PDA

View Full Version : Understanding how FET works...


Aurora
2008.07.11, 01:23 PM
Hi, all.

This discussion was originated from a discussion in Cristian's thread about his build in prep for Atomic World Cup. And not to hijack his thread, I would like to continue my quest and education here.

Building an AWD for the World Cup
http://www.mini-zracer.com/forums/showthread.php?p=308857#post308857

************************************************** **



In short cap, there are few things that I wonder:

1) What is the effect of mixing fets(for ex, 3010 and 4562)? CT has indicated that even mixing fets, the performance of the car will be improved.

2) Whether either the positive/negative fets will have a dominant effect on the forward motion of a car.



************************************************** **

For 2), I understand this might go against the conventional logic or understanding of how FETs work on miniz board. But local FET expert has confirmed that the positive fets have more dominant effect on the forward motion of the car, and negative fets on reverse. I don't know why this is so, but just like a fet looking on a spec sheet doesn't mean it will fair well on a track, I like to find out more too--


To (re)confirm my statement in 2), I just went back to a car with a ad band board that I have fetted and did the following experiment:



Parameters:

--using an xmod 4wd truck modded with a ko propo AD band board as a platform
--2x 4562 for the top negative fets on ad band board
--4x 4562 for the bottom positive fets on ad band board
--using Team scream matched battery
--Kyosho xpeed arm in a neo magnet can
--Tx setting will not be changed, except the 'reverse' function will be altered once
--Using the same Tamiya 4wd odometer as a tool to gauge performance


Experiment:
--Run on a Tamiya odometer for 10 sec, collect data set 1
--Rest for approx 30 sec
--Then switch motor wires (+ve wires connect to -ve on board, -ve motor wires to +ve on board), 'reverse' the Tx throttle, and do the same 10 sec run on the same odometer. collect data set 2.



Objective of experiment:
--If the either positive or negative fets have a dominant effect on 'forward' or 'backward' motion, then the reading should be different. And if there is no such 'forward/backward' fets, the reading should be the same.



Results:
--Data set 1 reads 22km/h
--Data set 2 reads 26km/h


What I am wonder here is if the data set 2 is collected after data set 1, there is no reason why the number it should jump 18%. Maybe the xmod truck is not a good platform, or the difference between 2x4562 and 4x4562 not that vast, but it somewhat valid the hypothesis that I made earlier.

Can any FET or EE expert chime in and either prove or disprove this?

ruf
2008.07.11, 01:54 PM
I'm an EE. The FETs we use in these applications are dual-channel FETs set up in a push-pull configuration. To use the terms "positive" and "negative" FETs is incorrect in a couple ways. Firstly, the correct terminology is "p-channel' and "n-channel". We screw this up a lot because we also use the terms "positive" and "negative" when referring to the motor output. Secondly, the FETs we use have both built in a single package, hence the term "dual channel". Now it is not uncommon for a dual-channel FET to have a higher performance p-channel vs. n-channel or vice versa.

Depending on how the board is laid out, our dual channel FETs could independently control "forward" and "reverse", but most applications I've seen is that the load is shared which makes sense from a power handling/heat dissipation perspective. Either way, there is usually some limiting factor in the circuit for forward or reverse: p-channel or n-channel. The interesting thing is that these live on different FET packages due to the push-pull configuration. I have seen interesting solutions where the "weak" channel FETs were stacked higher than the "strong" channel FETs to get rid of the bottleneck in forward. The way this is done is by having a taller FET stack on the left vs. right, not top vs. bottom. The way the boards are wired, top vs. bottom does not matter. Value of this staggered stack? Questionable.

One problem I see with your experiment is that you did not alter the Tx settings. The end point adjustments for forward vs. reverse could be very different. I know that they are on my radio settings.

Aurora
2008.07.11, 03:10 PM
Joe,

Thanks very much for chiming in here. :)

Would you mind elaborate more when you mention the endpoint adjustment?
To clarify a bit, I purposedly adjust both forward and backward to 100%, with no trim or subtrim, prior to beginning. And for run #1, I simply press the throttle to 100% forward. Then do all the switching and altering, and again pressing 100% throttle forward for the 2nd run.

********************************************

AD band board layout vs 2.4G board layout

The AD band board has a layout such way that the top two fets are 'stacked' together in a side by side configuration, and the bottom two fets are 'stacked' again together at the bottom.(just like you said, the burden is shared) So if I use a multimeter, I will get connectivity on the corresponding pin for the top two fets, and same goes to the two at the bottom.

The case is a little different for 2.4G, where I get connectivity on corresponding pin for the top and bottom fets closest to the left side of the car, and then the middle top and middle bottom fets are stacked together.

Just so we are on the same page...Ruf, could you confirm this with the multimeter?

