Zarko study

Question 1

With the Gate Timing MfgCmd (GTM), we learn ___.

Answer 1

where our writer is relative to our reader.

Question 2

Before anything has happened, put new SOC and PCBA on the drive, there’s some code initially that goes in so you can ___.

Answer 2

bring up hardware, and then makes it so you can do basic operations.

Question 3

The first of the basic operations is to load the heads on media and try to ___.

Answer 3

track-follow and seek.

A lot of tuning goes into even that, just to be able to hold the head on the same track fairly reliably.

Question 4

After tweaking the parameters enough that you can load and stay on track at OD, then optimize the base servo parameters so you can __.

Answer 4

stay on track at different places on the media.

Question 5

There’s the servo part of channel, there’s pre-amp, and there’s servo-tuned parameters that you need to optimize in order to reliably stay on track everywhere else.

Start by optimizing gain on ____, gain on ____, things like that, so that you can stay on track at different places on the media.

Answer 5

pre-amp side,

channel side

Question 6

There’s something in the drive called ___. It’s like a time state machine.

Because you know the rotational speed and how many servo wedges there are, you can calculate how much time it takes to get from one to another.

Answer 6

WCS - “writable control store.”

Question 7

Writable Control Store (WCS):
The ___ is found, which starts the time state machine. It asserts the servo gate so you start reading the information, it counts down the seconds, and then it knows it should be more or less be hitting the next servo wedge. It opens the servo gate, looks for the ___, and when it hits it, it starts the state machine again.

Answer 7

servo sync mark (2x)

Question 8

All of the servo information, like where the servo is, is based off of the ___.

Answer 8

reader

Question 9

So, what we do in the ___ MfgCmd is we learn where our writer is relative to our reader.

We do that by writing stuff, then reading, and learning how much we had to delay the read from the write. We learn it in the time domain, and then translate it to distance, cross-track and down-track.

Answer 9

Gate Timing Mfg Cmd (GTM)

Question 10

The reader and writer offset is called ___.

Answer 10

Mjog

Question 11

The Gate Timing MfgCmd (GTM) calibrates where the writer is relative to the reader across the stroke. Why is this important to do before ever attempting to write?

Answer 11

Because you need to not ever overwrite your servo information (the servo sync marks used in the writable control store, or WCS). That is completely forbidden.

Question 12

You can ___ after these steps:

1. Put new SOC and PCBA on the drive
2. Code goes in to do basic operations and bring up HW
3. Load heads on media and stay on track at OD
4. Optimize base servo parameters so you can stay on track at diff parts of the media
5. Optimize servo-tuned parameters to stay on track everywhere else (gain on pre-amp side, gain on channel side)
6. Gate Timing MfgCmd to know where writer is relative to reader across the stroke

Answer 12

attempt to read and write

Question 13

After you can reliably read and write and some of the parameters are optimized, then ___.

This is done in ___.

Answer 13

shoot for the highest aerial density you can write.

UFO - Unified Format Optimization.

Question 14

___ is a manufacturing command that optimizes TPI and KFCI for optimal HDD yield.

Answer 14

UFO - Unified Format Optimization

Question 15

In ___, you look for certain parameters in the read-back, and you keep trying to write more densely until you reach the threshold where if you try writing at a higher density, you’d exceed your read-back quality threshold.The output is TPI and BPI profile curves at each part of the media.

Answer 15

UFO - Unified Format Optimization

Question 16

After UFO, we have _____, which is a bit of a problem because you first optimize how densely you can write, and then you change the ___, and that may change the equation you just optimized, so you might need to change it again.

The other problem is that it takes a lot of time on 18 heads, so doing it again to re-optimize is time-consuming.

Answer 16

write current optimization,

write current

Question 17

There are tons of parameters to optimize. First we optimize them at nominal temperature. Then, we optimize them at hot temperature in ___.

Answer 17

SRST

Question 18

In SRST, we do the defect mapping, and then say, “ok, let’s ramp temperature down,” and do ____, and then we have ____.

Answer 18

MFG pack write and pack read,

Final test pack write and pack read

Question 19

DTM (Drive Test Module) is maintained by __.

Answer 19

TPE team

Question 20

There’s a ___ that goes from executing one module to the next. It loads the input information, kicks off a part of the code, and that code runs completely independently with some provisions from IPFW.

Answer 20

DTM sequencer

Question 21

MCCB, MCSB, etc. are lists of ___ for the DTM sequencer.

Answer 21

input and output parameters

Question 22

Every MFG cmd has a certain amount of memory pre-allocated for it, the ___ populates the input information and ___. The input information comes from an excel sheet that each of the MFG cmd owners populate for each of the new programs with parameters. Usually incubation team is involved. These are the knobs for that.

So, the input parameters get sent to the command, the command acts on those inputs and optimizes this way or that way based on the inputs, and then puts the values in MCSB.

Answer 22

DTM sequencer,

sends it to the cmd

Question 23

There is pre-processor, compiler, and linker, and the linker creates the executable binary.

MFG and IP are two separate and executable ___.

When you put them on the drive, it’s like executing two programs on the drive.

Answer 23

linked binaries

Question 24

What DTM does is loads the MFG binary to the ____.

