Видео

Glad you enjoyed it! Dave

Indeed! Looks like I have some binge watching to do :) All Hail the Algorithm!

I'm glad the videos are getting a wider audience. It will take while to go through them but I do generate the subtitles (not automatic) so they are pretty accurate and you should be able to watch at 1.5x speed most of the time. Hope they are of interest. Dave

The costs are very high. I paid several hundred euros for all the chips alone - 2 to 3 times as many. The dual-port RAM IDT7134 alone costs almost €50 with shipping costs from England. The breadboards at Mouser cost around €10 each including VAT. I have 20 of these alone, which I bought with and without power rails. I use the recommended breadboards from Breadboaard Systems, as recommended by Ben Eater and Dave. And then there are the measuring devices, whereby I only have a high-quality multimeter from Brymen, a 16-channel logic analyzer LA2016 and a PC oscilloscope. All in all, I estimate the total cost at around €2000. Dave will probably be even higher as he uses 3 different logic analysers, one of which is quite expensive.

The cheapest way to play with XT class machines with XI 8088 BIOS is via Book8088 on AliExpress which is about £200 for a pre-built laptop (though with modern surface mount components). As Peter below says the parts to build this are not cheap, I would estimate probably about £400-£500 for the current state but I tend to buy in bulk for spares & stock so if buying just the quantity needed then could be a bit cheaper. I am doing this for learning NOT for a cheap way to build a PC! The power strip bus is what Ben Eater used so just copied him. I have got some Veroboard/stripboard with 0.1" connectors to do a 'proper' 78 pin bus but not needed to use that yet. If I do a 32 bit i386 version then that will definitely be required! If you were to build the Nanocomp 8088 with LED display and keypad I think you could do this for less than £150 its the 8088 monitor ROM you need for that but I have put that on Github. An original 8088 Intel processor you can get for £10 on eBay. New 80C88 from Mouser is £43. The switches for the keypad were about £25 with the keytops, 5 breadboards about £35. Not found any 25 key keypads elsewhere any cheaper (4x4 16 keys lots). Rest of the chips you could get for less than £50. Just build stuff out in stages and buy things when you need them (but buy breadboards in 10's as they are cheaper that way). Cheap breadboards will just waste a lot of time! To build the project you will need a certain amount of test gear which you could already have but I have collected mine over the last 2 years (probably about £900 worth). You would also likely need a bench power supply, mine is currently set at about 3.5 amps with floppy disk drive so USB power adaptors not really up to it, EEPROM programmer, plus logic analyzer (I think 2 x LA2016 is the best price/performance). An Oscilloscope is useful but not essential, there are cheap USB ones about though the £210 Hanmatek standalone one I have is reasonable. Plus soldering iron, multi-meter, wire strippers...... The breadboards from Mouser are about £7 each when buying 10 or more and additional power rails about £2 so for 15 boards plus 5 extra power rails (you can take one power rail off each of the breadboards for the buses) you would spend £125 on the breadboards alone for the current project booting DOS. Most components can be found on eBay for $5-20 but certainly not cents (perhaps 20 years ago before they became 'vintage/retro'). Except for the dual port RAM which I got when still manufactured for about £12, but now more expensive. Also the 8288 bus controller was tricky to find and cost about $20 and took a while to be delivered. I am pretty sure the clock and bus controller chips could be done with PLDs but that's for another day! Hope this helps. I would start small and build out over time or use the XI 8088/Book8088. Dave

Tbh, it's not easy to follow along with every detail and then also add comments / questions. I see two options: Either build it by yourself (like people build the Ben Eater projects), but following this here would be a BIG commitment. I'm wondering if anyone managed to do that. The other option is just to listen and to do it in your head. But then you'll mentally skip many implementation details

​@@michaelj7677 Then perhaps I am an exception. I'll try to recreate the project. However, I will replace one or two components. For example, the 74LS574. These are very expensive. Therefore I use the original 8282/8283 chips which can also drive a higher number of TTL ICs. Similar to the 74ALS series. At the same time, I am also writing documentation - in German - for myself, in which all the chips used are explained so that I will still be able to understand the connections in 10 years' time. Unfortunately, I can't make any progress at the moment because the circuit diagrams on Github are no longer up to date.

