Mar 87 Letters
Those of you who have been around in the Computer field for a while will recognize the name of Loy Spurlock, contemporary of Dan Paymar (Lower Case Adaptor for Apple II), Randy Hyde (Lisa Assembler) and Dave Gordon (Programma International & DataMost), from the early Apple II days. Loy started Computer Forum in Santa Fe Springs and presently owns a service company in La Mirada called Computer Quick which installs Levcos MonsterMac. What makes Loys company unique is that they actually scope out dead Macs to find out what killed them, rather than just do board swaps. From this perspective, Loy tells us the truth about the Mac power supply problem. -Ed.
Macintosh Power Supply failures
Loy Spurlock, President
Computer Quick, inc.
La Mirada, CA
The Macintosh power supply has been getting a bad rap lately. I have seen articles written by writers claiming that the power supply is no good. The only reason those articles ever saw print was simply because they were written by a writer who normally writes for the magazine, not because they knew what they were talking about.
The problem is, these authors dont know from Adam how the power supply works, what goes bad, or anything else about them --or they wouldnt have written the article.
This whole hullaballoo reminds me of the same thing that happened with the Apple II switching power supply. The Apple II, much the same as the Mac is now, was a leader in the use of new technology. Up until the Apple II, computers were generally powered with the older transformer type power supplies.
Anyway, at one point, there was a barrage of articles written about the problems with Apples switching power supply --one such article was called, A Worm in the Apple. All the Authors were bemoaning the fact that it was a switching power supply. If only it were frame transformer, everything would be all right.
My guess is that when something is wrong, people are usually ready to throw the blame onto whatever is new. Since switchers were new to micros, that technology took the rap, rather than the amount of current it was designed for.
In actuality, the power supply was designed to produce a certain amount of current at each of the various voltages. If Apple had used a standard transformer type power supply of the same currents, the problem these authors were writing about would still be there, but they would be writing about an under powered power supply instead of the weaknesses of a switcher. Also, if Apple had designed their switcher to produce more current, there wouldnt have been an imagined deficiency, and the Author would never have written the articles in the first place. More about designed current levels later.
Why, then, did Apple not use a more powerful power supply to start with? At that time--and still today to a lesser degree--switching power supplies cost more than transformers. A transformer with twice the power could be purchased for half the money of the switcher. Why, then, did Apple sacrifice power and money just to have a switcher. Switchers have definite advantages, one being that they act like circuit breakers when something goes wrong. For example, if you drop a pin on circuit board and short something out, the switcher will switch off, rather than continuing to produce current and fry the circuits on the board.
Clearly, Apple was ahead of its time. Apple was right, the Authors were wrong. Today, switchers abound in high tech electronics. You can get switching power supplies in a variety of voltages and currents. If the switcher had been so bad, they would not have grown to such prominence.
Now, on to the Mac power supply. Is it weak? Was it designed poorly? In both cases, the answer is the same; maybe, but not for the reasons cited by the authors in the trade press. Lets take the scenario from the top and analyze why these authors think the way they do, and why they are wrong.
Jim (fictitious) Johnson is hammering away at the keyboard one Sunday afternoon, trying to beat a Monday morning deadline, and smoke starts bellowing from the top left vents of his Mac. With a panic snap of the wrist, he reaches behind the smoking gun and flips the switch off. Nuts, he mumbles, Ill have to drag out the old Radio Shack TRS80 to do this article.
Monday morning; he turns in his ragged TRS80 printout and drags his Mac to the local Apple dealer. Right now, this very moment, given all these circumstances, this usually objective writer is in a very rotten mood. It wont take much to set off his short fuse. Then the word comes from the repair tech, It was your power supply, he gleefully says with dollar signs in his eyes, we replace more of these than any other thing in the Mac. It happens all the time.
Now, you (you the reader of this article) are probably mumbling to yourself, Ill bet this really doesnt happen, Loys just made up a story to put his point across. Ive got news for you buster. It happens all the time.
