Hack Acelabs and Salvation and others

[Feklar]

Ptaak:

They need to add a hardware hack to their existing systems.

A USB or serial 5 Volt line power monitor and controller. It isn't hard to design a circuit that monitors the 5V line and feeds the voltage reading to the software through the USB port, and to have the software read that value and make adjustments back out through the USB port in .01 Volt increments to vary the 5V line from 4.8V to 5.2V, the entire tolerance range. The power controller chip would regulate a 12V to 5V DC-DC converter off the 12V supply to supply the 5V to the drive.

When something like Deepspar encounters a bad sector, it can activate the power controller and cycle it through the operating range while trying to re-read, and in some cases the little less or more voltage will make a successful read possible.

Very likely, this would allow more sectors to be recovered than otherwise. Of course, I may be reinventing the wheel here, they may have already done this...

If they haven't, and they use this good advice, they can send me the new version for free and for keeps to beta test out of respect and gratitude.

[BlackST]

Power cycling is the last resort when drive is not responding anymore, but these HW tools are more and more complex, so the price.
At least one have service area emulation to get data out, and deal with internal parameters of the drive.
If you can really want to make your own *WORKING* DDI you can, but I think you have to invest a lot of time....

If you want to do some experiment, get a failing Calypso and see what happens power cycling it a lot of times or shift power supply (completely useless as the preamp and other important stuff are post-regulated)... have fun !

P.S. I suggest you NOT to do it with Hitachi or WD as the fun will end too soon. Also, if you do it on some Seagate, you'll discover that........ oooooops !

[Feklar]

I'm not saying the power should be switched off and back on, only varied from 4.8V to 5.2V.

You are right about hard drives having onboard regulation, but that regulation is still dependent to a degree on the validity of the input power.

Therefore, a 10 percent reduction or increase from 5V may only cause an actual overall 2 percent reduction or increase in the drive's circuits past the internal regulation. Nevertheless, it may be enough. The amperage can also be monitored and regulated, which might have more of an effect.

Perhaps give a warning to try to recover first without voltage and amperage regulation and save all the good sectors. Then try voltage regulation, saving those sectors which were previously unreadable onto the clone. Then try voltage and amperage regulation on the third pass and save any that had been previously unrecoverable?

For higher end drives with very expensive but very tight onboard regulation, perhaps solder point images could be published showing the regulator bypass location on the circuit board. With a warning to get all the recoverable sectors off the drive before bypassing the regulation.

[BlackST]

It may have had more sense with pre-PRML drives, not on modern ones , in my opinion.
What's worst, the onboard regulation you're talking about is not very expensive albeit "decently" tight (the power supplies on PCs are usually crap so you can imagine what can happen).
Some times ago there were speculations about this kind of experiments , I made some research and found it almost useless (nothing changed dramatically, so you can't tell if it was weak head, ECC, temperature, or external factor) to vary +/- the supply voltage.
What I have noticed is that +5/+12 supply is not critical (although you have to remain in decent range) , BER changes if there is noise / HF noise superimposed but between a certain range nothing changes at user level otherwise few PCs would work , if you consider the quality of PS. If you have a DDA or a 1 - 2 GHz scope you can display the read channel, depending on the drive family you have to find a way to trigger the visualization (otherwise, capture and see it later).
Power cycling the drive is necessary when drive is not responding anymore, and it forces recalibration, re-read the service area and flush caching if not disabled, it has some beneficial effect in CERTAIN situations.
You can do some kind of action via SW, but the way to do it is brand/family dependent so a general solution is considered (power cycle).
Of course, in case of partial failure or on certain circumstances varying the power voltage "may" have some effect but it's an exception, not the rule.
Also, on modern drives, the decoding of data / servo etc. is done almost interely thru digital signal processing (SW) so the effect of mild power regulation is counterlimited.
Instead of playing with power, due to the design of modern drives, there are many other factors that can be "tweaked" or "played with" SUCCESSFULLY to squeeze some more data out of the drive.
Final consideration : due to the cheap design of drive electronics, it is not advisable to play with +5/+12 supply voltage beyond a certain SOA. I am quite sure a good % of drives are exposed "normally" to non safe power, so the failures and the wear out.

Why vary power has little effect :

1) if servo is locked it should stay locked.
2) the preamp is usually powered at negative voltage that is meant to stay constant (it is usually derived from a DC-DC converter or from a section of the motor control) and it will stay almost constant.
3) the parameters for preamp are set via SW and there is a certain amount of auto-tuning
4) all raw data that comes from the read channel is interpreted thru DSPing.
5) rotational speed will stay almost constant as there is usually a closed loop regulation either HW and SW controlled (BTW usually the VCM/spindle is powered thru the +12 line on 3,5")
7) due to the nature of encoding of data on surfaces, small variations have almost no effect (otherwise even the slightes vibration or problem could have catastrophical effects)
below a certain threshold, internal locks on R/W subsystem trigger to avoid corruption of data as the heads are always flying on surface. I don't want to know neither I want to try on live data what happens - for ONE brand I am sure that power failures lead to "zap" (magnetically) the surface or at least do servo damage.
9) below another threshold, the MCU is reset (POR triggers). During POR the entire drive is inoperational (this is for safety and to avoid malfunction).

