5” version - I haven’t had great luck with 2.5 adapters and I don’t see any 3.5” IDE SSD drives out there.
I love how he uses an old IDE drive as a bench to work on his SSD version.
I was thinking something similar, I think it's an IDE CD-ROM drive
First of all, your work is just amazing.
And about the soldering: first I found this "pre-tin then install SMD with hot air" technique a bit weird, since solder paste exists and it's not that expensive. It just seems more work than adding paste with a syringe (not talking about using masks, that costs extra), populating the board and just blasting with hot air.
But I got curious when I saw the resistor arrays. My main choice is 0603 with 0402 if really necessary (haven't gone down to BGA yet), but the arrays are so compact and useful, yet annoying to solder without paste mask, I always get a blob or two shorting pins which obviously needs repair. But this technique probably solves the "too much solder on tiny pins" problem! I'll give it a try today.
I totally forgot to answer, until a spambot commented here :D
It worked definitely better for the resistor array. Thanks @dosdude1 for the idea!
The Amiga community would love it if you designed one to plug right into the male header on the A600 and A1200. The bus is slow, but latency can help performance. The key is to put the flash chip on the opposite side of the board from the controller to keep the profile low. 32GB is plenty. Also having a female connector on the drive would be nice to plug directly into the motherboard. I am guessing that you could sell quite a few of them if you got those key features engineered.
It could be fun if you make a SCSI variation of SSD. For old servers and silicon graphics workstations.
Could probably base it off the BlueSCSI project and adapt it for SSD flash storage.
And as you can see, this is a really cool project.
That was remarkable to watch. Absolutely amazing!
Thanks for sharing this! :)
You invent compact flash second time. Congradulation!
Would you consider designing one for SCSI, or do you think the existing projects (BlueSCSI, etc) are enough?
This is actually something I was thinking about doing, but there is no off-the-shelf NAND controller for SCSI. I'd have to fully implement it myself by way of an FPGA or such.
Agreed, SCSI replacemet would be wonderful... for older RISC6000 systems.
"Simple as that. " Easy for you to say. :) But I found it fascinating to watch how this is done.
This is just amazing. Thank you for sharing all of this with the world!
You've answered questions that I didn't even know I needed to ask. Great video!
dude! thank you for the ZIF variation! the thing will save a lot of iPod's
Finally, a SSD for my TC1100!
This is an amazing project. Am I strange for wanting a 3.5" version?
I do too. I'd love to tell people I have a 2TB IDE SSD in my original Xbox. I'd get many confused looks.
@@ProjectswAlex I have a 80gb HDD in mine...this would be awesome to have a 256GB that is capable of maximum read/write
Interesting project! In the past, I've used mSATA to 2.5 in IDE boards, and they've seem to work fine. I'm curious how this compares?
I got one adapter running great in my os9 g4 mini
The cheap ones that are widely available in most places utilize the JMicron JM20330 IDE to SATA bridge IC, which tends to be pretty bad. It has compatibility issues with many machines, and on the machines it does work with, tends to be very slow (I've seen them max out at as low as 33 MB/s on some machines). In the future I plan to design my own IDE to SATA adapters as well, but utilizing a much better IDE to SATA bridge IC. Though for the maximum performance and compatibility, a true IDE SSD like this is the best option.
@dosdude1 I had a bad experience with that chipset, too. I ended up buying an adapter with the Marvell 88SA8052, and that worked much better.
@@AndrewMackoul Yeah, that's the chipset I intend to use when designing mine.
@@dosdude1 I would be very interested in that... I'd suggest the 88SA8040 and not the 88SA8052. Yes, you lose SATA300, but who cares, IDE is limited to UDMA6 anyways, which is 133MB/sec. And the 88SA8040 is easier to solder.
Nice, I was just going through the process of re-soldering of the damn U8900 chip on my 2012 MBP retina that has the dreaded video cut off problem caused by bad soldering of U8900. Half taken apart yesterday late, continuing today after I finish my usual 1l of coffee. Hopefully I don't kill it.
This is so cool to see ! This is the first of you video I see, so i dont know if you already did but it would be nice of you to go over the desing process of this board. It i very interresting to see someone making a SSD at home. it is crazy to that the parts count looks so low for something so complex. It make me when to try it ! Your channel look really interresting. Keep up the good work !