Most local Apple dealers have a service department to --what else?-- provide service for Apple products. They have this service department for three reasons; 1) Apple likes them to have it, 2) It is a necessary evil to selling systems (when a prospective buyer asks where he will get his $10,000 system serviced, the sales person only has to point to the service window.), and 3) It does make a few (sometimes a lot of) extra bucks.
In this service department, they usually have a service tech. However, because of the way Apple provides service support to them, they dont really have to know very much. The Mac is made up of a motherboard (digital logic board- the part that does all the computing), the internal disk drive, the infamous power supply board, the video tube, and a variety of cables for connecting everything together.
All the tech has to do to diagnose your Mac is swap the motherboard. If the computer works, it was the motherboard. If not, swap the power supply. If it then works, it was the power supply--and so on.
When he finds that the power supply was the culprit, he gleefully tells you so and collects your money.
After youre gone, he packages up your old power supply and ships it to Apple. Apple then has it repaired and ships it back to the computer store to be placed in yet another poor souls Mac.
Ok, heres the rub. This so called power supply board is more than just a power supply. It consists of the power supply and the video circuitry. Thats right, the video circuitry. The same kind of circuitry you see inside a video monitor that is used with the Apple II, and any other computer which uses a separate monitor. The problem is, when anything on this board goes out, the power supply automatically takes the rap because the board is generically called the power supply. Nothing is ever said about the video circuitry on that board.
As president of an honest to goodness service, repair and upgrade shop (we have no system sales) which specializes in Macintosh, I can tell you that 9 out of 10 power supply failures is the video circuitry--not the power supply--part of the board (the rate is dropping as explained later). Furthermore, 8 out of those 9 video circuitry failures is flyback transformer failure (This rate is also dropping for the same reason).
About 25% of the non-video circuitry failures is from cold solder joints--again, not a power supply defect. A cold solder joint is caused when not enough heat is used to solder a component onto the PC board. This usually happens in two specific types of instances; 1) When soldering a part into the PC board where the hole is part of a large foil plain, and 2) When soldering a part, which has extra large leads, onto the PC board. In both cases, more than the normal amount of heat is required in order to sufficiently heat up the larger amount of metal involved in the foil plain or large solder tail. Without sufficient heat, you get a cold solder joint.
This leaves us with about 1 out of 20 power supply board failures actually being the power supply part of the board.
Although I dont have an exact scientific study to support these figures, they do come from an estimation after actually repairing several hundred Macintosh power supply boards. We repair them for retail customers as well as for other dealers all over the country. Most Apple dealer service techs simply swap power supply boards and send them to Apple for repair--he really never knows what the failure was.
All early Macs had a small flyback transformer that would crack (probably due to heat). Ive cut failed flyback transformers open to see if I could determine what actually caused the failure. It appears that a chain of events happened: the insulation material (grey in color) surrounding the windings was heated, the heat caused the insulation material to expand, the expansion caused pressure within the epoxy casing (red), the casing cracked due to the pressure, and separation of the casing caused the winding to short, open, arc due to loss of insulation, or a combination of these. In any case, it is evident that the insulation material was sufficiently hot to flow because the material is pushed out into the newly opened crack of the epoxy housing --evidently with tremendous heat because a bubble grows on the outer plastic cover (black).
Next time your repair tech says, The power supply went out., ask him if you can see your old one. Look at the big black part (about 1 1/4" in diameter and 1 3/4" tall) and see if it has a bubble on its side. It should be cylindrical in shape without bubbles. One note here though, the lack of a bubble doesnt mean that the flyback is good, only that having a bubble means that it is definitely bad.
If heat expansion is what causes these early flybacks to go bad, clearly a fan could help keep it cool, possibly preventing the failure. I know that Apple claims a fan is not needed. Obviously it is not, if you can accept failures that maybe would not have happened if a fan had been used. However, my feeling is that no preventable failure is acceptable if the prevention is less costly than the failure. Do yourself a favor, get a fan. If noise bothers you, get a MacBreeze Piezoelectric fan.