Answer 24

memory area

Question 25

When MFG command needs to call an IPFW-provided function, it needs to go through the ____, and it says,

• “ok, since I’m in charge now, not IPFW, because you gave execution to me, I want to call this one thing from IPFW, so I have go and say, ‘hey, hey, I need you to do something for me,’”
• and IPFW says, “ok, tell me what,”
• and MFG sends a ___,
• and then IPFW goes and does that command.

Answer 25

special command layer,

special command description block

Question 26

___ function as an interface between MFG and IPFW.

MFG FW would have to implement too many low-level functions that IPFW already has implemented anyway for drives to operate, so why not share them and use always the right ones, the ones that IPFW supports.

Answer 26

special commands

Question 27

However, special commands can often be a problem because ___.

Answer 27

IPFW changes without MFG knowing.

There is a breakdown when IPFW changes how the function works and doesn’t tell MFG.

Question 28

You could say that IPFW, though owning special commands, also owns a part of MFG.
Servo owns both, and RSS owns both.

RSS used to mostly own MFG, but now there’s a few more things like ATI, DRP, and background scans, things like that.

It’s really __ vs. ___ in the code, and the organization is IP FW.

Answer 28

customer FW vs. MFG FW

Question 29

MFG measures a lot of things using these so-called ___ and it tries to find the min. It sweeps the curves, says, “ok, this is the ideal position, I’m going to stay here.” Many MFG commands work that way.

Answer 29

“bathtubs”

Question 30

There are parts of the reserved area where we store the ______.RA0 (he thinks).

There are other parts where we just use them as these ___, where, let’s say every 24 hours, we go and measure something on those ___ and things like that.

Answer 30

files with the data required for the drive to operate properly, 
test tracks (2x)

Question 31

The way the whole drive operates in order to preserve ____. Because we have a very tiny amount of it, we store only the information necessary for us to be able to read the reserved area, and from there we load the rest of the code, the rest of the parameters, and then we load that into DRAM, and then after that we can function.

Answer 31

flash space

Question 32

So, corruption of the ___ is extremely problematic because after that the drive cannot come up.

Answer 32

reserved area

Question 33

There’s a part of this MFG process called ___ where we optimize all this reserve area so we can finally save all the optimization data somewhere.

Answer 33

“pre-reserve”

Question 34

However, in L-H, “pre-reserve” optimizing of the reserve area so we can save optimization data somewhere wasn’t done well. They first had to optimize the full stroke, where in L-W we optimized one sliver very quickly so we could go ahead and start dumping data right there. In the infrastructure of L-H, they had to keep dumping this data to ___, and then ____ needs to give it back once it’s all optimized because there’s too much data. Now, on CCB9, that’s going to get solved.

Answer 34

the tester (2x)

Question 35

What is special about DHSMR?
Well, you need to optimize two times, once for CMR configuration, once for SMR configuration, because they actually operate differently.
___ and ___ are different between CMR and SMR.

Answer 35

TPI and BPI

Question 36

All of the output data of MFG is stored in ___

Answer 36

FIDs and RIDs

Question 37

FIDs - data that needs to be stored in ____.

RIDs - data that needs to be stored in ____.

Answer 37

flash,

the reserved area

Question 38

That’s where ____ come from. When a feature was done one way a couple years ago and the data was stored in the FIDs/RIDs one way for a certain feature set, and then there’s a new feature set you need to store more because more stuff needs to be optimized. The new code doesn’t work with the old structure of FIDs and RIDs.

Answer 38

incompatibilities

Question 39

To convert the RID structure from one to another, there’s a ___.

Answer 39

munge process

Question 40

Notches - every notch, the ___ changes, as a step function

Answer 40

BPI

Question 41

___ is on a curve, not a step function (and Zarko thinks this is a big mistake, but it is what it is)

Answer 41

TPI

Question 42

On the notch boundaries, we change ____ for different frequencies to accomplish different BPI.

Answer 42

channel settings

Question 43

These are the things that need to be updated every time we ____: channel band, AE Params, ATI/FTI ACRP curves, RRO, etc.

Answer 43

switch from one notch to another

Question 44

So when you have a seek that goes from one part of the drive to another, then you have to reload all of these ____.

We’re going from 64 to 192 notches, which means we’re going to add more data that constantly needs to be ____ in order for you to switch in runtime between these configurations, and we’re also going to spend more time in MFG drives calibrating all of this.

Answer 44

channel and preamp parameters

loaded into DRAM

Question 45

This shows the ___: what you have to do during the optimization of the media before you can dump your optimization on the media.

Answer 45

pre-reserve sequence

Question 46

OK, so what do we need in CI?
You see that this is an optimization for ___, and if you screw up one thing, it trickles all the way through.

For example, if you have incorrect ___ set at channel, then when you do all the other parameters sweep that’s fundamental, then everything else is going to be broken because ___ modifies how you see the values. So, you often don’t even know you have a bug in this. You just know you track pitch ends up being this, and you don’t know if that’s good or bad.

See Nagano-san’s email for the list of bugs in code that ended up with a significant ACC loss. So, Nagano-san was asking why we’re making integration find our bugs. Our FWTEST process sucks if it can’t find it.

Answer 46

an analog process, 
data gain (2x)

Question 47

Where are the IP FW and MFG FW binaries stored before MFG starts?

Answer 47

In L-W we used to load them directly into DRAM from the tester software. So, full FW would be in DRAM, just through loading through serial port. That’s where FW executes form normally.