Dude, love your vids! Clearly one of the best channels out there! Thanks a lot!
oh man, I've been searching for an IDE SSD for a while. and this actually seems like a better* idea compared to buying one! it's fun! ^^ thanks for making this
How you will deal with ahci for trim?
This is really awesome! Love to see more.
Sincerely yours,
a fan of all of your work
For lining up the connector pins, a bit of Blu Tack is very handy stuff.
Wonderfull homehacking of hw :)
I would suggest you get a "bare" stencil from pcbway next time though. Allows you to apply paste to entire board properly instead of tinning pads like you did.
This also allows the BGA's to sit a bit taller on the board once installed (correctly) and may improve reliability as well. Not to mention less thermal stress to heat and reflow one side once.
You really shouldn't tin BGA pads before soldering a chipset with solder balls already on it (in most cases). If I could get the stencil without holes for the BGA pads or the pads for the pin headers, that would probably be worth it.
@@dosdude1 The thin layer of paste on the pads melts while chip is on the pads to reflow properly. Using only solderballs does not provide the correct amount of solder.
th-cam.com/video/dz7ltWBDm7U/w-d-xo.html
Microscope video of a very misaligned chip being helped by the pad solder paste.
This is awesome. It seems great to work in my older macs with IDE drives as well as my project 1st Gen Apple TVs.
It seems that the 10 kohms resistor R16 in your design is the Toggle mode resistor. On most SMI SATA boards, this 10k resistor must be populated when using Toggle mode NAND (Toshiba, Samsung and most Sandisk), and should be removed if using ONFI (Intel, Micron) or Legacy mode NAND.
Interesting... Though I’m currently using Intel “ComputeNAND” (as shown). I’ll have to see if I can dig up a data sheet for it, and determine if it is indeed toggle mode. When I initially created and tested this design, I had some Toshiba NANDs I used that would not work properly... I may try those again with R16 removed and see if that makes a difference (I could not find a data sheet for those Toshiba NANDs).
Can you make hard disk with RAM memory chips which have unlimited readwrite resource (like gigabyte i-RAM device, but with modern sata interface)? I agree format disk after each power failure, for some tasks is ok. It is possible? May be you can use a cheap ddr2 chips for that?
I've thought about designing custom memory devices before, but I always figured it would be too difficult. Did you find it to be particularly hard to do?
This one was a bit more difficult, mainly because I couldn't get a datasheet for the SM2236 controller. As such, I had to figure out its pinout by reverse-engineering another board (from a Compact Flash card) that used this chipset. Routing the board was the next difficult part. My goal was to keep the board at 4 layers (two power planes and the top and bottom as signal layers), so I had to spend quite a lot of time doing the routing over and over again until I got it right. All in all, I put in about 3 weeks worth of working hours to complete the whole design.
pretty sick, but not easy to do due to the BGA packages. i think for simplicitiy most will stick to passive CF to IDE adapters or similar
Wonder if this will be more reliable in detection than IDE to Sata adapters. as i have a 486 that while it works with IDE to CF or IDE to SDCard adapters, the IDE to SATA adapter always fails to detect the drive properly.
It certainly will, though it may be necessary to put much smaller NANDs on it to make, say, a 1GB (or maybe even smaller) drive instead of 256GB to be more compatible with the BIOS of those older systems.
@@dosdude1 True, im running a XT-IDE BIOS which let me run a 40gb drive in this system, but i know that was a pain for some of my other past projects.
IDE to SATA adapters adhere to ATA2 standard and no longer support /IOCS16 pin/signal. "SATA HARD Disk in 286/386 Mobo: Is it possible?" thread on Vogons has answer and solution.
TLDR: old IDE adapters think IDE2SATA is trying to send 8bit data.
@@rasz Interesting thread, looks like they might have a working solution eventually, will have to watch this thread.
@@jjjacer As I said solution is already posted in that thread, 2 chips decoding and asserting /IOCS16 when needed. Lets you use modern SSD is ancient 386/486.
I didn't expect that you can melt solder with stencil, though it'll just solder to it.
