what to do with a HDD motor

[Simon]

I have taken apart an old HDD (well have 4 actually), can these be used as generators at all ? the shape lends to putting belts on them if they need gearing or I can get an impeller on it with ease

[SeanB]

As they are basically a 3 phase motor with a permanent magnet rotor I would guess they would make a small generator.  As to driving them you will need to have a 3 phase bipolar output and ramp it up slowly so that the motor follows the moving field without losing lock, which will make it stall. You can use it as a generator but the output voltage would be limited to around 5VAC at a guess.

[Simon]

it has 4 terminals, is that one common and three phases ?

[SeanB]

Probably, I never was able to get them apart in any fashion that was usable afterwards. Check resistances, if 3 are the same to one terminal then star connected with star brought out. Otherwise it is an earthing for the spindle.

[kfitch42]

obviously you could use it to throw pens:
http://www.sparkfun.com/news/910

[free_electron]

HDD motors are brushless three phase motors. Permanent magnet as rotor. They are configured in star with an optional centertap. The centertap is not actuated but used to track the rotor position.

Yes you can use them as generator. That is actually what a real harddisk does as well. in case of an unexpected power loss the drive electronics isolates itself from the incoming power rail and stops the commutation of the 3 phase H-bridge. the bulk diodes of the drive fets are used as a rectifier to regenerate power from the stored energy in the spinning platters. this provides enough power to execute a safety retract of the head, either to the landing zone or into the off-platter parking garage. once everything has gone through a controlled shutdown the bridge is reactivated on 2 phases to 'brake' the spindle to a standstill as fast as possible.

[Jimmy the Squid]

Wow. I've been a computer geek since my first Tandy TRS-80 (only just recently got into the actual electronics bit), and I never knew that about HD's. I only quoted the first bit, but I'm very impressed. That is totally cool.

My first encounter with an HD was in the 80's, and it ended with me saying something like "No, I just shut off the power, what does parking the heads mean?"  Oops. I was only 12, so I guess that saved me from the full wrath of the owner of a brand-new and very expensive paperweight.

[KTP]

star wired you should be able to hook one lead up to ground on your scope and see three 120 degree separation sine waves on three scope channels when you spin the motor (if you have a 4 channel scope).

[Psi]

Since you can control the speed and they are designed to spin larger objects quite fast you could have it spin a hex mirror (like what we saw in that photocopier)

Then hit it with a laser and do some scanning

Two units together and you could do X/Y scanning

[IanB]

I don't know about HD motors, but various people have rebuilt FD and CD motors as very powerful motors for radio controlled aircraft. You can apparently create small, light motors with a rather tremendous power output:

http://www.flyelectric.ukgateway.net/motors.htm

[vk6zgo]

I don't know about HD motors, but various people have rebuilt FD and CD motors as very powerful motors for radio controlled aircraft. You can apparently create small, light motors with a rather tremendous power output:

http://www.flyelectric.ukgateway.net/motors.htm

There is an article in this month's "Silicon Chip" on doing just that,with a CD motor!

[amyk]

I don't know about HD motors, but various people have rebuilt FD and CD motors as very powerful motors for radio controlled aircraft. You can apparently create small, light motors with a rather tremendous power output:

http://www.flyelectric.ukgateway.net/motors.htm

http://www.flyelectric.ukgateway.net/croc.htm
1.3kW from a 40mm diameter, 36mm length motor

They must rely on some pretty heavy cooling during the flight.

[Uncle Vernon]

I don't know about HD motors, but various people have rebuilt FD and CD motors as very powerful motors for radio controlled aircraft. You can apparently create small, light motors with a rather tremendous power output:

http://www.flyelectric.ukgateway.net/motors.htm

There is an article in this month's "Silicon Chip" on doing just that,with a CD motor!

Interesting! I've previously managed previously to fly these motors and in fact entire drives without any form of modification. The one proviso being that no return path was ever envisaged in any of those flights.

[vk6zgo]

After the SC article,the model aircraft Geeks will grab them in mid air like a dog with a Frisbee!

[Uncle Vernon]

After the SC article,the model aircraft Geeks will grab them in mid air like a dog with a Frisbee!

Since Gillard's new Green Appeasiation Tax has hiked up the cost of a skip bin by $100 a cube, the plane geeks are welcome to all they can eat and more.