************************************************** *****


So imagine this in an AM board layout, borrowing the chart from another site(Benmlee's?):

http://mini-z.home.att.net/mosfet.htm
http://img510.imageshack.us/img510/3122/minizboardix6.th.jpg (http://img510.imageshack.us/my.php?image=minizboardix6.jpg)

I believe we are looking at an AWD AM stock board. And the right side of motor connecting pad is to go to the positive of motor, and left is to negative connection on motor. From my parameters above, it will be like stacking 4 fets on the right, while stacking 2 fets on the left.


And what I trying to set out to do is after the first run--by switching the motor wires on those connecting pads, at the same time switching the TX reverse function from 'normal' to 'reverse'--it will be like swapping the 2 fets to the right while the 4 fets to the left...and given the same route of "signals" and "powers", to see whether there is any noticeable difference in performance.

One might ask why all the TX throttle reversing and motor wire switching--and why not just simply get a reading from going "reverse"? My reason being these moves are meant to avoid any potential mechanical or electronic difference if the TX is going forward vs reverse. Though the display say 'Hipoint 100%' and 'Brake 100%', I never know what really happens inside the TX. So by using same 100% hipoint for both instances, it will eliminate those chances. As for the motor wiring switching, again, it is to avoid the chance where the motor might perform differently when going backward vs forward, due to how it is constructed.


As for the result, if there is difference, then it might mean that a driver could purposedly stack different # or quality of fets on board to achieve different performance in forward or backward motion. If no, might it be some limiting factor due to the lower number of fets stacked or quality of the fets, a driver will not have any advantage to do the uneven #/quality stacking.


Man, this is one long-winded writing, and I must thank you for reading thus far. :D Also, much thanks to Cristian in another thread to continue to share and work, I too am just trying to work things out, and hopefully will be not viewed as argumentative. ;)

ruf
2008.07.11, 04:02 PM
Argument leads to progress. You should see me and Cristian in the pits at a big race! We have to challenge each other and our assumptions if we want to succeed. And if we all had the same answer, we'd never come up with anything new! :D

AD band board layout vs 2.4G board layout

The AD band board has a layout such way that the top two fets are 'stacked' together in a side by side configuration, and the bottom two fets are 'stacked' again together at the bottom.(just like you said, the burden is shared) So if I use a multimeter, I will get connectivity on the corresponding pin for the top two fets, and same goes to the two at the bottom.

The case is a little different for 2.4G, where I get connectivity on corresponding pin for the top and bottom fets closest to the left side of the car, and then the middle top and middle bottom fets are stacked together.

Just so we are on the same page...Ruf, could you confirm this with the multimeter?I don't have an AD board to check this. If this is the case with the FET orientation on the AD board, then you are correct. It's a matter of identifying the "weak" channel and stacking taller on the FET that uses the "weak" channel in forward drive. Again, people have been doing this since the MR-01 days. Theoretically it should give more punch and not add too much lag from input capacitance. I personally have never seen a lot of value in building a staggered stack, but I'm not much of a big power guy.

As for your experiment and EPA, if you are doing all the switching in the radio like you said, I don't really see an issue. As for your testbed, there are some intrinsic characteristics of your selections that may magnify results. First of all, the Xmod 4WD truck has a very draggy drivetrain which increases the current draw of the motor. Also your modified Xspeed is probably draws a good deal of current. High current, high load conditions will put more emphasis on the FETs. For the sake of the experiment, this may be a good thing.

Now as Cristian mentioned in the other thread, he actually uses the construction of the stack to modify power delivery. Think of it as modern engine management. Different tracks have different driveability issues, and making smart choices in your FET stack and battery selection can mean the difference between winning and losing. Cristian has been working on a FET tuning guide for a while, and we're hoping that it will spread some more light on this subject.

Aurora
2008.07.11, 04:54 PM
Joe,

Totally cool. I throughly enjoy the progress of thoughts, and the pleasure is all mine. :D

I myself had never a chance to experiment with the mix fet solution, and I think that is what I will try when I got time in between the point-series race hosted locally. Very interesting idea that you guys have proposed, and as said in another thread, I kinda wonder whether the difference of resistance might lead most(if not all) current to go thru the 8858s instead of 3010s?

And what is your thought on the turn-on and turn-off delay? Personally, I felt there is some difference when driving, but I am concern whether it is purely psychological.

Lastly, just a humble suggestion, I think it will really rock, if Cristian FET tuning guide could also relate or tie to motor and gearing, or even some simple track analysis, since I always think any choice of parts going onto a car should be approached by a holistic view and all these choices are inter-related. :)

ruf
2008.07.11, 07:24 PM
On resistance values are different from FET to FET, but not so much you would get really lopsided current division. For instance, an 8858 would carry 65-70% of the current if stacked with a 3010. The 3010 will always carry some of the current and have it's own timing of that current contribution.

