(this post is on Facebook too - apologies to everyone who's reading it twice - however I know not everyone is on FB)
I'm discovering that not all Dragon power boards are the same. It might be ageing components, or simply I'm driving beyond their specifications with my modified machines.
This all started when I noted only one of my three Franken-Dragon 64s would actually boot OS/9.
* Dragon 1 (32K, upgraded to 64K plus WiFi) won't boot unless I plug the CocoSDC into my MiniMPI. I've heard the ESP-01 wifi module requires a lot of power, maybe that's the issue there.
* Dragon 2 (64K plus DragonPlus) wouldn't boot even with the MPI. I swapped its power board for a D32 board, and now it booted, but crashed when it tried to access the memory on the DragonPlus board. Swapped for yet another D32 board and it works fine.
A few thoughts. I've heard of problems with the CocoSDC needing a lot of power before (John, this has caused you problems?). I've ordered a Gotek, and I'm going to see if that works better with an original DragonDos cart.
Is it simply a case of removing the capacitors and regulator on the power board and replacing with modern components?
Slight tangent - what the hell is OS/9 doing when it starts that upsets these machines? Flex and Editplus (on the Dplus machines) seem to work fine!
Dragon power board problems!
Re: Dragon power board problems!
Not sure about the PSU problems... hard to diagnose without measurements under load. It could be dry capacitors, or a worn out power switch, or a bad connection between the PSU and main board. The regulators are probably ok.
I don't think the Dragon power boards were designed to power much beyond a simple plug in cartridge. I remember Paul Grade being rather fond of his eprom programmer card even though it browned out the display while it was programming![Smile :)](./images/smilies/icon_e_smile.gif)
One possible supply upgrade is to replace the 5V regulator with a DC-DC buck converter. This would run a lot cooler. You're spoiled for choice with plenty available around the £2 mark. The most common type has an adjustable output so you need to be careful. I would be tempted to go for a type with a fixed 5V output, or modify the adjustable type by replacing the potentiometer with fixed resistors. I haven't tried it myself so can't tell you if there are any gotchas, (e.g. many converters are not stable with a high capacitance load), though one of my machines does run happily with an open frame switch mode supply in place of the entire power board.
Regarding OS9 crashing, I'm not sure if this is relevant but stock OS9 doesn't boot on upgraded D32s because it expects the D64 serial port to be present at $ff04. On a D32, OS9 ends up writing to the low pia instead and freezes the machine during boot. I think the D32 64K upgrade guide describes how to reset the machine and restart the OS/9 boot process so that it completes.
I don't think the Dragon power boards were designed to power much beyond a simple plug in cartridge. I remember Paul Grade being rather fond of his eprom programmer card even though it browned out the display while it was programming
![Smile :)](./images/smilies/icon_e_smile.gif)
One possible supply upgrade is to replace the 5V regulator with a DC-DC buck converter. This would run a lot cooler. You're spoiled for choice with plenty available around the £2 mark. The most common type has an adjustable output so you need to be careful. I would be tempted to go for a type with a fixed 5V output, or modify the adjustable type by replacing the potentiometer with fixed resistors. I haven't tried it myself so can't tell you if there are any gotchas, (e.g. many converters are not stable with a high capacitance load), though one of my machines does run happily with an open frame switch mode supply in place of the entire power board.
Regarding OS9 crashing, I'm not sure if this is relevant but stock OS9 doesn't boot on upgraded D32s because it expects the D64 serial port to be present at $ff04. On a D32, OS9 ends up writing to the low pia instead and freezes the machine during boot. I think the D32 64K upgrade guide describes how to reset the machine and restart the OS/9 boot process so that it completes.
Stew
Re: Dragon power board problems!
Though I believe one of Tony's upgraded D32s has my D64 serial upgrade board in it so that shouldn't be a problem, as this makes sure the PIA is only accessed $FF00-03 and the ACIA $FF04-07.sorchard wrote: ↑Wed Apr 29, 2020 3:30 pm Regarding OS9 crashing, I'm not sure if this is relevant but stock OS9 doesn't boot on upgraded D32s because it expects the D64 serial port to be present at $ff04. On a D32, OS9 ends up writing to the low pia instead and freezes the machine during boot. I think the D32 64K upgrade guide describes how to reset the machine and restart the OS/9 boot process so that it completes.