Since our first day in business (October 1985), we have used a newly designed flyback transformer. In all the hundreds we have installed, not one has gone bad.
Apple now uses flybacks from two new sources--one good, the other better. So far, we have replaced about 15 of the good ones, and only 1 or 2 of the better ones. It just so happens that the better one is exactly the same flyback we have been using all along.
You might think that, considering we have replaced 15 of the good flybacks, it isnt so good after all. But, also considering that in the same period of time, we replaced 50 or more of the older bad types.
With the flyback problems dropping, the percentage rate of failure between the power supply and motherboard is coming into line where it should be. Before, we would have about 15 power supply (with 12 of them being flyback transformer) failures to 1 motherboard failure. Now, it is about 10 to 1 and still dropping. As more and more of the older flybacks are taken out of operation, the power supply to motherboard failure rate continues to equalize.
To heap on top of all the misunderstanding, we have scads of memory upgrades being added to Macs. Were not talking about only a few here and there, were talking about thousands upon thousands of memory upgrades. Fully 25% of all our customers have a 2 megabyte MonsterMac by Levco. Although this is probably higher than the average, I would guess that the entire Mac world probably has at least 1 out of 10 original 128K/512K Macs with more than 512K. Remember the statement earlier in the article where I said, when something is wrong, people are usually ready to throw the blame onto whatever is new. So, although the power supply (flyback transformer) would probably have gone bad without the upgrade, it is the upgrade that takes the blame for making the power supply fail.
Percentage wise, our customers with MonsterMac upgrades have no more power supply failures than our customers with 128K or 512K Macs.
The Macs power supply has plenty of power to support reasonable upgrading. As far as I know, there is only one part which has to be checked when doing a large upgrade. It seems that there was one batch of power supplies built with a weak diode at location CR17. This can be checked when adjusting the power supply. If you can get the 5 volt line up to 5.2 volts without crowbaring, the diode is okay. However, if it crowbars prematurely, it needs to be changed (It doesnt need a different value part, only a higher quality part.) Since adding any upgrade to the Mac requires a proper power supply adjustment, doing this extra check is no inconvenience.
In adjusting the power supply, you cant simply set the 5 volt line at 5 volts and forget it. It must be balanced with the 12 volt line. If your upgrader does not understand balancing the voltages, you should look for a different upgrader.
This diode and proper power supply adjustment are only two possible upgrading problems, I can think of, short of out and out over loading it. However, just because the MonsterMac can be easily powered with the Mac power supply doesnt mean that all upgrades can. Remember the first statement in this paragraph, plenty of power to support reasonable upgrading. The power supply is not limitless. No matter how powerful Apple could have designed it, there would always have to be a limit.
The point is, Apple had to give it a limit somewhere. If you design a car, you dont put a truck power train in it. It would simply raise the cost, and subsequently the suggested retail price, too high. So, when Apple designs a power supply for a computer, they design it to power what they have, what they feel they will add, plus a little extra for good measures. That was true for the Apple II, and is true for the Mac. If somebody wants to pull a trailer with their car, it is incumbent upon them to install the necessary power train and cooling system. If somebody wants to double the current requirements of the Mac, it is incumbent upon them to provide it. Apple cannot be held responsible for what others might to with their Mac.
I, for one, feel that Apple provided ample extra current. It is well within reason. It will easily support two megabytes if properly designed. Remember, the Mac was originally designed to as an appliance, not to be opened and modified. Apples only real goof-up was with the original weak flyback transformer, which they have corrected. My opinion is that had a fan been included in the Mac, they wouldnt have had nearly as many flyback problems.