Hopefully if you follow the 36 odd detailed videos then this is not too much to consume at a time. The first 12 and the 1st summary will end up with a functioning Nanocomp 8088 with keypad and LED display. This summary was trying to squeeze more than 12 hours of video content into 1 video. The build sequences in the regular videos are sped up between 5 and 10 times so the videos in real time would be been closer to 40 hours in total! One of the reasons I am doing it is so that others don't have too (Peter is obviously having a go) as I am fortunate to have the free time to spend on it. I am just video blogging the process. At some point I may do some more tutorial type content but I am just doing this to 'scratch and itch' , to really understand how PCs work and not turn it into a million plus subscriber channel!

OK thanks, I remember the OS/2 Museum blog on this, it wasn't straightforward at all. May give it a go at some point. Thanks. Dave

Thanks for the info. I didn't think too much about the BASIC issue at the time but you are correct about the IBM BASIC ROM. Not really worth worrying about as from memory the cassette BASIC in ROM couldn't talk to floppy disks anyway. Will try GWBASIC from later DOS versions in the next video. Dave

Thanks, yes the BASIC I didn't think too much about at the time but I seem to remember the XT had cassette BASIC in ROM (which was unable to talk to floppy disks). I'm not going to stress too much about that. May try Microsoft GWBASIC from 3.3x but no point in messing with IBM BASIC. Dave

I will take a look. The main issue is the Breadboard PC is basically an XT with no DRAM (or DMA refresh), two interrupt controllers (like an AT) and PS/2 keyboard controller (with mouse support). Main issue would be test ROM may have trouble determining system type. DOS 4 with TEST1 variables set to 1 will output quite a lot of debug info. I think i still have a few DMA/FDC issues. So will try both approaches and update in the next week or so. Thanks for the offer ideas. Dave

Thanks for the info. Have downloaded a copy. Will take a look at it in the coming weeks. Dave

I think this summarizes the situation quite well! www.os2museum.com/wp/how-not-to-release-historic-source-code/ (Michal Necasek of OS/2 Museum). Dave

That would be good if MS could work with Michal Necasek to get a decent release out (for 2.x, 3.x and 5.x as well would be ideal). Best bet is check in ZIP files with the originals, its not as if they couldn't 'touch' the files with original dates that they needed to cleanse the comments. Dave

Sorry, it was one of the first videos I made so not everything went right. Will be remaking a tips video when the PC project is done. Will make sure everything is readable then.

Agreed about WinCUPL, it crashes so often you can't rely on it. Just checking the .doc files after compiling though would avoid the issue. Not seen anything out there to do CUPL/PLD for Notepad++ though I guess the comment syntax is common enough which might help. Will take a look at this when I do a PLD video in a couple of months once I've finished the PC. Dave

@@breadboardinglabs I don't have too many issues with crashing, that seems to be fairly rare in my experience. But when it does happen it'll be for some totally innocuous reason like hitting backspace during editing, and of course you never know when it'll happen. I just save frequently.

My first PC project was PC-AT with DOS 3.2 writing Windows 1.0 programs (so I'm old too!) I know 4.0 was buggy, was the default O/S on PS/2 Model 70 I used. BUT it does allow me to get detailed debug info as it boots up which will hopefully allow 3.30 and 5.0 to boot too. Dave

Thanks though I suspect if I tried to hammer those DMA chips one would probably break up and hit me in the eye! Bending a couple of pins back will have to do! Dave