I also expect that board thickness would have been closer matched to connector pin gap
Yeah, reballing BGAs with the stencil on is how it has to be done, otherwise you’ll get bridging. Unfortunately I couldn’t get an exact thickness match from the PCB fab... The next available thickness size would have been too thick.
Awesome design!!!!!
I have a few original XBOX's that I modded and put 80gb spinning hd in them for the games...this would be INCREDIBLE for upgrading them to SSD!
So epic!!!!
My cousin gave me a powermac G4, I wonder if this would work with that
I also would want a few! I have a couple of old macs with spinning rust in 'em who can benefit from this drive! Great job man really!!
Sir can you please make a video about how solder smd components like you do ?
Great work 👌
Thank you
Greetings from north Africa (Algeria)
This is a brilliant project and i'm just now scratching the surface of this sort of thing. I can't help but wonder if/why they didn't include a solder paste stencil? I'd think with that much to solder it would have been more time saving to do it that way? Either way this is exciting! Just found your channel and insta-subbed! 👍
They didn't include a stencil because I didn't request one... I don't see any benefit to using solder paste when assembling by hand. Now if I were using a pick-and-place machine, then yes, I would definitely need the stencil and solder paste.
@@dosdude1 Ah I see. Thanks for reply!
Would you consider doing a project for a full size custom keyboard?
You could solder with just solder paste and hot air, without so much flux and a soldering iron. It is necessary to solder on a surface that does not remove heat well, like cardboard, wood, matte rough ceramic tiles. But in no case not on metal, the metal quickly takes heat from the board, then the board requires more heating and there may be areas that need to be heated much more.
Makes me think of a few things, and I tend to have a futuristic point of view.
1. From a perspective of someone who doesn't make his own chips, it seems like a lot of people could do it with the adequate toolkit.
2. If it is possible, then what are the limits to the type of components? Considering powerful LLMs emerging, programming is even more automated, etc., couldn't you hypothetically make wacky components like a 3.5" SSD with 8Tb of memory, etc.
(I'd love to get some realistic considerations on these types of ideas and what type of limitations could arise in the process, generally curious what's theoretically possible)
Very nice, I need a few 2.5" PATA drives for some ancient test gear, this looks like a very useful project.
I don't know much about electronics of this complexity, but would there be any mileage in creating sockets for the Nands, or even an adapter for a daughter board? The flexibility of changing the 'platters' out sounds handy! Of course, you'd have to ensure that your nands were identical ... Hmm, food for thought!
Oh gods, an almost perfect instruction on how to make an SSD drive to replace the HHD in an Audi MMI 🙂
How big can we go with storage capacity? 512gb? 1tb? I assume it's up to the controller chip
I mentioned this on Action Retro's video about the device, but I wonder if this would work for Playstation 2 consoles? Right now people will take a SATA SSD and use a converter to make it IDE. It might also work in an original XBOX, but that console locks/unlocks the drive using commands, so it might be iffy.
Any plans to make a full-size IDE connector version so a converter would not be needed?
Also, do trim commands work over IDE? If so, would they work with this?
@dosdude1 supports the TRIM command?
yes yes, I need this to all my retro laptops! I will miss the hum, but most of my ide drives died already
Yeah, that is the biggest issue...the drives being so old and going bad, this will definitely fix that! (and add TONS of space to boot)
woold it betta 2 install as first the bga's
n then passiv komponents?
I have no interest in obsolete retro tech especially macs but it's very interesting to see how the manufacture process looks like regarding what happens after the PCB is soldered.
What's the advantage of this over industrial CF cards such as the Transcend CF170 or CF220I series?
Why wouldn't you put informations about select resistors on silkscreen?
would there be any performance advatage over IDE by adding some DRAM cache to the design?
IDE is too slow to make any use of it, which is why this controller (and all other IDE NAND controllers I’ve found) don’t have any support for DRAM cache.
Only performance improvement i could see would probably be seek times as there is no head seek, but as dosdude replied IDE is too slow to take full advantage of an SSD.
What i see is this is probably a more reliable alternative to old IDE drives or SD card adapters.