[Simon]

I try to break all my stuff down and sell the metal off and take the pcbs to the tip. But such a shame to throw the motor away

[KTP]

I wonder if there could be some way to turn a hard drive into a brushless servo motor using the existing circuitry.  I doubt though that they have 3 half bridges since they only need the motor to run one direction so you would maybe have to build the power driver.   But if you could write a bit pattern into the platters that could be read in quadrature fashion somehow (just off the wall, before morning coffee type talk) and then make a PID loop in a micro (that part is easy)...it would be very cool to build a 3 axis pcb drill small router type cnc machine using brushless servos made from hard drives :-)

[T4P]

Good luck with that 
Still figuring out how to drive a tape drive motor  it looks like it requires a shitload of 5V ...

[free_electron]

It is a full h bridge. We can commutate the 'wrong' way (active braking)
And there is a servo signal present.
One rotation ( called a mechanical cycle in hdd jargon) is divided in several electrical cycles. This depends on the number of pole pairs. Most drives have 4 , 6 or 8 pole pairs.
For every electrical cycle we have a zero-cross detection on the center tap. The floating phase ( only two phases are energized per time) is compared to the centertap. If the sign of the voltage there changes we know the rotor has passed.
A control loop predicts when this should happen and cuts the bridge temporarily. At this point the motor is a generator. A delay measurement between predicted and real gives us feedback to know if we need to speed up or slow down. So drive speed is checked several times per rotation.

As for a true positional marker : you can get the servo wedges stored on the platter.  Every block of data has a servowedge. This is a pilot carrier followed by a chunk of data, followed by the user data.
The servodata holds physical track and sector information. The pilot is a number of periods of a sinewave. This is used to lock the read pll to the speed of the drive.

If you tap into the head return signal you can see this fly by.

Your only problem doing what You imagine is the following:
The head return signal is very fast... We are talking up to 4GHz...
Writing your own pattern will not work since the drive would not be able to track ( heads dont stay in place on the track ) and the speed would be too slow for the headstack. ( heads don't work below 10MHz.. )
You could throw the system out and put an optical patten on the platters and read that. Plenty of surface, even for an absolute encoder.

But, second problem... The hdd motor is not made to run reverse. You will grind down the bearings in no time as the air bearing would not work. There is a mechanical bearing. During spinup the airflow in the motor creates the air bearing. Spin the other way and the air bearing is not created....

And the last problem will be talking to the motor controller... Those chips are proprietary and use a hdd-only protocol. Command transfer is quick with clockspeeds up to 120MHz these days. There are timeouts on this bus. Fail to send a command in the alotted time, or infrequently and the motor controller will shut down. It assumes the controller has gone wonky and will perform an emergency shutdown.

So you d have to make your own motor electronics...
Doable , but not very efficient. The motor is driven by a fast chopping pwm signal that creates a current waveform. This is NOT a sinewave... There are a few waveshapes in existence.
Multiple reasons: dissipation in the motor , physical construction of the motor poles , acoustical noise.

You can drive the motor with a sinewave but it will overheat over time.

There's a lot of technology in harddisks , and a lot of it is unknown to outsiders. A lot of the technology is under NDA as it is key to the performance and reliability of the drive.

Here's a couple of nice ones: how many head crashes does a drive incur ?.... Multiple a second. Not a problem. So where does the head crash myth come from ? The headstack can slam into the spindle or agaianst the outward stop. This causes heads to wobble and grind against the disk. The disk is coated with industrial diamond. ( diamond like carbon coating ) so very hard. Part of the ferrite material of the heads may break off and destroy the heads... This stuff gets everywhere.
And may eventually destroy other things.
Strikes against stuff on the platters happens a few times per second. No head damage there. The head does not slam into the disk , it slams against stuf lying on the platters. Only if you bang the drive mechanically can the head crash into the disk

Disk 'parking': there is a special landing zone that is made to be 'rough' using a laser . If the heads land on the platters in the smooth (data ) area the stiction (static friction) is so large the disk cannot spin up. The heads are almost glued to the platters. Try to spin up and you rip the heads off the arm.
Newer ( last 8 years) drive electronics can detect this and will not spin up.
The rough area prevents stiction from ocurring. A mechanical catch holds the heads there. Newer drives do not have a landing zone at all. They retract the heads off the platter completely and park them in a 'garage'. Simply because the landing zone can be used to store more data...