Good spot, when I bought the AS6C1008 from RS in the UK there wasn't much in stock anywhere (compared to now, and I tend to now use Mouser for larger orders). It wasn't a problem as I think from memory the voltages where around 3.5V anyway on the data bus (could always add 10K pull ups to the data bus if it was an issue). If doing this again I would probably use the faster 32K SRAM I used for the video controller (71256SA) as this are narrower and have fewer pins which makes fitting to breadboards easier. For the 8086 project I have used the 512K AS6C4008 which is TTL compatible. I would suggest getting a TTL compatible SRAM chip if you want to repeat this project. I think the AS6C1008 was the least worst option available at the time given 32K chips were out of stock. Dave

Not come across those buffers before and I have a tube of 74ALS245's so that was the path of least resistance! Once its all working I will try pulling out the extra 74ALS245 to see if it still works. I suspect the dud DMA 8237A controllers wouldn't have helped. Will take a look at these for the next project. Dave

Glad you enjoy them. Thanks for the feedback. Dave

Yes, just trying to do things methodically and not too fast! I've been lucy so far with the eBay chips I have bought so far, only been a matter of time before some duff ones turn up! Thanks for the feedback. Dave

It's sometimes a struggle to keep under 20 minutes, my 6809 ones tended to be 40+ minutes but not sure many people had the time to watch the whole 40 mins. The booting MS-DOS I suspect may take a little while to get right as it depends on the BIOS, DMA, Interrupt AND Floppy Disk controllers all playing nicely! Will release it as soon as it is working! Thanks for the feedback. Dave

The booting MS-DOS I suspect may take a little while to get right as it depends on the BIOS, DMA, Interrupt AND Floppy Disk controllers all playing nicely! Will release it as soon as it is working! Thanks for the feedback. Dave

Thanks for the feedback, glad you have found the videos useful. I am just doing this for my own interest and if others find it useful/interesting then that's a bonus. You can contact me at the email breadboardinglabs (at symbol) g m a i l.com . Replace the @ symbol and spaces. Dave

The floppy disk controller (FDC) chip handles all the decoding I think. We just get the sector bytes dumped in RAM buffer by the DMA controller. So the boot sector for example should get returned from the FDC to 0000:7C00 where it can be executed. Dave

There are a number of good books out there (as well as the original IBM documentation). Some topics are better handled than others. DMA was strong in the one I referenced in this video. Dave

Yeh I know its grown a fair bit but I find the decomposed diagrams more difficult to follow especially when you can highlight the tracks on a single large schematic. I think multiple sheets is less of an issue when using PCBs but for Breadboards I try and make the schematic match as much as possible the physical layout. I also re-order the symbols so they have the physical pin layout and not the default nice logical ones which doesn't help physical breadboard layout. I can't wait to get DOS running either but I suspect troubleshooting DMA and Floppy Disk Controller may take some time! Dave

I like the all-in-one-schematic. Somehow this makes it easier for me to get the whole picture. Maybe the +5V and ground busses could be removed in order to reduce visual complexity.

I try to make the schematics match the breadboard layout as much as possible (not 100% but say 90%). Helps when working out how physical pins are wired to ground or Vcc. This avoids all the effort needed to create a physical layout like is required in PCBs, especially when using the symbols I have customized to 'Breadboard' physical pin layout rather than the logical layout default. Makes physically orientating chips to make the connections easier. It works for me but I am not sure many folks are making such large breadboard projects (except perhaps for some Ben Eater CPU reproduction projects). I prefer the one large schematic as it makes it easier to see how things are connected (which is not as big an issue if making PCBs) especially with the highlighting tool. That gets a bit messy with multiple pages in a schematic. Dave

Its from the XI 8088 Speaker circuit which is 0.01 uF or 10nF. Have updated the working copy of KiCad diagram so will upload this when DMA and Floppy Disk Controller working. Dave

It's a VT82C42. This is a chip dedicated for Keyboard/Mouse decoding. It was used on old Computers like 8086-80486. You do not have to program the chip.