Today must be one of the "pro bono days" I don't usually discuss...
And now I expect also this post to go somewhere else on other people's name

[ppumkin]

Just reading the datacompass manual is says there that if you adjust the intensity value in DCEXP - that is like increases the voltage and only should be used on bad sectors not on normal recovery. Or is it just a hoax? I have never actually tested the voltage out using this feature.. but it did help me a few times when i adjusted it to max intensive.

[BlackST]

Did you try to tweak adaptives / interfacing to preamp ?

[Feklar]

Perhaps a plate tap between the HDA and the circuit board.

Imagine a spacer plate between the HDA and circuit board, say 1/2 inch thick, containing two extender headers to connect the spindle and data contacts, so the circuit board is still connected to the drive electrically, only 1/2 inch higher than it normally sits.

Then you could tap the data header, intercept the head and / or power supply lines, and apply (very) slight amplification or reduction of the signal or levels directly to the heads there, automatically cycling through the allowable ranges.

I guess they would have to sell about 30 different kinds of adapter plate... Even so, it still --> might <-- be worth examining as a thought anyway, if it means a greater percent of data recovered.

The big DR companies would probably be grateful to shell out $75 to $150 per plate and buy 30 different plates. Competitive advantage for those DR who recover for Fortune 500 companies and the like...

[BlackST]

Doesn't work that way, mister Forget analog electronics and Wile E. Coyote.

[ppumkin]

PS BlackST

Thanks for that valuable information

[BlackST]

You're welcome. The problem is that I don't know if it is correct or not

[Feklar]

Doesn't work that way, mister Forget analog electronics and Wile E. Coyote.

You are mistaken there, although the rest of what you said probably holds true in most cases.

Digital or analog, a signal is a signal. Plenty of high speed amplifiers exist (YAG transistors for example) that can accurately follow and amplify or reduce a high speed digital data signal. Also, the preamp supply voltage and current can be accessed directly that way, and reduced or amplified.

[BlackST]

Digital or analog, a signal is a signal. Plenty of high speed amplifiers exist (YAG transistors for example) that can accurately follow and amplify or reduce a high speed digital data signal. Also, the preamp supply voltage and current can be accessed directly that way, and reduced or amplified.

Good luck. Did you try it ?

[Feklar]

Dog begs dogpile.

I don't have the resources ACELabs or salvation has. I merely suggest to them a possible means of generating more income and increasing data recovery percentages.

The only potential problem would be the signal path lengths, getting them all to equal length, but any decent electronic design software can produce a printed circuit mask that does this.

[Doomer]

Then you could tap the data header, intercept the head and / or power supply lines, and apply (very) slight amplification or reduction of the signal or levels directly to the heads there, automatically cycling through the allowable ranges.

Amplification is not the only problem
There are some other basic things like timings, phase, current shift and signal-to-noise ratio

I don't think your proposal has any vital signs

Last edited by Doomer on September 28th, 2010, 12:13, edited 1 time in total.

[BlackST]

And beside the basic things, other MANY nasty things....

[Feklar]

Amplification is not the only problem
There are some other basic things like timings, phase, current shift and signal-to-noise ratio

I don't think your proposal has any vital signs

Strange, though, how i have seen with my own face, HDA to circuit board connections that were so corroded that one almost needed a jackhammer to remove the corrosion from the contacts, and yet the drive still functioned perfectly... which should have (and by definition, had to have had) greatly decimated phase levels, current shift and particularly signal-to-noise ratio. And as long as the current paths are the same lengths as each other and the length increase very small, I doubt the timings would be that much affected.

Ptaak: the servo index pulses come out the HDA connector, so where is timing the problem? They too can also be delayed by lengthening their circuit pathway by the same amount as the other lengthened pathways. The circuit board would neither know nor care. A buffer can be used in between.

(As posted elsewhere, the first thing I do when a drive is acquired, unless its new, before I ever power it, is to remove the board and clean the contacts and look for burned or blown chips... its somewhat of a lesser religion... in case you wonder why I removed a board from a perfectly working drive. That particular drive was before my new religion, and I tried its board on another drive to recover its data and discovered the corrosion.)

[guru]

Doomer has been watching "Scrubs"

"I don't think your proposal has any vital signs"

[Doomer]

Doomer has been watching "Scrubs"

"I don't think your proposal has any vital signs"

I like House MD more

[Doomer]

Strange, though, how i have seen with my own face, HDA to circuit board connections that were so corroded that one almost needed a jackhammer to remove the corrosion from the contacts, and yet the drive still functioned perfectly... which should have (and by definition, had to have had) greatly decimated phase levels, current shift and particularly signal-to-noise ratio.

and amplitude of course
but apparently not as "greatly" as you trying to impress here

[BlackST]

Drives are malicious objects made by insane minds . That's why they work even if they shouldn't