There is a lot going on in drives that people are not aware of... Like a heating element on the head to control flying height..  Or a second actuator to control the tangent of the head on the track... The heads can twist on the arm now so they are always perfectly in line with the track.

[KTP]

wow that was a lot of info about hard drives....did you work for a HD manu at some point?

I didn't know about the air bearing and reversing.

[free_electron]

i make the chips that sit on the drives

I did the preamps ( the read/write amplifier that sits on the arm with the heads ) for about 2 years. Then moved on to the controller.
Drives are incredibly complex these days in terms of 'trickery' to cram more data on the platters ( and retrieving it... storing it is easy... finding it is another matter )
To give you an idea of some the 'weird stuff' going on in drives :

- The write amplifier manages to send a 4GHz signal with 300mA of RMS current through the write coil. That is a serious RF amplifier .. especially since the little chip is barely 4 square millimeter...
- the read amplifier is gain programmable from 20db to 80dB in steps of 1 dB. The thing is gain-flat and phase constant over its full range.
- the read/write chip connects to all heads and has an on-board multiplexer. the chip is 'naked'. there is no packaging for two reasons : any bondwire or packaging 'slows down the electrical signal' due to inductive and capacitive parasites. the chip is non-traditional in the sense that it also has connections in the middle of the chip ( as opposed to only at the edge ). No packaging means lighter in weight which means the mass of the headstack is lighter which means faster seek times possible... Silicon-germanium technology.
- The reading element is a magnetoresistive element. The resistance of this element changes in function of the applied magnetic field.
- Head flying-height is tuned using a heating element. Writing causes the head to heat up. the 'lift' is subject to bernouilli's law. one of the factors is temperature. the flowing air
molecules dragged along by the spinning plattes cool the head. depending on reading or writing we change heater settings to tune the height for maximum field strength.
- The heads are so sensitive to static discharge that they contain a built in fuse to short the terminals. Once the heads are mounted on the arms , and the arms have been mated with the control chip a special command is sent to the read/write chip to blow the fuses. A verification process makes sure all electrical connections are ok and then the fuses are blown. after that a calibration step checks out the performance of each individual head and stores the optimum bias current for the reading and writing elemnts in a table for the firmware to use.

- the equivalent flying height ( scaled up ) is a Boeing 747 flying at cruise speed 1 inch off the tarmac ...
- The recording layer is not at the surface but buried under a diamond-like carbon coating.
- The platters are lubricated with a fine coating of an oil like substance. the drive platters are one of the smoothest surfaces created by man.
- While 'writing' the drive is essentially blind. The controller knows where he is , verifies the current track/sector by reading a servo wedge and confirming the positional data in it. it then moves to the next sector ( where writing will take place ) , syncs up with the servo wedge by listening an phase aligning to the pilot signal , skips the servo data block and turns on the writer. there is no feedback at this point. data is written 'blind'. writing needs to terminate on time or you corrupt the next servo wedge rendering a sector/track bad. there is A 'gap' at the end of the data block before the next servo wedge begins. the smaller the gap , the more data area you have. the smaller the gap the greater the accuracy needs to be, especially the control loops and the jitter in the servo's are important here.

A 1 terabyte harddisk contains roughly 80 gigabyte of 'lookup tables'. these tables are used by the driveware to find the optimum jump locations for the next block operation. the number of sectors changes from track to track. ( inner tracks contain fewer sectors than outer tracks ). sectors are not stored sequentially but staggered across tracks. this speeds up seek operation. Seeking is also a 'blind' operation. we know where we are ( track wise ) , calculate how far we need to travel left-right and slingshot the head there ( acceleration/decelration of the head follows a curve ) when the movement stops the entire system is tuned so that , in the time the head traveled left-right , the platter has rotated to exactly the spot where we need to be. if the process fails we need to wait a full spin of the platter to get the block to pass again. so the velocity curve of the slingshot is adapted that we arrive just in time to catch the wedge now passing .