@@dosdude1 ah gotcha
@@jjjacer IDE is plenty fast enough to make use of a better storage medium . Especially if the 20 -60 gig hard drive you started with runs around 1-3 mb/sec . The interfaces had a max theoretical through put of ATA 100/133 mbs per second(probably shared amongst all I/O). There are new ssd's that still don't run that fast yet. Turning a seek time into a zero also saves time EVERY operation. It has increased stability long term , will be more responsive and at some level will be faster .
@@tobiwonkanogy2975 True although the IDE device id be throwing one into is in a 486 so probably plain IDE or EIDE 8mb-16mbps which even most later mechanicals i have can do
But as all my mechanicals are getting on in age, ssd's look like a better choice for me even though the speed might not be much since they are able to be made new and im not dealing with used drives.
Thanks for your great work!!! I will certainly use this nice SSD
Great project! Serious question though - where do you source your IC components? Can't seem to find the controller or memory chips on any of the standard places (mouser/digi-key/ebay)
Capacity might be a little too much for a problem that I am still working on. I have this old RocTec RocHard-RH800c harddrive-ram sidecar for Amiga500. It has this tricky and strange ROM that only allow max 256 megabyte drives.
Could the controller ic from modern compact flash carts allow for larger capacity?
Would love to install this into my Mac Mini G4 1.5 and all my Powerbooks. I got these M.2 to 44pin enclosers and had a bunch of 128gb laying around from recycling depots. as for speed wise?
Hello! Sorry for my English, thank you so much for mptool, my CF card is SM2234, I want to increase the capacity, but I can't find mptool for it
Great video indeed , I can see quite big potential in this ; I'm thinking how many old gen console enthusiasts waiting to put an ide ssd into an old Original xbox
original xbox works fine with ide to sata adapter
where did you got SM2236?
Does the controller you use do automatic TRIMming? I suspect Apple nor Windows will let you TRIM an IDE drive.
There are 2 ways commercial systems do surface mount small parts, one is similar to your method called Pick n Place. However, I think you would be better off with a glue line style method where you glue all the parts down with an adhesive and then apply the solder. Admittedly they are generally applying the solder via dipping the board into a puddle of hot flowing solder with a part in a fixture to protect things that shouldn't get solder on them. I think you would just be better off gluing the parts down first in their proper places and then soldering them afterwards by hand.
The industrial soldering specification is IPC-601. I learned under IPC-601-D, but I'm not sure of the current revision.
where did you buy this nan cheps??
[Computer Programmer]
As a data recovery engineer, I use SATA ATA/ATAPI 1TX3-TB5D-R1RC4P2P2 driver to create and install an OS to your computer because it requires 8GB of minimum space to install it directly to your PC
SSDs have 5 chips and in total of 256GB of memory to install the MacOS Tiger and copy it to the newly assembled 2 and a half inch drive as a MacOS Tiger Post-Installer
Wonderfull work right there, where can these be purchased ?
If you ever get the itch for another project,
pcmcia linear flash memory, (so I can install different os to my omnibook)
ram upgrades for old devices, (I need one for my omnibook 425)
pcmcia I always wondered if a combined ata storate and ems memory, could be made. So you could use it with an hp lx200 series or similarly spec'd old palmtops/laptops with limited memory, could be neet, maybe even have room for things like a uart/esp32 and a yamaha sound chip or something "extra extra" (lol not asking for much, I can dream can't I?!)
it'd be neat to get some new design for something like the IBM5140 convertible, I wonderd if it were possible to perhaps stuff more memory into it it, and do some form of EMS mapping for it, (although for that machine it would likely be hard to do, since so much of the machine is inside of asics) I started a project quite a while ago attempting to use the internal modem pins to add a hard drive to it, but there's no way I can find to put any extra rom's into it, like xtide, easily.
I know so many of each of us nerds desires are one off and my ideas, might not be of any interest to you, but If any of that excites you, or you have similar projects, that cross with my desires I'd be happy to sponsor some of the costs. You do some nice work friend.
