LM78H05K fakes?

[ef1j95]

I've been having problems with a high current 78H05K regulator (the 5A, TO-3 version of the 7805) on an S-100 memory card. Took a gamble and bought three spares from Taiwan via ebay. They finally arrived, but after soldering one in, it failed immediately. It went into thermal shutdown. It never gets above 2V output from the 8V S-100 line.

I'm guessing these are fraudulently relabeled 7805K's. The parts that came were "mil-spec" which were different from what the seller showed online, which only further raises my suspicions.

The original 78H05K will provide a steady 5V for about a minute, but will drop as it heats up. I don't think it's due to anything more than the load of the fairly large number of chips on the card. There are no signs of shorts. No other hot parts, and I switched out the tantalum filter caps, which turned out fine.

I think my next step is to try one of the buck converter replacements.

 

[Dwight Elvey]

Replacing one that is going into thermal shut down will not help.
You need a bigger heat sink, not another regulator.
You might also consider dropping more of the input voltage.
You only need about 2 volts across the regulator. Anything more than that
is just wasted heat.
Many S100 cards were designed with insufficient heat sinks for 8+ volts on
the unregulated lines. Many transformers put out 10 volts.
You might add a resistor in series with the input. Another
simpler method is to put a resistor in parallel with the regulator
but you need to be careful to make sure the regulator is still passing
about 1/3 to 1/2 the current.
Dwight

 

[pearce_jj]

Or implement a smps

 

[Dwight Elvey]

It is possible that mil-spec means these are burn-in failures.
For mil-spec they are usually put on a board with a large number
of these and have their outputs shorted. They stay like this
in a heated chamber for many hours( 100 hours as I recall ).
Of course, a number of them fail. It could be you got some of
the failed parts and not fakes.
Dwight

 

[Chuck(G)]

If it was my board, I'd probably go with a switching regulator replacement. Vastly less heat and more stable output--easier on your PSU.

Here's an example of a LM323 drop-in replacement

 

[ef1j95]

Yeah, that's the switching regulator I went with. I'd like to keep things "authentic" but I'm guessing that the single high current regulator is a bit of a design flaw for this card. You don't see many memory cards with them... usually several (>4) 7805's spread out for plenty of heat dissipation. This 78H05 gets pretty toasty, despite the rather large heat sink. I should try a little convection to see if that helps... but it's not a long term solution.

 

[Chuck(G)]

I suspect that one could devise a shell over the switching regulator so that it was indistinguishable at a distance of 3 feet from the old linear part.

One of the big weaknesses of S100 boards is the need for locally-regulated 5V. MITS had this issue right from the get-go, using TO-220 regulators and having to shunt them with 5W resistors to keep things from turning to toast (e.g. the 4K DRAM board). I remember experimenting with an LM723 and a 2N3055 as a replacement switcher for a LM323K back in the day. The problem was that the thing was just too large physically.

 

[Uniballer]

I would try measuring the actual current draw, just to be sure you are really on the right track (e.g. something may have failed and is now drawing too much current).

 

[ef1j95]

Clipping the fuse 2 lead on the N8VEM extender board and inserting my ammeter into the loop turns out to be a convenient way of monitoring the current. The measured power on the 8V line is consistent with changes measured for the overall computer using a Kill-O-Watt.

With half the TMS4044 chips on the board, I measured 1.73A running a short program and 1.69A in a WAIT. My numbers are consistent with 50mA draw for each TMS4044, which means the total current draw should be around 3.25A for the full board. That should be well within the specs of the 78H05K if it is running properly.

Pulling the chips, I noted that most of the 4044 pins have been very corroded. Interestingly, there is one National Semiconductor MM5257N-L that is absolutely pristine. Did TI use crappy metals? I've been cleaning the chip pins as I restuff the board. Not sure if that will help?

 

[gekaufman]

I can't comment specifically about 78H05's, but counterfeit LM323's (5v3a) are well known.

TI is still making "real" LM323's in TO-3, but you might want to sit down before you check the price...

LM123's are the -55C to +150C temp version, and genuine parts are often more reasonable in cost as they are less well known (and less likely to be counterfeited).