Sure, Apple could play it safe, and only use proven technology like IBM does--and always be 5 years behind the current technology. But I dont think any Mac or Apple II owner really wants that. We have our Macs because they are superior to the blue in ways blue owners simply cant understand. When I find mistakes that forward thinkers make, I always remember what my daddy told me more than 30 years ago. The only people who never make mistakes are people who never do anything.
If you still believe that the Mac power supply is bad, there are several implications: you believe what the writers wrote, you will do nothing different than before, you will probably have problems in the future--with the power supply, motherboard, and everything else--, and you will continue to complain.
If what Ive written makes any sense to you, you will go out and have a fan installed (one that runs on 110VAC), youll probably have less failures, and youll be glad you own a Mac. Got a problem? Call us at (213) 941-7951 and well fix it.
Apple Computer, Inc.
After carefully sorting through and removing the clutter of my computer room at home, Ive decided that information overload has contributed to its demise; I simply have too many magazine subscriptions! Ive chosen to reduce my number of subscriptions to two, and MacTutor is my first choice.
How did I arrive at this decision? Quite simply, MacTutor is the very best and most significant Macintosh magazine to which I subscribe. It is the only magazine I read cover to cover, and I always get something valuable from it.
I read issue number one, volume one and subscribed just after that (ah, the old MacTech days....). Unfortunately, I didnt subscribe in time to receive issue one, volume one, so that is the only one Im lacking. [Number one is being sent to you, no charge. -Ed]
Please keep up the good work. Yours is the only periodical that I find myself referring to time after time. Enclosed, you will find another subscription order, as mine is nearly exhausted.
Mac Programmers Wanted
Level Five Research
Level Five Research is looking for Pascal programmers that have experience with the Macintosh. Level Five Research is an exciting, growth company in the AI field located on the east coast of central Florida. We currently have the largest installed base of expert system development tools in the world. Our products are available on PCs, Digital VAX/VMS and we will be producing a Macintosh version next. Need immediate help. Please call (306) 729-9046.
How the Mac Finds Things
In the ABCs of C column in the August 86 issue, Bob states that Lightspeed C does not include the QD structure that is used to determine the screen size. This is quite true, but does not mean that Lightspeed has a problem. Lightspeed follows Inside Mac and defines the whole group of variables as globals. The screen rectangle can be referenced by screenBits.bounds. These globals are defined in the MacTraps library, so this must be added to your project. You should also include QuickDraw.h, which declares the globals as external.
In the January 87 Basic school column, Dave Kelly notes that the MS Basic interpreter is required in addition to the new Basic compiler to compile programs. Although the two together would make a nice development environment (write and debug with the interpreter then compile it for speed), I would be very surprised if the interpreter is required to develop a compiled Bascic program. [Its not. -Ed]
Here are a few features of the 128k ROM or the Mac+ that have not been well publicized:
Poor Mans Search Path:
This appears to be the mechanism that causes the blessed folder and the root directory to be searched after the default folder. There is a function called via the HFSdispatch trap, called _SetPMSP, which can change this search path. Also FSPrivate.txt lists PMSPPtr at $386 that is a pointer to the list of directories on the PMSP. The SetPaths shareware utility, by Paul Snively, can be used to set the search path.
Sector tags on 800k floppies:
While using FEdit on an 800k floppy, I noticed that the 4 bytes of the sector tags that are defined as the time stamp of the write did not appear to be a typical time stamp. The highword was always zero. I think the 4 byte field has been redefined as the volume write count when the sector was written.
Inside Mac Volume IV says that there is a new custom clock chip in the Mac Plus that adds more non-volatile RAM. The original clock chip has 20 bytes of Parameter RAM. The new clock chip has a total of 256 bytes of PRAM. The additional memory seems to be undefined so far, except for 4 bytes that hold a validity status flag for the new memory locations. The value for the bytes that says its valid is Bugs. There are new traps called _ReadXPRam and _WriteXPRam.
