Tape Drive

Magneto-Optical (MO)
CD Rom II

CD Rom Audio Connectors

CD-ROM Type	1	2	3	4
Toshiba	Gnd	R	L
Sony	R	Gnd	Gnd	L
Mitsumi	R	Gnd	L	Gnd
Panasonic	Gnd	L	Gnd	R
MPC	R	Gnd	Gnd	L

MPC = Multimedia-PC

>But there again, if this is more like a takka-takka-takka... sound and only occurs once per drive after cold boot/power-on then this could be the drive doing a (re)calibration (self-diagnostic) after spin-up/START from the controller BIOS.
  The same sound might re-occur after some runtime however. The process is called "thermal recalibration"  and is done to maintain the exact tracking data for the servo head (lowest head) after drive warming up.  Nothing to worry about. 
   But: if the drive makes this "glass marble bouncing on stone floor" sound of death - then there *is* a problem existing. Especially when it starts playing mad in the middle of nowhere "just so" - after being running for hours already. 
   This theme has been discussed very often in the past - and the "IBM oldtimers" like Ian Brown and me (and  several others) know this problem too good. The 320 / 400 MB drives, certain series of the 1GB 0662 and even some 2GB 0664 suffer on this "disease". It is -technically- a loss of the synchronization signal from the servo head and the heads being shot over from inner to outer stopper, trying to recalibrate and find a servo signal. That was a problem already when these drives were used in the AS/400 and the "IBM repair kit" consisted out of half the electronic boards used with the drive - and it was a hell of work changing the boards .... sometimes with no effect or with no durable effect at least. Sometimes you were just able to back up the data until the drive starts this circus once again.
    I -personally- have found out that the problem seems to be better if you use a very short SCSI cable (less  then 10"), use the drive at the end of this cable and make sure to either have the original DataMate inline terminator (grey plug between cable and drive) and install it at the end -or- install it as close as possible to the controller. I had one of the "critical" 0664 2GB-drives which often "pinged" in a Mod. 95 with 30" flat-ribbon cable - but it stays silent in a 9577 with only a 10" cable. Adapters in both cases are the IBM SCSI with cache ...!
    And: I set the "Motor Start Enable" jumper - to force starting the drive with power on and not wait for the SCSI controller to start it.
     Hope that sheds some light onto the whole mess.

Guess I should drop in here again .... :-)

>Yes, after shining a flashlight into the mounting holes of my new drive, I discovered that 2 of them had scratched paint and nice shiny metal showing thru them, leading me to conclude that the mounting screws I had used to mount the drive were too long and had apparently grounded the platter cage, causing the "marble bouncing sounds" (that's exactly how it sounded too) that Peter describes. 
      
There was even an IBM notification for the dealers that mentioned this point. Accidently some of the earlier  0661s had been delivered with too long mounting screws - which cause a lot trouble. And I had my share of trouble with these drives as well back in 1991 or 92: drives that ran perfectly sitting on the desk all alone did no longer work after installing them in the drive sled ... after some time it dawned at me that the failure occured for the first time after I checked one of the fixing screws for perfect fit ... that was exactly the quarter turn too much. Released it a bit - and voilá: drive worked again. About 6 (!) month later IBM verified my (and other field technicians) complaints and brought out a general warning, which still can be found in IBMs database.

   First of all: the 066x-family drives are sensible in many cases. They should not be installed in a 77 at the top/right position (danger of overheating). They should also have mounting screws which are *very* short - the black platter cage within the drive may not be attached to GND under no circumstance, which automatically  happens when too long screws are used in the *side* holes.
   The screws used in the holes from the underside can have any length (well - almost - if they are too long the drive will "stand" on them ...). Check this first.
    Next: use the drives one by one and make sure that you use the passive inline terminator IBM P/N 64F4774. This is a passive terminator - and the one which originally came with these drives when they were installed as first / single drives in Mod. 90 & 95. 
    These drives have rarely been used in the 77 AFAIK - the "Bermuda"-77 came with the flat black IBM drives (100 - 212MB) and the Maxtor MXT-540S or the WDS-540S, which all had a much lower profile and much  lower heat emission. The 9577-xxF models had these drives - but also had an additional air-baffle from the power supply to the top/right 3.5" bay, which helped to improve the airflow over the drive. There was an ECA  from IBM on this problem. 
    Various versions of the low level format have a known "bug": once it detects an SCSI-drive, which was already formatted with the IBM PS/2 signature it reformats not even the first blocks but only runs a sort of "data  integrity test" on the rest of the drive - even if you definitely selected "Format drive". This results in a very quick LLFormat (few seconds only) and the counter jumps from 0 to 100% in the blink of the eye before signalling "done". This is pretty annoying: the content on the drive stays intact. I had that various times with 0663 / 1GB and  0664 / 2GB drives. They could not been low level formatted with the CRTL-A advanced diags.  I hooked these drives to an Adaptec AHA-1640 and used a DOS-bootdisk which contains the ASPI4DOS.SYS and ran Adaptecs SCSIFMT on them - which took much longer but really cleared the drives.
       Nontheless: I don't think that it will help much, because the "glass marble on stone floor" noise is technically caused - from the track-following servo circuitry, which lost signal from the lower-most head. This head acts as servo-head and signals on which cylinder the drive actually is. If the signal is lost the drive tries to recalibrate with running to the inner stopper, against the outer stopper and then tries to re-read the servo informations. It it fails it repeats this procedure several times ... until the tolerance counter reaches zero and a "drive defect error code" is passed to the controller. This is the end of the way for this drive then. 
    There are two known causes for this misbehaviour: the already mentioned mounting screws, which ground the  platter cage and a cable / termination problem. This problem causes the drive to enter recalibration repeatedly due to a communication failure with the controller - to which the drives' firmware reacts with a drive reset.
    A potential third cause is "sudden power loss". The drives' aging causes a higher power draw especially on the +12V line, which is the one used for the platter motor mainly. The DC-contacts tend to oxydate or get loose (DC-plug contacts bended open). This causes temporarily power-drops on the DC lines which also causes the drive to reset itself and re-calibrate.
     Enough stuff to think about, I guess ...