As for the lag, it's definitely NOT psychological. I've seen some really silly FET jobs that drove awful.

As for the FET tuning guide, we'll take those suggestions into consideration. 1/10 companies have allowed for current control in thier ESCs for years now via programming, current limiting chips, etc. It's harder to do with a FET stack on a Mini-Z, but it's the same concept.

cowboysir
2008.07.11, 08:26 PM
This is an interesting toipic that i'll be keeping an eye on...I have very little knowledge of what you're talking about but I like big words.:D

One question though...when you switch the power wires on your 130 sized motor did you take into account that most machine built 130's are built/made with a reverse timing and that might account for your differences in speed?

ruf
2008.07.11, 10:06 PM
That should be negated by reversing the throttle in the transmitter. All of his testing is done in motor forward if I understand it correctly.

Honestly, LBRC and woahnelly (as well as Cristian) have a lot more practical FET experience than me. It might be nice if they chime in. I just know the theory with some application.

CristianTabush
2008.07.11, 10:21 PM
Good stuff here...

Technical stuff is not what my expertise is in. I know the practical pretty well.

In all honesty, I don't worry so much about how to get the most speed out of my car with FETs. To me FETs are used to tune power delivery as Joe has stated. And the stack numbers that I have selected allow for the most driveability and control of my car.

People that understand electronics are better suited to explain than me. I understand the concept, but I am no electronic engineer so as seen in the other forum I am no good to explain this...

I am actually learning as I do this.

So my question is then: Is the 2.4 board wired the same as the AM board and the AD board is the different one? Or are the the AD and AM boards done the same way? If the case is that the AM and 2.4 board are the same and the AD band is different, it might be that I am wrong about the 2.4ghz board then. i have to take a look at one to find out I guess...

It is good that this awoke discussion, because I believe we ALL learn from dialogue.

And yes, Joe and I often fight (ending in me throwing a tantrum and blaming Joe if something doesn't work...) in the pits. He takes a lot of crap from me!!! hehehe

ruf
2008.07.12, 02:01 AM
As I understand it from Aurora's description, the AM and the 2.4 boards are laid out in a similar fashion. The AD board is the odd ball with each "stack" on each side of the board. The AD board is the only board that can be "stagger stacked" on the same side of board.

AD: stagger top vs. bottom
AM and 2.4: stagger left vs. right

Aurora
2008.07.12, 03:12 AM
Ruf,

I think to clarify it a bit further, I will write up something with detailed description and pictures, for anyone who is following this path. :)

And basically, to elaborate a little more on what you said:

--Stacking can be done by piling FETs on top of each other or by the board layout

--AM board has no layout stacking, only two pads for soldering FETs. So to 'stack' an AM board, the FETs have to be piled on top of each other.

--AD band board. The top two FET pads are designed for a "layout stacking" side by side, so the top two FETs are 'stacked' together. The same goes to the bottom two FETs.

--2.4G board. Also has a 'layout stacking' design, but different from an AD band board. Instead of "side-by-side" layout-stacking, it is "top-and-bottom" stacked. So the two FETs(top and bottom) closest to the left side of the car are 'stacked together', while the two FETs closer to right side of car are "stacked" together.

Hope this helps. :)

LBRC
2008.07.12, 04:05 AM
...One question though...when you switch the power wires on your 130 sized motor did you take into account that most machine built 130's are built/made with a reverse timing and that might account for your differences in speed?
Unlike slot-car and other higher voltage DC motors most Mini-Z motor use zero timing, so there is no difference between running clockwise and counterclockwise.

for anyone reading who is not familiar with the “timing” reference about halfway down the page on the following link is a very good explanation. http://www.professormotor.com/FAQmotors.shtml

Hybrid example, otherwise known as the cheap stack solution.
Shows how even one 4562 can cut the On-State resistance in half and increase current handling by one amp.

http://lbspd.com/v-web/gallery/albums/album164/Hybred.jpg

Note that each of the “FETs” are actually dual channel mosfet packages each containing one P-channel and one N-channel mosfet. Like Ruf said for best possible heat dissipation to go forward it uses the P-channel from one FET and the N-channel from the other, the opposite for reverse.

You can see that the P-channels are the weak links having higher resistance and consequently lower current handling.

As Aurora said above if you want to switch which P and N pairing is used for forward and reverse all you have to do is swap the motor wires to change it’s direction then change the throttle direction on your radio.