The Mac Plus motherboard ROM sockets have one more address line than the 512 motherboard, so 256K worth of ROM can be installed on the Plus. The 128k ROM startup code uses this fact to determine the hardware configuration of the Mac and sets the bits in HWCfgFlag (the word at $B22). Bit 14 is set to signal that the new clock chip is present and bit 15 is set to signal that the SCSI interface is present. (See Tech Note 37)
Santa Cruz, CA.
The Feb. 1987 issue of MacTutor had a letter from Kim Hunter claiming that there was a bug in MS. FORTRAN 2.2 because it incorrectly found the SQRT(2). To quote from the manual:
Intrinsic functions contained in the math library do not follow the typing rules for user defined functions and cannot be altered with an IMPLICIT statement. The types of these functions and their arguments list definitions appear in table 9-1
Why then doesnt the compiler return on error at compile time? The answer is simple. FORTRAN passes addresses, not values. Suppose I had set I=2, and then said Y=SQRT (I) The compiler would pass the starting address of the 32 bits of I to the function. These 32 bits would now be assumed to be in real format, and would be so converted. I know of many Fortran programs that use the fact that FORTRAN passes starting address to pass integers to reals and vice versa assuming that there will be no conversion of the 32 bits where the variable is stored. This may bother people who program in other languages, but it is a feature of FORTRAN in many compilers. To had the square root of 2, therefore, conversion is necessary to a real variable.
I feel it is worth-while to warn people that as stated in the manual FORTRAN intrinsic functions must have arguments of the correct type, but I would hardly consider this a bug.
32K Segment Limit
S.C. Kim Hunter
Mission Viejo, CA
I recently started using Lightspeed Pascal and was immediately slapped on the hands for defining procedures and arrays that were too large. Digging into the manual, I find that Lightspeed Pascal has a fundamental limit of 32K for all procedures and for any data structures defined within a procedure. This is attributed to the fact that the Macintosh restricts the size of a code segment to 32K bytes stated in the Lightspeed Pascal manual on page 5-4. Somewhere else in the manual I recall a statement that the 32K limit was supposedly fundamental to the MC68000.
I was struck with the fact that nowhere in the MacFortran manual is any mention made of any limit of 32K, nor anything about segmentation. Why? Turns out that, according to Absoft, MacFortran was programmed without any use of the Macintosh segment loader. They didnt even bother to figure out how to use it. They just load the program and run it. So there is no constraint at all regarding the 32K limit imposed on the segment loader.
So my question is: How come all the other programming environments are hung with this ridiculous limit? TML is. Turbo Pascal is. MPW is. Surely Absoft is not that much smarter than all the other folks. There must be some reason the others have to use the segment loader. Hopefully someone can tell me more than just because its there. Meanwhile, Im trashing all my Pascal compilers and going back to Fortran.
[The segment loader and the other compiler products use a 16-bit offset to address the various procedures from a base address. While it is true the MC68000 only allows a 16-bit offset to a base register, obviously, this is not the only way to address the various subroutines in memory! Hence, Fortran, which was ported from another computer, can address subroutines anywhere in the 16 megs of memory of the 68000 because they dont use this particular addressing mode to keep track of things. The other companies simply took the easy route and followed Apples defective lead. Apple turned the linear addressing space of the 68000 into an Intel type partitioned memory! -Ed]
Here are some benchmarks Ive run using a simple SQRT(2.0) operation:
Number of Square Roots per Second
Apple II (VisiCalc) 0.55
HP-41 Calculator 4.3
Apple IIe (Spreadsheet) 4.7
Apple IIe (Applesoft Basic) 21.5
IBM-PC portable (1-2-3 ver. 1a) 49.0
Mac 512 (Excel) 74.0
Mac Plus (Excel) 91.0
Mac XL (Lightspeed Pascal) 245.0
Mac XL (MacFortran 00) 602.0
Mac Plus, Novy 68020/81 (Fortran /20), cache 40,667
This benchmark shows an 80,000:1 ratio over the 10 years represented by the various microcomputers!