As Peter said the VIA chip is implemented in hardware so does not need programming like the 8042/80C42 microcontrollers which were generic chips needing the BIOS Keyboard support programming into them and I think these may be tricky to come by. Video #27 (part 5) included this text on one of the slides from the VT82C42 datasheet "The VT82C42 is a compatible direct replacement for the Intel 80C42 BIOS version of the Keyboard Controller. The VT82C42 is fully implemented by hardware logic so that it has a very fast response capability for any command issued by the host. In addition to keyboard support, the VT82C42 also offers PS/2 mouse support." Thanks for the feedback. Dave

Thanks, just planning out the floppy disk controller and DMA then we should be able to boot DOS! I got the LA1010 first as it was the cheapest but it had limitations as it has no onboard memory so can swamp the USB bus at higher capture frequencies and can only handle a few leads at video/VGA frequencies (fine for address and data bus). The LA2016 I think is the sweet spot in price/performance, can do 200M Samples per second for all 16 lines and is a moderate price. This is great for most things but when trying to view 16-20 bits of address & 8 bits data bus then gets tricky syncing results from two analyzers (especially via CSV export if debugging boot failures ). The LA5032 is great and does 500M samples per second but costs a lot more (mine was about £450 and I got part funded as a gift so helped justify it!). I have 8 way Dupont connectors and cables for 16-24 of the LA5022 lines to make it easier to monitor address & data buses, so when I want to have more flexibility I use the LA2016. The LA1010 I am using with the POST port now. So I would recommend starting with the LA2016. I probably would have bought a second LA2016 if I didn't have a chunk of gift cash to put towards the LA5032 (it was a real treat!). My only mild criticism of the LA5032 is that the Kingst VIS app only has 16 bit decoders not 20-24 bits. I may try one of the open source applications to see if this is better. But Kingst VIS is pretty good for my needs (except it crashes for me when applying glitch filters which can be needed at higher capture frequencies). You can download Kingst VIS and try it out before buying one of the devices as it generates demo data if no physical device attached. www.qdkingst.com/en/download. Hope this helps. Dave

Agreed, once I have a DOS command line running, it will be just like coming home after a long journey! Thanks for the feedback. Dave

GitHub just been updated now it is mainly working! github.com/breadboardinglabs/Breadboard8088PC Dave

Excellent. Thanks! 🎉 I am beginning to understand the enormous amount of work you are putting into this project. Thank you very much for it. ❤

@@breadboardinglabs All GAL Chips are 15ns?

There are a variety of speeds from 7.5ns to 20-30ns from the Mouser link www.mouser.co.uk/c/semiconductors/programmable-logic-ics/eepld-electronically-erasable-programmable-logic-devices/?q=ATF22V10C&mounting%20style=Through%20Hole The 15ns tend to be the cheaper ones (7.5ns similar) however getting chips faster than what is needed could cause ground bounce spikes and other problems as rise and fall times will be faster than what is required (I am a hobbyist not a professionally trained engineer!). Dave

@@breadboardinglabs I would like to come back to the speed of the GAL modules. I saw in an earlier video that you used a 10ns GAL. My understanding of the speeds of the 8088 circuit is that all GAL devices can have 15ns cycle time. Do you agree with this? The background to the question is that I have a lot of 15ns GAL devices (22V10, 16V8, 20V8) and a few with 7ns. I do not have 10ns. Hence my question. Otherwise I would have to order 10ns components.

It would be possible but I find duplicating information makes it more likely that it will be inconsistent. Most of the IO Port details are defined in the PLD configuration which are copied from the Excel document summarizing a lot of this info. As things change, its more likely to become inconsistent. I could annotate the schematic once its all working. Dave

That makes sense. I hadn't even thought about that before because I hadn't looked at the PLD files (yet). I made the suggestion because Sergey Kisilev does the same thing. But yes, when everything is finished, that's better. Thank you for taking the time to answer my suggestion in such detail.