Aurora
2008.07.12, 07:57 AM
Thanks LBRC chiming in, and a beautiful chart to vividly depict and show how things work. This actually leads to more interesting discussion of how by mixing/uneven no. stacking, people could have options of how power is delievered on the same board. Especially when manipulating the TX reverse function and motor wires switching...

To follow what I mentioned in Ruf's post, I'm not an EE person, the following is what I knew:



Terminology and the layouts:

-Just be sure we are all at the same page(anyone has EE knowledge do feel free to help to correct my terms)

+ve FET---I will call those FETs that has their 5,6,7,8 pins connected(as shown in the picture borrowed from At@micmods below) to the positive motor pad/wire as the "+ve FET"

-ve FET---those FETs that have their 5,6,7,8 pins connected to the negative motor pad/wire as the "-ve FET"

http://img411.imageshack.us/img411/1672/minizfetstacking003vz0.th.jpg (http://img411.imageshack.us/my.php?image=minizfetstacking003vz0.jpg)
We are looking at AWD AM board, and here Pin 5, 6, 7 and 8 are soldered together for these two FETs leading to the motor pads. The FETs on the right is what I called '+ve Fets' and on the left is the '-ve Fets'


Stacking---meaning the pin 1, 2, 3, 4 of those FETs are separately soldered together. Purpose of stacking is by rearranging the number/quality of FETs on board, to achieve a more desirable power delivery to the motor. The 'stacking' can be achieved by either piling the FETs on top of each other or via the board layout. The 'pile' stacking is shown in another borrowed picture from At@micmods:

http://img411.imageshack.us/img411/8742/minizfetstacking002uh7.th.jpg (http://img411.imageshack.us/my.php?image=minizfetstacking002uh7.jpg)
Notice pin 1, 2, 3, and 4 are soldered together for these two 'stacks'; since we are looking at the same AWD AM board from another direction--the pile/stack on the right is the '-ve FETs stack' while the pile/stack on left is now the '+ve FETs stack'


2x2 3004, 2x3 4562, etc--this is a common way of how people address how they 'stack' their FETs. Usually, the long number at the end is the serial number for the FETs, and numbers in front means how it is stacked.
For example, 2x2 means the total no. of FETs for the upgrade is using 4 FETs. And there are 2 pads for soldering the FETs, and on each pad there are 2 FETs piled/stacked. Hence, the 2x2. Common FETs and configuration found on AM board is the 2x1 3004 or 2x1 3010.
As for AD band or 2.4G configuration, probably it should be addressed as 4x1 3010 (4 FET pads, 1 3010 FET solder on each of those pads).



Different boards' FETs Layout:

Kyosho stock car AM board:
-2 pads for FETs
-original config is 2x1 3004 or 2x1 3010, total FETs on board = 2
-looking from the rear end of AM board, the FETs on the right side is the +ve FETs and the left side has the -ve FETs


Ko Propo AD band board
-4 pads for FETs
-original config is 4x1 3010, total FETs on board = 4
-layout-stacking, the 4x1 3010 is equivalent to a 2x2 3010
-looking from the rear end of the AWD AD band board, the two -ve FETs on top are 'stacked' by layout side-by-side, and the two +ve FETs at the bottom are stacked side-by-side


Ko Propo 2.4G board
-4 pads for FETs
-original config is 4x1 3010, total FETs on board = 4
-layout-stacking, the 4x1 3010 is equivalent to a 2x2 3010 (but layout is slightly different from AD band)
-looking from the rear end of the 2.4G board, the top and bottom +ve FETs on the left are stacked by layout; and the top and bottom -ve FETs on right are stacked by layout.

Aurora
2008.07.12, 09:55 AM
LBRC,

May I ask a few questions regarding the chart you have provided?

1) I am always confused as what data should be used for "continuous current" capacity for the FETs? I usually calculated it based on the total of P and n channel on-state resistance and power dissipation(usually is 2w, calculate by P=I^2*R)...and that number is usually smaller than the spec, but I take it as more 'real' and conservative than the spec sheet nominal number? As for 4x4562, my number will show the con't current will be 23.49A.
Similar calculation can be found here:
http://mini-z.home.att.net/mosfet.htm


2) Imagine we put in a stack of 4x4562 at where you have put the single 4562, then:

4x 4562, since they are parallel:
p-channel mosfets will have resistance = 0.0165 ohm
n-channel mosfets will have resistance = 0.0125 ohm

p-channel mosfets can handle nominal current of 24.8A
n-channel mosfets can handle nominal current of 28.4A

What will happen then? In terms of forward and backward current?

LBRC
2008.07.12, 08:32 PM
Additional variables and rounding up, data sheets are half advertisements directed at who ever is the primary customer for a particular device. Some data sheets are better than others for our purposes, some can be misleading if you miss a minor note or detail like what voltage they used for max continuous current etc.

