@bytex64@awesome.garden on awesome.garden
So hey, I kinda lost the rest of the month, there! I actually got these some time ago and I have all the parts, I just haven’t felt like putting them together. So let’s at least get started on that.
Earlier in this thread 5 posts
@bytex64@awesome.garden on awesome.garden
Hand made edge connectors, just like mom used to make.
The slots in the edge connectors are too thin for the board edge mill so I left them uncut. I cut them myself with a jeweler’s saw, and this is definitely not my best work. In fact, the slot in the larger one was so narrow that it gripped the key in the connector and ripped it out when I removed it. 😬 But they do both fit fine.
@bytex64@awesome.garden on awesome.garden
And yes, I hand chamfered the edge as well. :) That’s all I have time for tonight, as I also spent a lot of time cleaning up my workspace. Ready for tomorrow. :)
@bytex64@awesome.garden on awesome.garden
I have the one surviving connector key, which pulled out easily. Plastic milk carton is about the same thickness, so I cut out a replacement. Works well enough!
@bytex64@awesome.garden on awesome.garden
Finally getting back to the MFM emulator. Let’s do some hot plate work!
What’s my paste technique? “Uh, just put that anywhere.”
Conversation (35)
Open reply Open reply
Chip
@bytex64@awesome.garden · Jun 23
There is, predictably, a reason you don’t do this. Unless you get really good at micro-metering the paste, you tend to make a mess that takes you almost as long to clean up as doing it manually.
It does look like the RP2040 self-centered just fine, though.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
But I have made another mistake that helps here. I went to order more solder braid and didn’t think about the size, I just got the smallest one I could find. Last time that was 2.5mm. This time it was 1.5mm. Which is almost unusably small unless you’re cleaning up things like this.
Some wick, tacky flux, and drag soldering, and it looks much better.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
Some of you may be asking, “how on earth do you see those pins, let alone get an iron on them?” You don’t. At this size, you load them up with flux, put a blob of solder on your iron tip, and pray to the chemistry gods that your flux and solder mask keep the solder where it’s supposed to go. I just slowly drag that blob down one edge and check it with a loupe afterwards. I use a meter to check shorts at the other end of these traces, and I have a program that strobes each GPIO to make sure they’re all connected.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
I think that’s everything I need to test the RP2040. For some reason when I designed this, I thought I had 27 ohm resistors in 0603. I don’t. So I’ve just bridged them. It’s in the USB 2 data path, it’s probably okay.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
#*%$@! I wired the CC pins into the SBU pins. This is going to be an annoying bodge.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
For short bodges like this, I like to strip piece of wire and use the wire itself as a handle when soldering. Then bend it where it needs to go, solder, then cut off the rest of the wire. This went pretty smoothly, thankfully.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
Hm. Test program isn't working. Looks like it starts to access flash, then stops. External oscillator never starts up.
This is a newer flash chip (AT25SF128A instead of the usual W25Q080). I wonder if I compiled this with an old enough SDK that it doesn't know about it.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
Oh right. This isn't designed to run from flash. It runs from RAM so I don't need to have the flash installed. It's been a while since I've used this.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
The tool is pretty simple. It outputs N+1 pulses on GPIO N. You check every pin in sequence, counting up one by one.
All pins tested good. 👍
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
It doesn't seem to want to run anything from flash, though. Hm. A problem for tomorrow.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
So it’s even weirder. It will run from flash when you start in bootloader mode and do a picotool reboot, but it doesn’t run from a cold boot. Still fails when I roll back to the 2.0.0 SDK. I wonder if this is a quirk in this particular flash revision.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
Slapped a Winbond chip on it, let’s see how that does.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
Same problem. But it does start to work when I increase the flash SPI clock divisor. So it could be a signal integrity problem. But there's another clue.
The below scope traces are from the "hello_uart" example. Running this several times shows that the timing is changing from run to run, and it shouldn't be. At the 115200 baud rate, only the latter part of it is decodable. If you increase the baud rate, the earlier character start to decode.
I think we have a clock stability problem. I kinda just put 33pF load caps in here. I bet that's for the HC49 cans, and I need smaller ones for the SMD crystal I'm using.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
I just removed the load caps, and it completely fixed it. Rock solid now, even at the usual SPI flash clock divisor (2).
I'm not sure I've used this footprint with a full ground flood front and back before. I guess those wings for the SMD HC49 are providing the needed load capacitance? Now there's a fun PCB hack. 😆
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
Hm. According to some calculators I'm finding online, these patches should be around 0.25pF, so I probably just got lucky (OTOH maybe the parallel plate model doesn't work with plates this small). I'll find the proper load caps and see how that goes.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
The load capacitance of this Würth crystal is 15pF, so 22pF is a lot closer to what I need. And I have those… in 0805.