What are the dimensions of the pcboard? I'm keen for an SSD for an MP3 player I have (IAudio X5), which has a Toshiba styled hard drive. I'm excited to see if this would fit :)
We had ATA 133 but most drives only did 40MB/s back then
@dosdude1 Is it possible to do the same thing with memory chips recovered from phones/computers?
why not use a stencil for solder paste for your passives and such? You can just set the parts in the paste and hot air them. Pretty easy and clean. That's my preferred method since when I'm not doing small prototype things, I'm stenciling the whole board and running it through my pick&place so I'm just used to the process. If I don't feel like warming up my conveyor whole reflow oven for one or two boards I just hot air them.
On todays episode of "Geeks with too much time on their hands".... lol. Awesome to watch, and I kinda want an IDE SSD now. lol.
This is the next level of ssd
My friend algorith did his majic and i landed here. Internet moment xD
Awesome vid, awesome job and hope you can get something out of this great idea.
As a data hoarder, I would love to be able to custom assemble SATA class or NVME class SSDs from bulk parts. It would be great to be able to revive hardware RAID by having recoverability built into the drive via Raid5-like structure, with replaceable NAND and ARM controllers, so you can just replace whichever chip is going bad and rebuild on the new chip entirely on an internal high speed bus.
Beautiful work
nice!
What are the ideas for using g4 macmini these days? is it at all suitable for anything? may be some lightweight linux as a logging device?
Hey Dos, loved the video. Just a quick question, what is the benefit of doing it this way as opposed to buying a $15 IDE - MSata adapter, then running an MSata SSD?
Compatibility, reliability, and speed, compared to common JM20330-based adapters. See my previous comment about this for details.
this is a great idea, but wouldnt it be easier to just use an existing Disk On Module design?
Excelent work! Thanks for sharing!
Curious, would like to see if this works on old PC's like Intel Pentium or AMD 64, DOS of course?
question about voltage regulator
most of 3 pin linear voltage regulators can be adjusted to some point by installing diode or resistor in the ground path of the regulator, why would you want to cascade 2 regulators when all you need to do is either put a jumper or a pasive component on the regulator path ro ground? the board could be much simpler and if you give 1,5v offset to ground to a 1,8v regulator it would magically become 3,3v regulator anyway
I'm not sure that would work in this case, as I need both a 3.3V and 1.8V rail.
@@dosdude1 oh then that is the explanation :] i didn't realise you need both rails controller takes 3,3 and nand take 1,8v or is it like dual voltage cpu with core taking one voltage and io the other
An Eletronic God, just like that
If you want to improve the drive's stablility, try to play with RDT settings.
GREAT video!
Honestly as a mechanical engineer maybe I didn't understand even 30% of what you did, but I enjoyed the process itself!
One thing caught my attention, the programming of the SSD itself, I have a problem with the Samsung 850 Evo 500GB SSD that is giving me headaches
A friend of mine brought it to me so that I, as someone who "knows computers" can try to unlock it, the laptop he had is broken and ended up in the recycling bin and problem is that he doesn't remember setting any code to it.
The symptoms are as follows: when I connect it, it won't turn on - it asks for a code, when I connect it via USB as an external device - it turns on, but Windows doesn't see it and won't initialize it.
Is there a program to at least erase it all with that damn code, there is nothing important on it anyway except old games. I hope he can at least use it for a few more years.
We tried through Samsung and we didn't get any useful information "you locked the device and it works as such"
Does anyone know a way, we read the forums and nothing worked and I honestly thought I'd throw it away until I came across your video
Greetings from Serbia
If you can find the specific MPTool for the controller it has, then yes, you can remove the lock. Be aware, though, that doing so will also delete all data on the drive.
@@dosdude1 First of all, thanks for the quick response, as for the rest name saving of the data - it is completely irrelevant, there was the old installation of windows, drivers and games, and maybe it would be better for him to wipe all data clean to get rid of old viruses.
As for the rest, I'm in absolutely unknown waters here, I studied machines not software and electronics, I don't know what or where to look.
Maybe you can help me, where should I look
It is a 500GB Samsung 850 Evo
P/N MZ7LN500
Model MZ-75E500
MSIP-REM-SEC-MZ-75E500
Do you need PSID: S/N or WWN numbers?
@@dexyco76 You'd need to open the drive and see what the part number of the controller chipset is.
@@dosdude1
I hope this is it
S4LN062X01-Y030
S3Z78MMG
U1605 ARM
MAIA