Take the P-channel of a 4562:
Data sheet claims that at an ambient Temp of 25 C it should be able to handle - 6.2 Amps.
Further more it claims that the drain to source On-State resistance of the P-channel when the ambient temp is 25 C, gate to source voltage is -4.5 volts, and the drain current is 6.2 amps should not exceed 0.033 ohms.

Did a little practical ad hoc testing on the 4562 a couple of years ago and found that the performance of a single 2x1 layer on a MR-02 CB was for all practical purposes on par with the claims made by the spec sheet.

You can use ohms law and this information to compute additional useful info but by itself ohms law does not pull unknown values out of a hat.

http://lbspd.com/v-web/gallery/albums/album101/HOMSLAW.sized.gif

For example; if our motor was drawing 6.2 amps and the drain to source resistance is 0.033Ω then we know from ohms law that:
• The drain to source voltage loss across the P-channel will be 0.2046 volts.
• And that 1.26852 Watts of power will be lost (as heat) due to that internal 0.033 ohms of resistance.

Now let’s compare the 3010, but right off the bat there’s a problem when you look at the data sheet, we find that some of the test parameters are different as well as the way they list the specs so it has to read a little differently when the P channel data is spelled out:

Data sheet claims that at an ambient Temp of 25 C it should be able to handle - 5 Amps.
Further more it claims that the drain to source On-State resistance of the P-channel when the ambient temp is 25 C, the gate to source voltage is -4.5 volts, and the drain current is 2.5 amps (1/2 of the -5A maximum load) should not exceed a maximum of 0.130 ohms.

Ok -2.5 amps is a far cry from the max -5 amps or the 4562's -6.2 amps but it’s not as bad as it may seem when you look at the 3010 data sheets P-channel graphs and notice that the on-state resistance is still nominally in the 0.1Ω ball park at -6 amps.

• The drain to source voltage loss across the P-channel will be 0.325 volts.
• And that 0.8125 Watts of power will be lost (as heat) due to that internal 0.130 ohms of resistance.

But what happens if we use the same 6.2 amp motor in the 4562 example and apply ohms law to the 3010’s 0.130 ohm max On-state resistance?
• The drain to source voltage loss across the P-channel should be 0.806 volts.
• And that 4.9972 Watts of power will be lost (as heat) due to that internal 0.130 ohms of resistance,

Ouch, HOT!!! 2.9972 Watts over the 2 Watt limit on the 3010’s data sheet. But wait, ohms law can tell us something I’d like to know; that 0.130 ohm resistance was a maximum so what would it have to be to get under the 2 watt limit? So Power in watts divided by Current squared equals 0.052 ohms, interesting, but then what abut the 3010’s claimed 5amp limit so ok P=2W and I=5A and gee 0.080 ohms very close to the listed 0.090 ohm Rds-On but still not right, or is it? Yes when I look again the -10volt Max RDS-On value on the 3010 data sheet is listed at 0.076 ohms and they didn’t specify a voltage when listing the max current did they?

But what about stacking?
When you solder one FET on top of another resistance is cut in half (that’s good) and current handling is doubled (even better) or at least that’s what would happen if nothing else changed, but alas when we stacked our FETs things do a change. If you look at the data sheets the 25 C ambient temperature 2W power dissipation claim for both FETS are based on them being surface mounted on a circuit board to help dissipate he heat, in a 2x2 stack effectively the top FET is sitting on a heater while the bottom has an electric blanket covering it so ambient temperature and power dissipation change drastically, and talk about huddling together for warmth what about the middle FETs 2x3-plus stacks. Which is why the 4x1 will handle more current than a 2x2 and a plosable reason why the stock ASF CB’s are handle more current and are faster than 2x2 3010’s.

Note that the data sheets typically list two Rds-On values one for and ambient temp of 25 C (77 F) and the other for 70 C (158 F) when you stack for all practical purposes you might as well go ahead and use the 70 C values, while remembering that the effective temp could be much more.

…Imagine we put in a stack of 4x4562 at where you have put the single 4562, then:

4x 4562, since they are parallel:
p-channel mosfets will have resistance = 0.0165 ohm
n-channel mosfets will have resistance = 0.0125 ohm

p-channel mosfets can handle nominal current of 24.8A
n-channel mosfets can handle nominal current of 28.4A

What will happen then? In terms of forward and backward current?

It sounds like you are forgetting that the mosfets work in pairs, regardless of the direction, each circuit needs to use a P-channel from one FET and the N-channel form the other FET so the direction doesn’t matter.

Two stacks of 4 (4x2) are required with 8 total mosfets on the ESC portion of the CB:

For resistance 0.033 ohms for each FET divided by 4 parallel mosfets = 0.00825 ohms for the P channel plus 0.025 ohms divided by 4 parallel mosfets or 0.00625 ohms for the N-channel, which gives you a theoretical total resistance of 0.0145 ohms for the circuit (P+N).