Thankfully we’ve got extra pads to work with. Footprints are just like, an idea, man.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
That's still failing at a SPI clock divisor of 2. With a divisor of 8, it works. The UART trace is better, but still jittery. Let's try 18pF? That's the only other thing I have.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
18pF definitely looks better, but still failing at a clock divisor of 2. I dunno. I'll just get rid of them for now. Certainly something I'll have to fix if I want to put these into production.
Open reply
Chip
@bytex64@awesome.garden · Jun 23
Thread reply
And works with the original AT25 chip, too.
Open reply
Chip
@bytex64@awesome.garden · Jun 24
Thread reply
Yes, I have forgotten to get 100 ohm 1210 resistors for two orders now. :/
Open reply
Chip
@bytex64@awesome.garden · Jun 25
Thread reply
Built the software without too much drama (what compiler defines ulong by default?), and loaded it up. There are no instructions on what the hard drive image needs to be, so I trial and errored my way through that. And when it appeared to load the image and work properly, I got a whole lot of nothing.
Then I thought about it for a second, realized that this thing works on raw flux streams, and I fed it a file full of zeroes. I rewrote the image with random data and voila! Garbage in, garbage out.
Oh also I wired up the STEP pin and the piezo clicker works. :D
Open reply
Chip
@bytex64@awesome.garden · Jun 25
Thread reply
It appears to be attempting to find track 0 by seeking repeatedly. But it’s stuck.
Open reply
Chip
@bytex64@awesome.garden · Jun 25
Thread reply
It appears that the DIR pin is operating backwards. When asserted, that should step in, towards track zero, and outwards when deasserted. So the T1200 is at track 0 and wants to step to track 1, but it's instead trying to go the other direction and staying at track 0. Repeat ad infinitum. I think I can pretty easily invert the GPIO signals.
Open reply
Chip
@bytex64@awesome.garden · Jun 25
Thread reply
Manual testing shows it working now (thanks to the RP2040 designers for putting selectable inverters on both inputs and outputs before the mux), but now the laptop doesn’t want to start so I guess I’ll be testing it tomorrow.
Open reply
Chip
@bytex64@awesome.garden · Jun 25
Thread reply
Well that sounds more promising. The T1200 does give up on it after a bit and continue now, so I’m guessing this is just no data found. Time to find a low level format tool.
Open reply
Chip
@bytex64@awesome.garden · Jun 25
Thread reply
Hmm. Not working. Just keeps making that chirp sound over and over again as it accesses the drive.
Open reply
Chip
@bytex64@awesome.garden · Jun 25
Thread reply
Debugging says it’s banging against the upper cylinder limit. So basically the same problem as before, but the other direction. I wonder what it’s looking for.
Ordered some longer IDC cables so I can debug this more effectively.
Open reply
Chip
@bytex64@awesome.garden · Jun 25
Thread reply
Part of the problem might be that the drive end of the cable is a slimline version that doesn’t actually fit well into this standard jack (I had to clip out some bits to get it this far). So maybe some signals are just not connecting properly. I’ve also got a male IDC to fit into the cable so I can plug in the Glasgow for some more detailed signal tracing.
Open reply
Chip
@bytex64@awesome.garden · Jun 27
Thread reply
I built a little rig to test this thing manually. Buttons for drive select, head select, direction, and step. LEDs to indicate ready, track 0, seek complete, index, write fault, and servo gate. And one extra line for the read pin.
I spent some time tracking down things plugged into the wrong pins, but it didn't seem to show any problems with the device itself. Polarities look right, track 0 behavior is right, seek complete works right, index pulse looks right, servo gate looks right. So yeah, maybe a better cable will fix it. 🤷♂️
Open reply
Chip
@bytex64@awesome.garden · Jun 30
Thread reply
Side note: an eBay seller had ten of these 26-pin cables for $5 each plus like $9 shipping. So I uh, cleaned them out. 20 cables for ~$24 shipped.
Apparently they were for the Raspberry Pi, which originally had the 26-pin GPIO header. And they were made almost instantly obsolete by the “plus” models’ 40-pin header we commonly associate with the Rapsberry Pi today.
I don’t think I’ll sell or even use this many, but a pile of ribbon cables and IDC connectors is a good raw material to have around.
Open reply
Chip
@bytex64@awesome.garden · Jun 30
Thread reply
26 pin IDC is also useful for early Sony floppy drives, and I have one of those.