On 7/28/2014 9:41 AM, Sean Breheny wrote: > Hi Neil, > > If you take, in this case, roughly 0.5 ohm total (high-side plus low-side= ) > resistance, 5 A and 3 Kelvin per Watt junction to case thermal impedance, > you get a temperature difference between junction and case of (5 A)^2 * > (0.5 ohm) * (3 K/W) =3D 37.5 degrees C (or K, makes no difference for a > delta-T). This means that you need a heatsink which can remove (5 A)^2*(0= ..5 > ohm)=3D12.5 Watts of heat while keeping the IC case below (150-37.5)=3D11= 2.5 > deg C. That is going to be a challenge given the small area of contact > between heatsink and the part (also, the thru-hole part would be better > because it would allow direct mounting of the metal back of the part to t= he > heatsink). The through hole part is another issue because I would like to avoid=20 components "floating" above the board due to significant vibration it=20 will endure. Instead, I want to just use discrete FETs as the RdsOn=20 would be far less -- in the 0.005-to-0.010-ohm range from a quick=20 search. Extra components for FET drivers are not an issue. > > In real life, you actually will want more margin than this for life, > reliability, and also because there will be some additional sources of he= at > generation inside the part, AND because the Rds_on will get higher, as yo= u > point out, as the temperature rises. In addition, you will want to accoun= t > for the max ambient temp being higher than 25 C. Exactly... as it heats, it gets worse. > > Where did you see the 3.5 Watt max power spec? I looked at the Infineon > datasheet and didn't see it. Oh, by "max" I meant at "max RdsOn" (350-mohm per switch), and that was=20 supposed to be 3.5V drop or 3.5Vx5A =3D 17.5W. And that's not even=20 accounting for RdsOn increase as it heats up. > In general, you MUST perform the full thermal budget calculation (like we > are doing here) when selecting a part like this. You cannot just say "the > part is rated 5 Amps continuous so I can use it at 5 Amps" because power > semiconductor datasheets are notorious for giving VERY optimistic max > continuous current ratings based on keeping the case of the part at very > close to 25 C, which would require and amazing heatsink and often liquid > cooling. So there's a back story to this, but I didn't completely select that=20 part. Someone else used it for this TB (throttle body) and said it=20 worked fine, but they only bench tested to see how much speed they could=20 get based on supply voltage. They were not an electronics person, nor=20 checked for heat, but did measure supply current and it got close to=20 5A. I've tried using a smaller throttle body to get it functional and=20 it worked fine, only getting mildly warm (no prob with my finger on for=20 a while). But on the actual vehicle with the far larger TB, things=20 (heat) escalated quickly. > In this case, it looks like they were not quite so overly optimistic and = 5 > A may be achievable but very challenging. > > You are absolutely correct that the Rds_on for this part is terrible. Is > this a very old part? Being phased out, but the replacement TLE5206-2 is really the same part=20 and specs so far appears to only be different in the logic-states that=20 control it. Apparently very common for TB control, though this is a=20 much larger TB than usual (it's over 100mm dia). I've not had this TB=20 with me to be able to experiment with it and see the effects of=20 different PWM speeds on it, but hoping to get a hold of it soon. From=20 my high-level math, I'd just veer towards a discrete-FET H-bridge, but=20 for now I'd like to see if I can tweak any settings to get it to work a=20 bit better (not overheat) until I can spin a new board with a better=20 control mechanism. Speed is not an issue, so thinking I can run it=20 slower, which should reduce the average current through it. The issue=20 here is that unlike any other closed-loop servo mechanism, this one has=20 a noticeable spring tugging on it continuously, so it takes some power=20 to keep it fixed at one position. FWIW, I have tried one of these motor controllers on the same large TB=20 and it worked fine (minimal heating)...=20 http://www.pololu.com/product/1373 . If I can adapt that to run off=20 this board, I wouldn't mind doing that. My thought is to have the PWM=20 lines generate RC-servo control PWM to feed this motor controller. All=20 I really need now is to get it functional to prove a concept and I can=20 re-design with a better motor driver later. > When you are driving a brushed DC motor, the motor looks kinda inductive, > kinda not. This is because some of the inductive energy is dumped at ever= y > commutation step (governed by the mechanical brush commutator inside the > motor). Still, I would think that the goal would be to use a PWM frequenc= y > which was several times higher than the commutation step rate, AND high > enough that the PWM period is much shorter than the L/R time constant of > the motor, so that the PWM does not cause a large current ripple but > instead operates the motor like a buck SMPS. This would be more efficient > than allowing the current to be chopped on and off completely during each > PWM cycle. Interesting... I was thinking otherwise -- ie: to switch on and off=20 completely, so the driver FETs were in-between on/off states as little=20 as possible. But then, I know little about motor control. > The above may not be possible given the frequency limitations of this > device (if it is indeed 2kHz as you suggest). This is the other doc (see the table in section 2.2.1, which has just a=20 blurb re: fmax =3D 2kHz)... http://www.infineon.com/dgdl/SE_0899_electronic_throttle_control.pdf?folder= Id=3Ddb3a304412b407950112b41714ad221d&fileId=3Ddb3a304412b407950112b41714e9= 221e > > Sean > > Thanks, -Neil. --=20 http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .