[TriEmbed] Running old hard drive motors with a MC: DRV11873 and DRV family ICs

Thu Jul 24 09:56:15 CDT 2014

Hi Collin and List,
    I've owed you more info since the meeting, Collin. But I've been 
mired in a sea of misinformation that would be alternately hilarious and 
sad, so I'll just leave most of that out and report my conclusions after 
poking into this some more. I wanted a little more time to be a bit more 
definitive, but the executive summary starts with some very basic 
details. I realize you know most or all of these already by now and 
include it for the sake of other readers who might be interested, but 
hopefully some of this is going to be helpful to you.
     1) Without using back EMF, hall effect sensor signals, or a coil 
and magnet (or maybe some clever communication with spirits of the dead 
as mediators) you can forget about doing much of anything reliably, let 
alone spinning up to and holding a high speed. If there was a way to do 
that I'd have found it by now, having run many nested for loops varying 
a god-awful collection of parameters.
    2) Controller chips for these motors are not designed to start up 
reliably with high loads. They're designed to start up reliably with 
fans or disks or the like attached to the spindles so the power of the 
motor is to do with high speed. So the controller chip approach is 
doomed for low speed and/or high load-medium speed work without at least 
some means of external control allowing special case handling of this 
scenario.
   3) You can't just run X volts DC through the coils without 
significant cleverness: either PWM, current limiting, or kludgy 
dependence on over-temperature shutoff of the motor power source. These 
are 12v (or 5v for the 2.5 inch drives, I think) DC motors that are 
completely dependent on their winding *impedance* to run at reasonable 
current/power levels: stalled, or at very low speed there is no 
"impedance", only the DC coil resistance and they emulate tiny, enclosed 
toaster ovens. I think this aspect might be why some folks on the 
Internet insist these are AC motors.
    My theory is that the vast majority of enthusiasts get one of these 
motors, mess around for a while,  then set it aside after initial 
frustration or determining and being unwilling to implement the 
necessary drive scheme. The small fraction (but no doubt large number) 
of folks using these motors successfully only occasionally provide 
quality information, and not a single source I've found so far covers 
all the critically important bases. The vast majority of folks out there 
are too smart to try to use these things at low speed. (In my defense I 
haven't already been using stepper motors 'cause I couldn't arrange the 
mechanical coupling without having a setup that is grossly inappropriate 
for calibrating a magnetometer! But I'm rethinking that avenue now!!)
     From a standing start there is no back EMF or Hall effect signals, 
of course, so the controller chip has to use "something" to handle this 
scenario. The chip I'm using has an oscillator that sequences through 
the winding energizing order until it starts seeing the back EMF zero 
crossings. It appears the TI DRV11873 
<http://datasheet.octopart.com/DRV11873PWPR-Texas-Instruments-datasheet-13024530.pdf> 
fan controller chip you pointed to uses "sensorless, proprietary 
control" for this. With this chip there is also "lock protection", so if 
something prevents fast enough startup the chip will power off the motor 
and optionally try again later.
   But also, with this TI chip *note well* that it doesn't contain 
diodes to protect the semiconductor junctions from high back EMF 
potentials. As per page 9 of the datasheet you MUST add external 
schottkey diodes to protect the chip or the magic smoke will escape with 
"high speed motors". But here's the possible "too good to be true" 
aspect of this chip: It's in such a small package that proper thermal 
connection to the metal pad on the bottom of the chip is critical. 
Without that proper connection the chip is likely to overheat for your 
(relatively high speed, presumably 12 volt) application. So a custom PCB 
and good surface mount soldering might be a must (also per page 9 of the 
datasheet). This is relatively straight forward, but involves an hour or 
so of CAD work and about two weeks to get the boards and solder paste 
stencil for a few bucks. I or somebody at NCSU can solder a simple 
"breakout board" for this part. An alternative to this might be to bond 
a heat sink to the bottom of the chip and solder it "dead bug style". 
But those are fine pitch pins and this would require professional grade 
soldering skills and be delicate as all get out (but I know a pro who 
might be able to do it for you).  A hybrid alternative might be dead bug 
soldering and fast moving air for cooling. It would be very interesting 
to know from you or your buddies that have the physics understanding to 
say whether blowing enough room temperature air at the chip could work. 
There is pretty decent thermal info in the datasheet.
   As for the CCW/CW question, it depends entirely on the control: 
nothing in the motor itself limits it to rotation in one direction. It 
would just take an alternative set of six coil energizing sequences to 
turn the other way, but the sequence appears to be fixed for a given 
chip for the cases I've seen.
    The 555 seems like a no-brainer for the PWM with this TI chip, but 
IMO you could combine something like an ATTiny and a potentiometer 
easier, with fewer components, and with the capability to goose the 
other chip options and play with the info coming out of the chip if that 
appears interesting later.
   I have a tube of Philips TDF5140A 
<http://datasheet.octopart.com/TDF5140AP/C1%2C112-NXP-Semiconductors-datasheet-11806991.pdf> 
chips and would be happy to drop a couple off so you can bypass most of 
the horsing around exercises I've been doing with raw FETs and H-bridge 
chips and at least get to the "seeing this thing spin" state very 
quickly. These plain, 18 pin DIP chips have internal schottkey diodes 
and smart current limiting so you can connect a full 12 (or 5) volt 
supply for a given motor without worrying about overheating. The chip 
itself can run hot, though, without a heat sink. I use a 3/4x3/4" square 
copper heat sink and the whole thing seems to stay in the 30-40C range. 
They cost about $2.50 at quantity 10.
   The bottom line is that only the first drive motor I started with 
would turn slowly but reliably (with heavy load) using a series of +12 
volt pulses on it's windings (i.e. just using FETs, not half H-bridges). 
I just got super-unlucky at the start. The more plentiful Western 
Digital drive motors I've been using since then won't cooperate with 
blind control of their winding voltages. In fact they won't cooperate 
with any combo of bipolar energizings of one or two coils, except with 
the latter (using the typical sequence mentioned below) they will turn 
reliably at a single speed for a given load and pulse rate in some 
cases. But they won't reliably start up to that speed: I have to spin 
the device close to the target speed for it to reach equilibrium. And of 
course I'm talking relatively low speeds: I can't spin an 8 ounce 
assembly of aluminum to 1500 RPMs with one finger flick.
   In other words, I've found no way to just do it "blind", as so many 
Internet sources about these motors suggest, except with one motor that 
can't be removed from the disk drive frame, and I have no clue about how 
that motor might differ (if anybody knows, please enlighten us!).
   The Philips chip above is designed for three phase motors in a "star" 
winding configuration. I can confirm that with a lightly loaded motor it 
will spin the thing up very quickly and reliably, depending only on the 
values of three caps that govern the start up oscillator frequency and 
the expected min/max target speed and electromechanical forces involved. 
Selecting these capacitor values for uncommon situations is the tricky 
bit and I'm still characterizing the limits of that for low speeds.  
Page 10 of the datasheet shows the coil energizing sequence. If you use 
this sequence and reasonable cap values (the datasheet suggests ones 
that "just work" for me) and an unloaded motor doesn't immediately spin 
up, then the winding connections are wrong. For the motors I'm using the 
four pins are common, winding 1, 2, and 3 in sane order. I suspect it 
would constitute inhumane treatment of engineers for a motor to have 
other than low to high winding connection order (e.g. wired "1 3 2"). As 
many other Internet sources point out, an ohm meter will read 2X ohms 
between a pair of windings and 1X ohms between common and one winding 
and that can be used to determine which is common. If there are not four 
pins, you're on your own. :-) No, seriously, apparently "delta" 
connections are used some times and the key there would be three 
windings forming a triangle and no common (i.e. Hall effect or spirits 
of the dead used in this case). I've seen references to motors with two 
windings, but don't know how they would work and wonder if they're just 
noise.
   And I knew almost zero about brushless DC motors just a few weeks 
ago, so take all this with a lump of salt. And finally, here's the most 
useful page I've seen (found by Erik Auschhaug). The three links at the 
bottom should not be missed, but watch out for the fact that these were 
written years ago, contain proprietary (and possibly obsolete) tech 
references, etc. But the "Sensorless Startup" paragraph in page 2 of the 
first Agile paper makes clear the subtleties involved with that. And the 
second Agile paper goes into even more detail about what's really going 
on at low speeds. (The Freescale link no longer works, but suggests that 
TI and Philips/NXP are not the only vendors making these chips)

