LCIII C22 polarity

[speakers]

So I picked up an LCIII for repair. The TDK PSU was toast - caps are good but the optocoupler is open-circuit and I'm awaiting a replacement. So I bodged into service a spare Astec Beige G3 PSU and the machine booted-up fine on its aged caps.

I quickly recapped with tantalums (as you do) but the machine then refused to run .. the PSU shut down after a second or so. I fiddled around and discovered that the -5V rail was the problem and the machine booted with this disconnected. Now, looking at 68kMLA, I find that the polarity of C22 is mis-marked https://68kmla.org/bb/index.php?threads/lciii-recap-apple-design-fault-47uf-reversed.27834/. Evidently the original aluminum caps were wrongly installed, didn't work, but didn't really care; but tantalums behave badly with their polarity reversed (and may emit magic smoke).

I flipped C22 around and all's well

 

[JustG]

That's interesting I recently recapped a LCIII with tantalums and followed the factory silkscreen for capacitor orientation including C22. I haven't done a lot of testing with it since recap but was able to get a healthy chime, video, cursor and flashing question mark. I seem to remember @Branchus mentioning that particular cap in one of his videos and stating that he was going to install it in the same direction as it was installed originally. Keeping an eye on this thread in case I need to flip that cap around!

 

[speakers]

Before I flipped this cap, I measured a load of just about 0.5A on the -5V rail before the PSU tripped. After flipping, I see only 10mA.

So, >0.5A being sunk by an incorrectly polarized C22 is not a good idea and it certainly exceeds the spec of both the cap and the supply. The -5V rail on the API-6120 PSU that I'm using is rated at 0.3A. Some people report that the -5V rail of their TDK units (rated at only 75mA max) is pulled down to around -2V with a reversed-polarized tantalum C22. These caps can be reverse-polarized to 15% of their rated voltage (16V) and that's 2.4V. In other words, the original TDK PSU doesn't trip and the -5V rail droops to a "safe" level for an incorrectly installed tantalum C22 to survive -- although surely the printer/modem ports won't work.

 

[JustG]

That computer has a TDK power supply. It's of course stuck back in storage, I'm going to pull it out next time I'm over there and test the -5V rail.

 

[speakers]

I got the original TDK PSU back in operation; blown fuse, switching FET, current sensing resistor and opto-coupler.

Here's a link to a post that includes a mostly correct schematic for the 699-0153 model which was close enough to my 699-0003 unit: https://68kmla.org/bb/index.php?threads/lc-tdk-699-0153-psu-reference.39601/

 

[speakers]

Oh .. and here's a recent 68kMLA post about how a reverse-polarized C22 indeed results in modem and printer port failure: https://68kmla.org/bb/index.php?threads/lc-iii-modem-port.42380/

 

[JustG]

Was able to test the TDK power supply voltages at the connector plugged into the logic board with the LC III turned on earlier today. +12, +5 and -5 all looked good with C22 orientated per the silkscreen. I'll need to build a loopback adapter and test out the ports next time I'm in front of the computer.

Were all LC III board marked incorrectly or were there revisions?

 

[speakers]

I’m glad your machine checks out.

There may be a dependency on the tantalums used. For the record, my misbehaving cap was a Kemet from Digikey - here’s the part info:

 

[JustG]

You may be right, I'm using 25V Kyocera AVX caps. Maybe the higher rated voltage has something to do with it.

https://www.mouser.com/ProductDetail/KYOCERA-AVX/TCJE476M025R0050E?qs=d0WKAl%252BL4KakoCbCQXOB2w==

 

[RetroViator]

I'll add my voice to those who had this problem with the TDK PSU and recapping C22 with a tantalum (I did not order the caps, but they look like Kemets). When following the silkscreen, I was getting -2.2V, but once I flipped C22 around (opposite of the silkscreen), I got a solid -5.1V. I actually have two different TDK PSUs, and saw the same behavior with both (using the same LCIII board).

 

[harrowm]

I get exactly the same results as @RetroViator