About Palmers HDDs heads swap

[FAsTec]

Hello friends

I had this kind of info:


"translator is in some manner "linked" with the serial number of the heads

so if you will change JUST (only) the heads, you'll read ONLY x00 from all users data LBAs"


do you know something about this?

have you succesfully swapped heads on WD10SMZW or WD10SPZX (both have board full size model 800069) or on WD20SMZW ( board small size 800067 and a sticker as cover as for the Seagate Rosewood )


ALSO related with this ---> https://forum.acelaboratory.com/viewtop ... 686#p36686

Thank you

 

[Jared]

I've heard, from a reputable source, that these drives can somehow detect when a new preamp is installed and they will re-generate a new (blank) translator that will result in everything coming back as 00 unless the translator is somehow manually rebuilt.

I guess these things build the translation on the fly as they add data. So when the drive is new, it doesn't even attempt to read sectors the drive thinks are unused. Thus the reason it'll be returning all zeros if it resets itself.

 

[FAsTec]

Hello Jared

so looks like the point will turn on:
how does the "new preamp installation" is detected

maybe the heads serial number is read and compared with some record in some of the modules?

A guy here
http://forum.hddguru.com/viewtopic.php? ... 12#p271712
replied that he solved already many Palmers,

I asked some clarifications

though to discover a new translator, some record comparison has to be carried out "somewhere" otherwise it can't be aware of the substitution.

Anyway, is weird the similar behaviour

from (round about) 25％ of the capacity on, the two heads (this is a 1TB drive) read at 90MB/sec

below that, so first 500.000.000 LBAs are (just the early ones) slow , then impossible to be read. (early ones mean the ones in from there 500M side)

Also these drives like (and more than) Samsung, do have huge LBAs chains per single head
but that , at the beginning of the drive
progressively the chains lenght decreases

so it does require some calculations to guess if the two heads become slow at the same distance from the Spindle or if one is weaker than the other

... but weak ... why? If for 75％ of the surface they read at full speed??

Also it's hard to bet it is a translator issue, since if it were, it should shoot all errors and not read slowly. What do you think about this?

Some strange G-List architecture? Which causes a Slow Responding effect only in the bad sectors area?

 

[FAsTec]

Anyway head #0 looks better than #1 at least reading modules

I have spared SA Regions to #1 and it doesn't give the ID, though with some hangs it reads all the critical modules

so I have spared to #0, here things are fast and smooth, but no difference on the data LBAs

 

[Jared]

[post]14041[/post] from (round about) 25％ of the capacity on, the two heads (this is a 1TB drive) read at 90MB/sec

That's because it's not even attempting to read the platters at all in the latter 75% of the LBAs. As I said before, these drives build the translator on the fly as data is added to the drive. So if the last sector of data was written at the 25% mark, there's no translation built past that point. When a sector past where data is written is requested, the PCB will automatically return all 00 00 without ever even moving the head to that sector because it assumes it's blank.

This behavior is why the data can instantly become all zeros if the firmware gets tweaked and the translator reset.

 

[FAsTec]

WAOW that would explain the 90MB/sec of speed even in the last chain

Head #0 in the chains Around 400M and 500M does return sectors with HEX content, so they are not blank

but in the same area head #1 is already hanging

on DFL forum a kind user confirmed he swapped heads between Palmer drives ( avoiding ANY firmware action )

 

[Blizzard]

I've heard, from a reputable source, that these drives can somehow detect when a new preamp is installed and they will re-generate a new (blank) translator that will result in everything coming back as 00 unless the translator is somehow manually rebuilt.

I guess these things build the translation on the fly as they add data. So when the drive is new, it doesn't even attempt to read sectors the drive thinks are unused. Thus the reason it'll be returning all zeros if it resets itself.

I agree that SMR drives are updating the translator as new data is written, but I find it hard to believe they have developed the firmware to detect new preamps and wipe the translator. I am very interested in how they came to this conclusion. I can understand the drive having issues if there is a head compatibility problem, or some SA damage. I can also see them adding a VSC to initialize the secondary translator. Don't get me wrong, I'm not saying this isn't the case, I just can't comprehend why WD or HGST would code it to detect a parts change and then erase the translator. If your source has disassembled the firmware and seen evidence this is truly amazing, otherwise I suspect something else is going on that may have lead to a wrong conclusion.

 

[Jared]

I have no idea about the why or how, and I'm not sure anyone does at this point. All I know is, I'm playing it safe until Ace further develops for these models and can give us some reliable guidance.

 

[pclab]

I think that if WD does that (translator initialization when parts are replaced), that could be also bad for them if they want to do a recovery on their own, right? Because that would also make it impossible to be done by them....

 

[FAsTec]

That is my supposition too

But It Could Be possible that They have some log in the other modules

This would be crazy and they're the performances aspect

With locks I mean they can record records about the dynamic translator