Thoughts on Swappable Drives, OSs, and UPS problems in Japan

by Todd Boyle

Lately my boss has been telling me I should consider a different line of work. Taking his advice to heart, I have decided to become a programmer. (Just kidding...)

I just formatted a blank hard disk, and installed Windows 3.1 onto it in order to support installing Win95. Went without a hitch. After installing ASPICD.SYS, ASPI8DOS.SYS, MSCDEX etc. I was able to access the CDROM to install Win95, and then installed the eight diskettes of MS Access 2.0, and finally installed Visual Basic 4.0 from the CDROM. It all went perfectly - it was truly a no-brainer.

I just made my first Hello World program in Visual Basic, which was just a screen with a text box. Starting tomorrow, for 21 days I will begin the SAMS book with the 21 day do-it-yourself course in Visual Basic. Let's see if I can make some progress. If VB is as hard to learn as C++, it's going to be veni, vidi, shelfi again....

The setup I have is one of these ¥4,000 removable SCSI brackets. The SCSI drawer is just a hunk of plastic with a couple of connectors and LED indicator. You screw your disk into the drawer and then it can be pulled out anytime. I am using these removable hard disks for experimenting with OSs. My real disk is DOS 5 running on an external SCSI enclosure (An HP 2-gig SCSI disk).

I've had some nasty experiences with operating systems, most recently with Warp going into my DOS disk and screwing it up, and then when I had booted from my DOS disk and loaded my ASPI drivers, somehow the drivers or the DOS got into the OS/2 disk which was running HPFS, and it killed my Warp desktop.

I had set up my SCSI controller not to scan the bus or initialize the SCSI ID#2 of my Warp disk. But that only stopped the SCSI controller BIOS from doing it. The ASPIDISK.SYS driver paid no attention.

A very similar thing happened earlier to my IDE drives, when I installed Warp. Imagining that my IDE drive would be safe from the install program, I went to the motherboard BIOS and entered "no drive C:" but Warp comes with drivers that supercede the BIOS and got into my DOS IDE drive and wrecked the Windows setup.

These removable drives let me install an OS like Win95 on a whim. I can install a cheap SCSI disk into one of the extra plastic drawers I have in around five minutes. If the terminators or jumpers are wrong, you can jerk the thing out, and change the jumper in 30 seconds. It's really convenient, Highly Recommended. They have IDE versions of these removable disk drawers, too. I honestly don't understand why all computers don't come with them as original equipment.

When installing Microsoft (or anything else, for that matter), I switch off the power to my external SCSI disk. That way, Mickey can't possibly get into my other drive and wreck all the non-Mickey software. Also, Mickey can't get all confused what video resolution I want, or make a database of all the software I'm using, etc.

After foolin' around with Win95, I turn off my PC, slide the Win95 SCSI out of the bay about an inch, turn on the External SCSI and turn on the PC again. Now I can play with DOS, Qmpro, Netscape, etc., and Mickey can't break them, or make me run them under Win95!!! It is physically impossible...Yippee!!

Regular readers of the TPC BBS will recall that on several occasions recently, while messing around with OS/2, my OS/2 software or installation went out and found my Microsoft partition and ate it. On other occasions the Microsoft or Adaptec ASPI8DOS driver somehow got loose and ate my OS/2 partition.

Getting tired of having my drives destroyed by these maniacal, antagonistic software companies, I concluded that the only way to run more than one operating system was to keep them on different hardware! (reading the manuals being out of the question of course....)

I took the plunge and bought a set of those removeable hard-disk holders from ATTLA in Akihabara. The drive bay was around ¥4,000 and I also bought a 2nd drive insert for ¥2,800.

Construction of the bay and the drawer appears a bit cheezy. The entire thing is plastic. The interface between the drawer and the drive bay is a Centronics SCSI-I type of connector.

When I installed the 1-GB SCSI-II drive it didn't work right, until I put this apparatus at the END of the SCSI chain. Don't even think about putting it in the *middle* of your SCSI chain, because it has several inches of ribbon cable stuffed in it, between the two connectors each mounted on a printed circuit board (one towards the disk and one towards the controller).

Anyhow it does seem to be working normally but I am worried about the heat. I left the top and bottom metal covers off the drive drawer in which the SCSI drive is mounted, and it seems to have the same airflow as it would have had before... but it is now up in the 5.25 drive bays sandwiched between a CDROM and another drive—the airflow is nowhere near what it had down in the 3.5 inch drive bays in my tower case.

Anyways, it is now physically impossible for the OS/2 system to ever see the Microsoft partition! Whew....

Using U.S. UPSs in Japan

There are all kinds of folklore and legends regarding why UPS power supplies from the US burn out prematurely in the Tokyo electrical environment. You already know this:

	Voltage	Frequency
U.S.	120	60Hz
Osaka	100	60Hz
Tokyo	100	50Hz

For many UPSs, such as American Power Conversion (APC) U.S. model Smart-UPS, the cause of burnouts is NOT the frequency or the voltage. It is the fundamental difference in design of the Tokyo electric grid, and the occurrence of nasty types of voltage spikes.