Good question without a simple yes or no I'm afraid! The worst case scenario is that if configured as the PC-XT, the Timer OUT1 and DMA DREQ0 are connected for refresh which would take CPU cycles away without doing anything (the DMA would take control of the buses, read a row from the Static RAM but not do anything with it). This MIGHT be useful if trying to make the PC cycle accurate with PC-XT to run some Demo programs like 8088 MPH www.pouet.net/prod.php?which=65371 with CGA video (not done CGA yet). However the XI 8088 and Breadboard PC don't need DRAM refresh as you say as they use SRAM. The Timer OUT1 is connected to a flip flop in the PLD U35 and then via REF_DET to Port B. This is then polled by the BIOS for more accurate delay loops based on the PIT_DELAY config setting which looks for the refresh_flag on Port B Pin 4 in units of 15us. So in summary the BIOS does not need configuration to disable DRAM refresh. Timer 1 is still used to generate 15us intervals, it is the fact that Timer 1 OUT1 is NOT connected in hardware to DMA DREQ0 (channel 0 which is used for DRAM refresh) which disables the refresh activity. The PIT_DELAY setting enables the BIOS to use the refresh timer for delay purposes but if this is not enabled then the DRAM refresh activity would be unchanged, this can only be enabled in hardware. Hope this helps. Dave

You are welcome. Just struggling with 8253 Timer I/O wait states at the moment. Next video soon after that! Dave

Glad it was helpful! Dave

It will work but the design would need to be modified a little to no longer need to use A0 to switch between character and attribute. The chip select for the CPU access would also need changing to include A0 and !A0 to switch between the two 2K RAM chips

Thanks. Working on BIOS video at the moment. Will hopefully publish Sunday/Monday. Dave

Thanks, There are a couple of summary videos for the 8088 Breadboard PC and MDA Video controller too (or the whole series in the Playlists). Dave

Thanks, yes the first beep (or at least multiple beeps for some issue) will be the first time I have ever wanted to hear them! I used to disconnect the speaker on my old PCs. Dave

I will upload the schematics and updated PLDs by the end of the week. I want to get the new board working first before releasing them as there is a good chance they will be wrong in some way. As far as SPK on bit 4 of Port C I think that was just me highlighting speaker related pins (no deep underlying insight I'm afraid). It looks like pin 4 was documented as Spare as you say but SPK on Sheet 8 of the system board is taken from the output before the 33R resistor so is the 'sum' of the Out2 from the timer or the SPKRDATA from Port B. I guess might allow a background TSR to monitor the speaker output from another program? Not sure this is wired in on the XI 8088, nothing before the 33R resistor. Will take a look for the next video. Thanks for the feedback. Dave

You are welcome! Dave

40 Mhz is fine as VGA 640x480 is 25Mhz. 40 Mhz would do SVGA 800x600. The others will be XGA and higher resolution.

Thanks for the feedback, I'm doing this primarily for my own interest but if it helps others to understand how this stuff works then great, I've not been actively promoting the videos so just organic views at the moment. Once I have a working DOS Breadboard PC I will do some more shorter videos on the running it rather than building it as I think this will have a wider appeal. I have been spending a fair bit of time recently sorting out the design for the final phases to get DOS booting. Will update the files and Github later next week after I have managed to get the next video out and the schematic is pretty close to final version. Feel free to comment on the individual videos, I may need to re-watch them to remind myself of what I did! I am very much learning as I go a long so happy to express opinions on issues (though I may not be correct!). Dave

I'll be striping out the 68xx components and LED Display and Keypad once I have the BIOS and MS-DOS loading, they are just to help bootstrap the BIOS as I already had the 6809 Monitor (though saying that the MC6850 does 56Kbs and the 8251 only seems to do 9600 baud!). Recent videos I have added mouse pointer highlighting, will change the Notepad++ font size as you suggest. Sorry to hear about your internet speed, not sure if you can download the videos for offline viewing (may need RUclips premium now or utility like ClickGrab?). Thanks for the feedback. Dave

@@breadboardinglabs 😃 Sounds great! Thank you very much..

Thanks for the info, I was aware of better ways of doing things but have tended to stick to what I had which already works. I will be doing a PLD video at some point so will include this info as well. Dave

The build sequences will be 12x speed, 1min=5s. Real time would be tedious I can assure you! Dave

@@breadboardinglabs That's correct, but sometimes you are talking so fast that it's impossible to follow you if someone is not a native English speaker... So I slowdown your Video to 0.75 or 0.5 too. 😀

I suspect it was that the buffers at that time (LS series) may not have had enough power to drive all the onboard logic AND allow for 2 TTL Loads per expansion card (16 TTL loads) or the capacitance of a single bus could have been too high so they split the buses. I suspect the Engineers being IBM made sure it met all specifications and didn't 'wing it' so I suspect it could be over-engineered. ALS logic vs LS has half the max input current so could do twice the fan out. Compaq DeskPro 386 and PS/2 70 uses ALS chips I think. I thought originally X-Bus was related to DMA but I can't see any way how it is (we'll see when I try it!) after spending many hours looking into it. As far as LS and ALS TTL vs HCT, I was trying to use early 80's components where possible when I started the 6809 project and stretch to early 90's for things like VGA DAC, Floppy disk controller, keyboard controller etc. The LS and ALS seem to be easier to find in DIP packages (certainly with local component supplier rather than Mouser/RS etc) and once I had a number of them it was easier to keep going! I do have some HCT inverters for crystal oscillators but not really any good reasons not to use HCT just kept going with what I had, I'm sure you will be fine. Not sure if the HCT may be more sensitive to the capacitance on Breadboards at higher frequencies (a power strip bus has about 20pF between rails) so rise times might be slower with lower HCT currents? Might be interesting to do a video on it a some point? I have PLDs in the list already! Comparing LS, ALS and HCT logic families on breadboards! Thanks for the thought provoking feedback. Dave

Found this link from Sergey forum.vcfed.org/index.php?threads/ibm-xt-8th-slot-video-card.24433/ Which says "The additional buffering of the X-Bus is required because buffers / transceivers on the ISA bus (between CPU and ISA) won't be able to pull all ISA slots and I/O devices on the system board together.", not a problem for this project for the reasons I gave below.

Yes I think in my thinking about "PC compatible" I mean 8088/8086 and MS-DOS, PC-DOS and perhaps DR-DOS. To make something 386/486/Pentium+ compatible is completely different (though these all still power on in real mode like the 8088/8086). Its only with 286/386 versions of Windows (and probably OS/2) I think that the drivers took over from the BIOS (I am not an expert in Windows drivers!). I suspect all the stuff like PCI, PCI Express and USB would be a major undertaking to be able to do the same as I am doing here, far too many pins on chips and too complex for 1 person's brain! I am pretty sure Windows 1.0 and 2.x just sit on top of DOS with no VxD drivers etc. No plans to make a PCB as Sergey Kiselev has done this already for the XI-8088 and other than the PCB's I made in 1981 I haven't made one since! I was mainly interested in building large projects on Breadboards when a lot of 'professional' advice said not to try it as it wouldn't work (reliably)! Dave

I will do when I get to that bit, I think in answer to your question "what precisely makes a computer PC compatible" I think it is mainly the BIOS as this can hide most incompatibilities. There are a few things like where the default location of the RAM for a video card is but the card can override that by providing its own video BIOS. The PC was a clever expandable design which I think explained its success even though the original 1MB RAM was a limitation in later years, in 1981 it was a massive amount! Thanks for the feedback. Dave

15ns PLDS are what I have been using, you don't need the faster ones for what I have been doing. I have tried to get ALS for some Video Controller chips where available as they theoretically could be used for SVGA 800x600 at 40Mhz (I am am planning to try this at some point) whereas LS TTL would run out of steam around 30MHz but both are OK for 25Mhz VGA. Also ALS have better fan out for buffers etc so don't need so much buffering for larger projects like the Breadboard PC one. Dave

Ok, thanks!

I couldn't find larger than 4 x4, certainly not 5 x 5, just remember that the scan codes for the keys will be different if you use the Nanocomp monitor keypad scanning code. Dave