For current 4 FETs in parallel x 6.2 P-channel amps each equals a total 24.8 amps for either direction. The 28.4 amps for the N-channel is not a factor because the weaker P-channel for the circuit will always go first.

Except that the heat generated by 4 stacked mosfets changes the resistance values and current handling considerably. Not to mention the input signal, you are now pushing 4 devices with one driver.

Aurora
2008.07.12, 10:22 PM
Wow! Awesome material! Thank you so much!

Gotta head out the door soon, so I probably have to come back and re-read the first half to appreciate it... :p

But the quick take away is that if I were to have a funky setup such as 4x4562 on one side + 1 3010, without taken into account of temp or the input signal or other variables, the FETs can provide 24.8 amp for the motor to go forward and backward. But this is one extreme example, and there sounds like a lot of loose ends and questions regarding the actual performance.


One thing I do observe is if I were to stack an AD band board two 3010 on top, and then four 4562 at bottom, when ever the car crash and to need to go reverse, it has a very severe lag before it can go backward. The forward is never an issue...And if I were to bring the top and bottom FETs closer in number and/or quality, the lag will be reduced or even eliminated. That do seems to confirm your comments on potential issues to deal when having an uneven stack of 4x4562 + 1x3010.

Usually, I have seen the local FET expert replacing the bottom 2 FETs with two 4562s or three 4562s(or equivalent) and leaving the top two 3010s alone , and that seems to jive quite well.

LBRC
2008.07.13, 02:16 AM
...Usually, I have seen the local FET expert replacing the bottom 2 FETs with two 4562s or three 4562s(or equivalent) and leaving the top two 3010s alone , and that seems to jive quite well.

So lets look at what you get with that on an AD band board asuming that your right and one motor pad goes to the top two FETs while the other pad connects to the bottom two.

2p4562 x 2p3010 (no stacked FETs)
Direction 1 = 0.039 ohms, 12 amps
Direction 2 = 0.0775 ohms, 10 amps

Not bad with direction 1 being forward, and direction 2 set for reverse, especially since you hardly run around WOT in reverse for long so 10 amps should be more than sufficient no matter what motor you are running.

Now we add that third 4562 to the top of one of the 4562’s.

2+1p4562 x 2p3010 (one 2 FET 4562 stack and one single 4562 in parallel x 2 single layer 3010’s)
Direction 1 = ~0.037 ohms, 12 amps
Direction 2 = ~0.077 ohms, 10 amps

Wow that’s not much of an improvement. No surprise with the current it wasn’t going to change because the 2x3010 are the weak link. But I bet you didn’t expect almost no resistance improvement did you. Ok so the resistance values are estimates due to the two stacked 4562’s but it’s probably not to far off the mark and there is even a good chance that if they are running warm it could be worse than just 2 single layer parallel 4562’s. Disappointed? Ok wait just a minute, these are only maximum values that assume high motor current levels, what if the FETs are not getting hot? Then you could have an Rds-On as low as ~0.026 ohms for direction 1 until they start getting hot. Either way IMO you’d be better off adding the third 4562 to the top of one of the 3010’s.

The following is a summery of how I estimated the 2 Stacked +1 not stacked parallel 4562 values:

I use the 70C temp current and power dissipation values for the 2 stacked 4562’s to adjust for the expected loss of heat dissipation. Normally I wouldn’t bother with all this when comparing similar stacks of different FETs but here we are talking about comparing parallel single layer performance to stacks so IMO its time to plunge into the heat issue even if its less precise and the math makes the old head hurt.

For the two stacked P-channel FETs
Current = 9.8 amps for the stack (4.9 amps per FET)
Maximum Power Dissipation = at 70C is1.3 watts
RDS-On = 0.0255 ohms (1.3W/4.9A^2=0.051Ω per FET,) see ohms law is useful :)
Same process for the N-channel.

Added to the single un-stacked 4562 for the totals:

3x4562 P-channel total:
Current = 16 amps (9.8A+6.2A)
Rds-On= 0.01438 ohms (1/Rtotal = 1/.0255 + 1/0.033)

3x4562 N-channel total:
Current = 18.9 amps (N/A)
Rds-On= 0.01111 ohms

The 2x3010 is then easy
P-Current = 10 amps
P-Rds-On = 0.065 ohms
N-Current = 12 amp
N-Rds-On = 0.0225

And that’s how I ended up with this rounded up for the above:
Direction 1
Total P+N Rds-On = .03688 ohms
Max Current = 12 amps (2x3010 N-channel value)

Direction 2
Total P+N Rds-On = 0.07611 ohms
Max Current = 10 amps (2x3010 P-channel value)

Aurora
2008.07.16, 05:40 AM
Finally get a chance to sit down and read this post throughly, LBRC, can't thank you enough to make all this calculation straight forward for a layman like me!



To summarize things I have learned:

1) When considering how to stack, beyond of merely looking at the spec sheets, there are many things to be bared in mind--the layout of the board, 'weak' link of the stacking solution, temperature, actual performance of the FETs used, etc

2) When considering 'cheap' stack solution for those 2x2 layout boards(AD band or 2.4G), replacing two FETs linked in parallel (either +ve or -ve FETs) with better performing FETs, might improve the performance quite well. But additional FETs add on top of those high performing FETs might only result in marginal improvement.

3) Track performance rules.



LBRC or any onboard FET expert,

One last thing--I notice there are quite some difference between FETs used when concerning how linear is the throttle response.

Does the linear feel more or less relayed to the turn-on delay and turn-off delay? What will happen if we have 3010+4562 for 1 direction, and 3010+4562 the other direction?

Based on what I learned from you regarding the calculation, it should now not be difficult to calculated the math. (thank you!) But linear response could be another story...

ProfoxCG
2009.01.16, 03:59 PM
from what I am gathering,

If I wanted to do a FET upgrade for FORWARD only, Then I could leave the
-ve 3001 fets from the factory alone, and only replace +ve fet with something more robbust, 2X2 4562, on an am board,

on a 2.4 board, I can get away with replacing the +ve fets only (left side of the board when installed in the car?

http://i10.photobucket.com/albums/a146/Profoxcg/Mini-z/FETQUESTIONFORRICHARD.jpg

LBRC
2009.01.16, 08:09 PM
from what I am gathering,

If I wanted to do a FET upgrade for FORWARD only, Then I could leave the
-ve 3001 fet from the factory alone, and only replace +ve fet with something more robbust, 1X1 4562, on an am board,

on a 2.4 board, I can get away with replacing the -ve fets only (right side of the board when installed in the car?

Yes but… each “FET” is actually a P+N dual channel mosfet and to go in any one direction you need the P from one FET and N from the other so one 3010 and one 4562 only increases the current handling ability from 5A to 6A, an improvement but not dramatic. However if you’re referring to an older Mini-Z with the 3004 FETs, you go from 3.5 amps to 5.5 amps and 0.45 ohms to 0.143 ohms which can make a major difference when trying to use some of the hotter “stock class” motors.

http://lbspd.com/v-web/gallery/albums/album164/Hybred.jpg

http://mini-zracer.com/forums/attachment.php?attachmentid=26512&d=1232157947


On a 2.4GHz ASF CB the stock non-stacked parallel (2x2) 3010’s already give you 10 amps and replacing just the forward P-side FETs with 4562’s would only increase the current handling to 12 amps, it would be better and considerably easier to leave all 4 stock FETs in place and stack the two 4562’s on the two top side 3010’s for a 3x3 equivalent.

ProfoxCG
2009.01.16, 10:44 PM
well i have an ASF board which I have already replaced the FETS with 4562's, however i want to add another layer to the top of the board (I guess top or bottom do not matter)

- If i want to do this, I need to remove the already installed 2X1 , stack them and re-solder to the board correct?

I have another question:

Why is it that car with upgraded FET's doesnt run as long as car with stock FET's? what is exactly happening? I want to understand this so that i know where to stop "fetting" and be able to run a 15-20 main and have power untill the end. I am running Orions 950 with w (2) 2X1 4562 ASF and I feel the car has alot of punch for a few seconds (one min) then the punch drops and the car begins to feel slower until it just dumps eventually.

LBRC
2009.01.17, 12:57 AM
well i have an ASF board which I have already replaced the FETS with 4562's, however i want to add another layer to the top of the board (I guess top or bottom do not matter)...

On the ASF board the top and bottom FETs are in parallel just as if the bottom FET was stacked on top of its corresponding top FET for a 2x2 with better heat dissipation.

- If i want to do this, I need to remove the already installed 2X1 , stack them and re-solder to the board correct?...

Not necessarily, for example it’s relatively easy to stack two 8858’s right on top of the top two stock ASF 3010’s for a surprisingly nice 3x3 without even removing the circuit board from the car.

...I have another question:

Why is it that car with upgraded FET's doesnt run as long as car with stock FET's? what is exactly happening? I want to understand this so that i know where to stop "fetting" and be able to run a 15-20 main and have power untill the end. I am running Orions 950 with w (2) 2X1 4562 ASF and I feel the car has alot of punch for a few seconds (one min) then the punch drops and the car begins to feel slower until it just dumps eventually.