The electric power in Tokyo does not have a connection between the ground side of the 100 Volt plug and the center pin ("Earth ground") on a 3-prong plug. As a result, there can be minor differences in voltage between the ground side of the plug, and the center pin. In fact, when you have a voltage surge, it can create several varieties of voltage differences that do NOT occur during voltage surges in other countries:

1. Differences between the "Hot" power pin on the plug, vs. the center grounding pin, but not the "ground" hot pin, or the "ground" hot pin, but not the center grounding pin ("Normal mode" spikes, EMI, RFI, etc.)
2. Differences where BOTH power pins on the plug simultaneously spike outside the allowed voltage, when compared with the center earth ground pin ("common mode" spikes) but in different or opposite amounts.

These voltage differences would not be possible in the US because the neutral pin of the plug is connected to the Ground cable in your fuse box. It is virtually impossible for any big difference to arise between them. (Small differences do arise, in relation to the size of the loads around the building).

The UPS has several components for filtering the input power of against spikes, EMI, and RFI (Electromagnetic or Radio Frequency interference). You have to ask yourself, what could any component possibly do, to prevent them?

The answer is, the component can either break the circuit instantaneously, for microseconds during the input waveform, or it can attempt to short out for those microseconds, and connect those spikes somehow, to a ground.

In either case, the component is in serious danger of destruction and unless you want to pay for HUGE capacity components throughout the UPS which are quite expensive, the engineers need to know the design of the power grid, appropriate ground and other limitations of the environment to size the components and design the circuit properly.

UPSs in general are very fascinating. In case of power failure, the manual for my UPS says it kicks in within 2 to 3 milliseconds. Since a 50Hz wave is 20 milliseconds, that is around 1/10th of a wave. Not bad.

In Tokyo, the spikes cannot be drained off to the center grounding pin, which is the usual circuit design in the US model UPS's; thus they can destroy certain parts of the input filtering in the UPS. One component is called a MOV, Metal Oxide Varistor, a neat type of semiconductor which automatically decreases its resistance very quickly when the voltage exceeds a certain threshold.

MOVs are not necessarily expensive; they exist in almost all of the cheap surge-suppressors and are built into many power strips. Good OV's however, need to have high capacities, at least 150 to 200 joules, and *fast* clamping, at *low* voltages. They are less useful, if their threshhold is way up in the hundreds of volts.

MOVs disintegrate over time. Each surge they are subjected to, changes the electrical qualities of the semiconductor until it eventually fails.

Modern PCs and monitors have switching power supplies which build input power for the computer, by switching on and off very instantly during certain parts of the input wave, in order to drain off only the 5 volts or 12 volts required. This type of load confuses the heck out of a U.S. model of UPS, when there is no proper ground, because the components in the computer are shunting their waste currents off to the center pin of the power cable, which isn't grounded properly in Japan. Thus the waves travel back thru the UPS thru the fake "ground" pin, and get the input devices all screwed up, in the UPS.

But the basic failure mode seems to be this:

1. Because the design is wrong for Japan, the MOVs are subjected to frequent surges.
2. The main varistors (or smaller varistors wired in different parts of the circuit to protect against particular types of surges) lose the ability to short a surge to ground.
3. A surge comes along and either it travels into the UPS' hot pin, blasting unprotected components, or, it triggers the main varistors which shunt it to the "earth ground" pin. If the other varistors are dead, which protect against common mode surges, that creates a new surge, which travels along the fake "earth ground" and back up into every livin' component all over the circuit, which expects the ground to be the basis for measuring, shunting excess voltages or whatever.

Our U.S. version Smart-UPS 1250 at the office blew over the New Year holiday, and among other things, the surge travelled into the circuits that control the voltage switching, battery charging, etc. and scrambled the firmware in the ROM chip.

What to do:

1. If you don't have a surge protector, get one, especially if you aren't going to buy a UPS. Research indicates you shouldn't buy one if the specs are not shown, or if the capacity is less than 170 joules.
2. Don't buy a U.S. model surge protector, unless you can ascertain it has the right design for Japanese wiring. Japan does not have a bonded ground.
3. Don't buy a U.S. model UPS, even though they are half the cost.
4. If you already have a U.S. model UPS:

A UPS is particularly vulnerable during the times it is in voltage boost mode. This is a mode when the UPS senses a "Brownout" of maybe 90-95 volts which is not uncommon in Tokyo, and kicks in special circuits to increase the voltage supply to your computer or server. For this reason, and because of the additional protection against very quick spikes, your U.S. Model UPS will last longer if it has a step-up transformer in front of it.

Get a high quality surge protector designed for Japan, in front of your UPS. Expect to pay at least ¥15,000 or more.

© Algorithmica Japonica Copyright Notice: Copyright of material rests with the individual author. Articles may be reprinted by other user groups if the author and original publication are credited. Any other reproduction or use of material herein is prohibited without prior written permission from TPC. The mention of names of products without indication of Trademark or Registered Trademark status in no way implies that these products are not so protected by law.