Cheers.
Phill.
Re: Dragon power board problems!
Yes, I should have mentioned this! So my D32 is almost a D64, except for the ROMs. I did have this problem with my other D32/D64 (which came with an original DragonPlus and was upgraded back in the day) - it wouldn't boot without a patch.
I'm currently getting round these problems by using 'external' power - my mini-MPI allows the CocoSDC to work just fine, and it's fun to have a use for my Race expansion box at last.
Mike Miller (@bluearcus on here) has been doing some thinking on how we could upgrade/uprate the boards. I tried new capacitors, as I had some to hand, but that didn't make much of a difference. Perhaps something more serious is required? As our machines get older, yet we're offered more fun things to plug into them, this issue might get worse!
Cheers, Tony
Re: Dragon power board problems!
I'm inclined to agree with Mr. Orchard here... measurements under load are key. I don't even know what some of your franken-dragons might draw on the 5V line, but on my oldest, most heavily used and customised machine power up is unreliable (won't boot, works after reset) with its original power board and some combinations of mass storage solution. Swap the power board, all OK.
Obvious possibilities are the main smoothing caps, which Tony replaced, and which although things have improved, don't seem to have completely cured his woes. Or a failed diode (scope time on the rectifier output).
If it is simply a question of not enough voltage out on 5V at the high loads of a Franken-Dragon with CoCoSDC then updating the voltage regulator would seem to be the next element of a possible fix. The LM309 really needs 7.25V out of the rectifier and caps stage to do its job. Below that, it's passing on some of the ripple. A more modern LDO regulator might manage with less, but picking one is a bit beyond me.
How about the MIC29310-50WT. 5V fixed, 3A. Should cope with much lower voltage in before losing regulation.
Obvious possibilities are the main smoothing caps, which Tony replaced, and which although things have improved, don't seem to have completely cured his woes. Or a failed diode (scope time on the rectifier output).
If it is simply a question of not enough voltage out on 5V at the high loads of a Franken-Dragon with CoCoSDC then updating the voltage regulator would seem to be the next element of a possible fix. The LM309 really needs 7.25V out of the rectifier and caps stage to do its job. Below that, it's passing on some of the ripple. A more modern LDO regulator might manage with less, but picking one is a bit beyond me.
How about the MIC29310-50WT. 5V fixed, 3A. Should cope with much lower voltage in before losing regulation.
Re: Dragon power board problems!
An LDO regulator would certainly make the supply a bit more forgiving in the event of sagging input voltage. There isn't much to pick between them, any 3 Amp device should work. Just check the datasheet carefully for notes regarding the output capacitor as they have a nasty habit of oscillating if the ESR of the output cap is too low (Been bitten by that one). Ideally you also want the mounting tab to be ground so that you can attach it to the heatsink without an isolating kit. Looking at the package drawings, it looks like a TO-220 package could fit in the space occupied by the TO-3 package quite neatly. The tab would be the 0V connection via the heatsink and the outer two legs could be connected via short wires to the existing input and output pads.
One thing I've always wanted to know is how much current is actually required, so I thought this a good excuse to take some measurements (Results attached). I found the differences in the 5V current a little surprising, though a quick glance at DRAM datasheets reveals that these are responsible for most of the difference.
The maximum design current on the 5V supply is 1.4A, and I would expect this to make the heatsink, rectifier diodes and external transformer get really hot. A small fan blowing air onto the heatsink and diodes would work wonders here, though the transformer would still be at risk of blowing its safety thermal fuse.
Please be aware that a different linear regulator is not going to run any cooler, regardless of power rating or type. That's just the nature of linear regulators: As the input current is nearly the same as the output current, the heat generated is a direct consequence of the change in voltage from input to output.
A switching regulator however has a fundamentally different principle of operation and will run cooler due to greater efficiency. I happen to have a cheap ebay buck converter so I quickly tried it in place of the 5V regulator to see how well it would work. I can confirm it did work albeit with some interference on the display. Admittedly I did have it hanging off the end of some crocodile leads so it would probably have given a cleaner display if fitted properly.
I recall there being a switch mode drop-in replacement available for TO-3 linear regs. These came from the USA but I'm not sure if there's a UK source.
One thing I've always wanted to know is how much current is actually required, so I thought this a good excuse to take some measurements (Results attached). I found the differences in the 5V current a little surprising, though a quick glance at DRAM datasheets reveals that these are responsible for most of the difference.
The maximum design current on the 5V supply is 1.4A, and I would expect this to make the heatsink, rectifier diodes and external transformer get really hot. A small fan blowing air onto the heatsink and diodes would work wonders here, though the transformer would still be at risk of blowing its safety thermal fuse.
Please be aware that a different linear regulator is not going to run any cooler, regardless of power rating or type. That's just the nature of linear regulators: As the input current is nearly the same as the output current, the heat generated is a direct consequence of the change in voltage from input to output.
A switching regulator however has a fundamentally different principle of operation and will run cooler due to greater efficiency. I happen to have a cheap ebay buck converter so I quickly tried it in place of the 5V regulator to see how well it would work. I can confirm it did work albeit with some interference on the display. Admittedly I did have it hanging off the end of some crocodile leads so it would probably have given a cleaner display if fitted properly.
I recall there being a switch mode drop-in replacement available for TO-3 linear regs. These came from the USA but I'm not sure if there's a UK source.
- Attachments
-
- Dragon-PSU-Measurements.zip
- (40.71 KiB) Downloaded 339 times
Stew
Re: Dragon power board problems!
Interesting measurements Stew! I assume those were all bare-machine, no disk interface and on factory original 4700uFs?
The TO-3 packaged switch mode converters ... this looks like them...
https://www.ezsbc.com/index.php/psu5.html
Too tempting not to try, have ordered a couple.
The TO-3 packaged switch mode converters ... this looks like them...
https://www.ezsbc.com/index.php/psu5.html
Too tempting not to try, have ordered a couple.
Re: Dragon power board problems!
Interesting, please keep us posted!
Re: Dragon power board problems!
That's right, the PSU boards are factory original, and no cartridges or other devices were plugged in other than a monitor. It would have been nice to get some scope captures of the ripple voltage but I can't get in to work at the moment.
That TO-3 device you've found does look like the one I had in mind. I'm intrigued to see how well they work.
Stew
Re: Dragon power board problems!
Well, the TO-3 LM309 replacements arrived.
and I (with some trepidation) took the de-soldering iron and braid to the power board on my workhorse D64.
Actually, I probably didn't need to desolder the main through heatsink GND bolts, they could simply have been left in place and the top side nuts and washers removed, but anyway.
Old school linear and new switched mode:
And new switch mode installed in the power board. There's a trim pot to set the voltage accurately.
Picture is good, possibly picture quality is a little improved overall, and stability with time / temperature seems better.
Booting Os9 from SDC seems more reliable (even with my new WifiSD card, which was really problemmatic).
Heatsink is now never more than warm.
I have an urge to try swapping the 7812 next, and then maybe even the 7912...![Smile :-)](./images/smilies/icon_e_smile.gif)
and I (with some trepidation) took the de-soldering iron and braid to the power board on my workhorse D64.
Actually, I probably didn't need to desolder the main through heatsink GND bolts, they could simply have been left in place and the top side nuts and washers removed, but anyway.
Old school linear and new switched mode:
And new switch mode installed in the power board. There's a trim pot to set the voltage accurately.
Picture is good, possibly picture quality is a little improved overall, and stability with time / temperature seems better.
Booting Os9 from SDC seems more reliable (even with my new WifiSD card, which was really problemmatic).
Heatsink is now never more than warm.
I have an urge to try swapping the 7812 next, and then maybe even the 7912...
![Smile :-)](./images/smilies/icon_e_smile.gif)