http://bldc.wikidot.com/p-esc-motor

Here's a picture of my setup using the "motor in plywood" mount Erik 
made. This picture file is about 200kb, so I couldn't inline it or it 
would exceed the 100kb per message limit for postings to this list and 
block on moderation.

http://triembed.org/blog/wp-content/uploads/2014/07/turntable3.jpg

If anybody is interested in access to the web site to upload stuff like 
this just drop me a line.

One more, possibly important detail about that Philips chip:

Digikey is discontinuing this chip and when their stock is exhausted or 
we reach September 30th they won't be for sale from them anymore. I 
suspect that relatively speaking these chips are so outdated the insides 
are gray. Also, if you just look up the part number Digikey will claim 
zero available. You have to use their part number 568-5788-5-ND to find 
them.

-Pete
(PS your Google address is now an alias for the ones you subscribed to 
the list last week and your postings using this address should go 
through from now on without waiting for moderation: no need to subscribe 
it too.)

On 07/23/2014 09:45 PM, Collin Ladd wrote:
> Hey guys,
>
> I was asking around at the last meeting about getting good RPM 
> performance out of a salvaged hard drive motor. I had to do some 
> shopping around for the lab, and I thought this might be of interest 
> to anyone with old hard drive motors sitting around.
>
> This IC <http://www.ti.com/lit/ds/symlink/drv11873.pdf>sounds almost 
> too good to be true for me, but for about 2 bucks you get back EMF 
> sensing speed control, lock sensing and a handful of other features. 
> The best part is the low parts count required--just a few caps and 
> resistors. It takes a PWM input and is pretty microcontroller 
> friendly. I will probably just use a 555.
>
> 1.5 A cont isn't going to spin up anything huge, but TI has a family 
> of higher power ones... I think (haven't had time to look around). 
>  http://www.ti.com/product/DRV11873/compare 
> <%A0http://www.ti.com/product/DRV11873/compare> << probably can find 
> one here.
>
> Depending on when I make our order, and how many I ruin during 
> prototyping, I might have a few extra to hand out at the next meeting.
>
> I read that HDD spindles are made to spin CCW only. Any truth to that?
>
> --Collin, the liquid metals guy
>
>
> _______________________________________________
> Triangle, NC Embedded Computing mailing list
> TriEmbed at triembed.org
> http://mail.triembed.org/mailman/listinfo/triembed_triembed.org
> TriEmbed web site: http://TriEmbed.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.triembed.org/pipermail/triembed_triembed.org/attachments/20140724/4b2049b4/attachment.htm>