Providing all the solder joints are clean and no damage was done to the FETs in the soldering process a 2x2 will have a lower on state resistance than a 1x1 of the same FETs, end result less voltage drop across the FETs, motor draws more current and the car goes faster, but the flipside is that even though the FETs are more efficient you don’t escape the fact that drawing more current and driving faster also drains the battery faster. However if you run two identical cars side by side at the exact same speed one with a 2x2 and the other with a 1x1, providing again that the solder joints are clean and the FETs are undamaged, the 2x2 would then run longer than the 1x1 with the same mosfets. Bottom line (with a few exceptions like improving rolling resistance) is that when a car goes faster it drains the batteries faster.

That’s in a perfect world but unfortunately most of use live in the real world where we also have to deal with the possibility of static and or heat damage. Mosfets are perfectly capable of operating with minor static or heat damage to one or more of the FETs in a stack, just not as efficiently as undamaged FETs, often a stack with minor damage will still handle more than enough current to run even a modified motor but you might notice that the stacks are getting hot and running one or more power hungry electric-Z-heaters along with your motor can empty your batteries fast.

ProfoxCG
2009.01.17, 09:35 AM
so basically i always need to add or upgrade FETs in pairs, because they both control forward and reverse?

its funny though that sometimes my car seems more powereful in reverse.

LBRC
2009.01.17, 04:03 PM
so basically i always need to add or upgrade FETs in pairs, because they both control forward and reverse?.

Yes


its funny though that sometimes my car seems more powereful in reverse.

That can be caused by a degraded FET but it’s pretty easy to check out, you can change which FET pair is being used for forward and reverse by swapping the motor wires and changing the transmitter throttle direction, if you still have the same problem it’s not the FETs.

ProfoxCG
2009.01.17, 05:46 PM
...if you still have the same problem its not the FETs.

what do you mean, believe it or not this happens on my stock ASF board for my MR02.

EMU
2009.01.18, 03:09 AM
Meaning, that if the motor is still faster in reverse after switching the wires on the motor terminal, and switching the direction on the transmitter... that it is the motor that is faster in reverse, which usually means the motor timing is retarded fwd and advanced rev.

Most of my motors are faster in reverse when I have the reverse EPA equal to the forward EPA on the transmitter... I lower the reverse EPA on my transmitter to have less brake effect and slower reverse. One thing I cant stand is going into reverse to back off a corner and backing out too far into traffic (its mainly for smooth braking though). I also have a button on my transmitter that I set to full reverse, in case I need to handbrake/reverse hard out of a wall on a bank or with a motor that has a lot of drag.

ProfoxCG
2009.01.18, 11:07 AM
gotcha. =)

MINIz guy11
2009.01.18, 09:28 PM
Wow, this thread is nice but I also have a question about FETs and their use with the Brake function.

I recently installed a set of 4501 FETs on a spare board. I get foward fine, revese takes a longer time to get up to speed, though. I also don't get the brake function all the time, as after I get off foward throttle and wait to apply brakes, I get reverse!

Exact steps I'm taking:
Foward Throttle on -> Foward Throttle off -> Wait 10 seconds -> Apply brake but get reverse (slowly gets up to speed)

Is there something wrong with the FETs itself causing this problem?

Edit: By the way, these are the Vishay 4501's, not the Fairchild ones.

ProfoxCG
2009.02.04, 11:01 PM
I need some help comparing,

I know from reading this thread that RDS @4.5V is one of the values we are looking for in terms of a "better" fet. The lower the better. This is the value the get cut in 1/2 when you start to stack.


What other piece of information is important?

Also, lets say I have 2X1 then the DRS for the P-chanel is .03XX for each FET STACK so when I add another FET on top to each stack, I take 1/2 * 0.3xx per stack. this equal 0.015xx... at this point is I add a 3 fet, whats the math involved then to get the RDS of my stack?

so far looking at specs, the FDS8885CZ is comparable to a Si4562Dy in term os RDS.

and the FSD8962C seem a bit inferior...
@4.5V 0.044 N-Channel
@4.5V 0.080 P-Channel

MikeL
2009.06.13, 06:20 PM
Well once again I should have found this thread before doing some work, lol. I replaced a 2.4g 3010's with 4562's, I'm trying to decipher if I should bridge all 3 pins on the output side, or just 2 (I have just 2 done on all 3 at the moment). I can see 2 traces in the board I think, so I was worried about bridging all 4. Can you bridge any of the input pins, I would think not. What's the hot fet of the day this year?

So far it feels smoother than stock, not sure if it's really any faster though.

Skv012a
2009.06.13, 09:15 PM
Would anyone happent to have diagrams with FET pad traces on the board or know where I could trace the legs if the pads get damaged?