The drive software consists of 3 things :

-the firmware: this spins up the drive, loads the heads and does all the mechatronical work. Once up to speed the firmware goes off and finds the 'driveware' and loads it.
-the driveware : this is the program runnig on the drive controller that does all sector translations and host interface. This driveware is stored in 3 to 5 locations. Should one load fail the firmware will switch to the alternate and attempt loading that.
- the drive tables : a bulk of data written during manufacturing with the optimized jump tables for the particular drive. this is specific per platter. no two drives are alike.

- a 'virgin' drive requires half a day of initialisation ( some people call this the low-level format , but this is even lower than the low level format .. without this there is no formating as there are no tracks written yet. This is a mechanical contraption that positions the heads. If you look at drive you wil see  silvery sticker on the side, about the position of the center of the platters, that covers an oval or rounded rectangular hole. If you remove the pcb you will find another one that looks like an angled or curved slit just outide of the platters and relatively close to the pivot poit of the arm. These are the openings where the drive formatter grips the headstack and inserts it's head to record the sync track. A second head is inserted during anufacturing that writes a synchro track on the outer perimeter of one of the platters. Once written this clock track is then continuously read and the info used to control the writing of the servobursts. Once the servoburst are written the mechanical interface lets go and the software takes over. It tracks the servoburst and writes the track/sector number. Servoburst+track+sector becomes the servo wedge. Writing the track+sector is the real low-level format. Writing the servoburst is below that and cannot be done using software. It is purely mechanical.

And we haven't even touched how your byte of info you want to store is mashed up , scrambled so the overal magnetic field is zero , and retrieved...
All this for 80$ per terabyte ...

and we are moving into an exciting new era of magnetical recording. SSD doesn't have a chance. it may beat us on speed , but it cannot win on density or cost.
Patterned media and HAMR open the door to drives that can store tens of terabyte per platter. There is a prototype of a laptop drive ( 2 1/2 inch ) that hits 1 Terabit per square inch of data. ... pack 3 platters in the strack and you have a laptop disk that can store 3 terabyte of user data. This was a lab-design by Fujitsu in 2007. And this is 5 years ago... you imagine where the 'lab protoypes' are today ... the drives in production today top out at 1 terabyte. because we have not begun to deploy petterned media or HAMR yet. They are the stick behind the door to beat SSD into oblivion should they impede too much.

SSD cannot overtake magnetic recording. they can't make structures small enough to pack them as close as magnetic data. even if they use mulitlevel ( more bits per cell as you store an 'analog' voltage per cell ) they are an order of magnitude away.
And there is another thing at play that people don't realise : the cloud and streaming are disruptive.

We saw a continuous decline of flash prices as the demand for size went up. around the 8 Gigabyte per chip suddenly a disruptive element entered the maket : the ipad/iphone and the concept of storing stuff in the cloud and retrieving on the fly. there is no more need for larger 'local' storage , so there is no demand for larger flash chips... so there is no development of larger flash chips.. only price erosion...
local storage is on its way out . sure there will be local flash drives of 64 to 128 gigabyte . to hold the operating system , applications and stuff that needs to be on-hand. but price will always remain high as beyond that there is no demand. people are moving away from computers with local storage. only the essenitals are locally stored . the rest is in the cloud. and that is magnetically stored.

a second effect is that this cloud is distributed. where your pictures were stored once on a drive in your computer and maybe one other drive for backup , they now reside on mulitple independent datacenters. each of your pictures is stored not in duplicate ( 1 original + 1 backup ) but acros 20 or 30 copies in 20 or 30 datacenters ... so you need 20 or 30 drives where you had only 1 before. so the worldwide demand of harddisks is actually exploding due to this storage model. flash will shrink in size to become bootup + app storage oly. the rest will live in the cloud.

[KTP]

I now feel incredibly guilty that I sometimes cut up hard drive platters to make CO2 laser mirrors (they work almost as well as gold plated first surface mirrors costing $$$).  I had no idea so much engineering went into such a cheap device.

[amyk]

I now feel incredibly guilty that I sometimes cut up hard drive platters to make CO2 laser mirrors (they work almost as well as gold plated first surface mirrors costing $$$).  I had no idea so much engineering went into such a cheap device.

Economy of scale drives the prices down. It's the same with ICs, even a little 6-pin MCU for example contains hundreds of thousands of transistors connected together.