Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

What is the best heat transfer method

Status
Not open for further replies.
As requested and more, in the .jpg, the red circle are around failed items, the red arrow is the shoulder washer that melts. I forgot to take a photo after I replaced all the FETs . The 2 .1 ohm resistors were replaced with a 50 watt .1 ohm resistor, similar but larger. I doctored the .asc to help you understand the file. You won't have some of the components in your library, so you will have to add them.
The heat sink shown goes in an aluminum box that has fins mounted to it.
Spice say my frequency is 11kHz, I was thinking it was closer to 20, I pull it out and measure and photo it, may take a day or 2 tho.
EDIT, made a few changes to the .asc file
 

Attachments

  • 2014-01-13 PWR STEERING PHOTO + SCHEMATIC.png
    2014-01-13 PWR STEERING PHOTO + SCHEMATIC.png
    3.9 MB · Views: 238
  • 20140115_110254p.jpg
    20140115_110254p.jpg
    508.6 KB · Views: 206
  • POWER STEERING 2014-01-09 POST .asc
    14.2 KB · Views: 155
Last edited:
The IRF3205 has a gate charge of 146nC, which is fairly large. That capacitance driven from your 2K pull up resistor is a very slow combination. So your mosfets are spending to much time in their linear mode, not fully on or off. This makes them run hotter than they would if the turn on time was much faster.

You really need a higher current to driver the gates up and down more quickly. You can make one with a couple of BJTs, or use a gate driver IC. I'm using a Fairchild FAN7842 in a product of mine. A pair of those would do well for your application. Also, since they have their own charge pump, you don't need to generate the +25V for the highside mosfets.
 
According to the Data Sheet, The IRF3205 Requires 10 Volts Gate to Drain, to Fully Turn On.
Since your Supply Voltage is Just 10 Volts, It is Unlikely the LT1017 will deliver that 10 Volts.
So you Won't get a FULL TURN-ON, on the Mosfet.

Also you should have a Gate Resistor of about 75 Ohms between the LT1017 and the Gate of the Mosfet for best performance.
 
Hi,

After reviewing the other replies and your last reply to my previous reply i have come to the same conclusion as other readers here. That is that the gate is not being driven correctly. To get the MOSFET in a PWM application to dissipate low power you need at least two things: 1. Enough gate voltage, and 2. Enough gate current.

Most of the time we have the correct gate voltage (5 or 10 volts or whatever) but many times the current is not high enough to turn the MOSFET on and off fast enough. IF the MOSFET switches too slow then the power dissipation goes wayyyy up and heats up the transistor too much. That means a good gate driver scheme is required. This could be a MOSFET driver IC chip which can put out 1 amp or more or sometimes we can get away with a two transistor driver using one PNP and one NPN transistor both as voltage followers with the emitters tied together.

The switching MOSFET is at it's best when it is either fully on or fully off, because both of these states means very low power is being dissipated in the device itself. If fully on it passes lots of current but has little voltage drop, and if fully off it has lots of voltage across it but little current through it, and both of these states means the power P=I*E is low because either I or E is low.

The problem is that when the MOSFET first starts to turn on the drain provides some feedback to the gate through the gate-drain capacitance, and that acts to decrease the gate drive voltage (negative feedback) which looks like a tiny 'step' in the voltage when it should be going straight up to the max gate voltage. During this time the MOSFET is neither on nor off completely and that means lots of power heating. To make this time as short as possible a low impedance drive is necessary and this usually has to come from an active driver circuit. IC's have been designed just for this purpose and so that is a pretty good idea, or you can try the dual transistor follower scheme to see if that works ok first.

Another interesting thing to note is that the gate does not take too long to pull high (or low) when there is enough drive current available. That means the driver chip only has to put out a high current for the time that i takes to actually change the state of the gate. After that, the driver only has to put out a small current (perhaps microamps) to keep the gate at that state. So for say a 1ms gate pulse the higher 1 amp current might only flow for 10 to 100us and after that the driver just has to 'keep' the gate high and that takes little or no current.
If we were to look at this current profile we would see the following sequence:
1. Gate driver initiates the high pulse.
2. Gate driver output starts to ramp high very fast putting out 1 amp.
3. Gate driver max voltage output is reached.
4. Gate driver output current goes down to very low value, microamps or less.
5. Gate driver output stays high for the duration of the pulse.

After that the gate driver will initiate a low going output pulse but that's the same thing except the gate voltage is falling instead of rising.
 
Last edited:
Gate Current is NOT an Issue for Mosfets.
Typical Gate Current on Mosfets is MUCH LESS than 1 uA.
But a GOOD Waveform, with a good Rise and Fall is ESSENTIAL.

DOES he Actually Have a GOOD SQUARE WAVE, Or is he driving with a TRIANGLE Wave?
A Triangle Wave will Create a LOT OF HEAT in the Mosfets.

I usually just drive my Mosfets wih a 555 and Never have any Problems.
Even at Quite High Frequencies.
 
Well spoke Mr Al and I agree, I shall add totems first to the PWM, low side, high side is strictly an on off affair while the low side does the PWM and it is the low side FETs that have failed.
A minor point is that Spice works better with models from Linear than some of the other sources, so I use models supplied with Spice, the LT1017 is actually an LM2901N
chemelec, the triangle wave mid point is varied by mixing with the input from the Hall and is compared to one of 2 set points to produce the on off times for the PWM, the out put of the LM2901N comparator is the source of the square wave. Do you have an explanation of how you use the 555 as a driver?
 
First, What is the Purpose of your Circuit?

Is it Just to control the Speed and Direction of Rotation of the Motor?
 
don't forget the gate drive for your top fet needs to be driven 10v or more above the supply voltage.
I find the diagram a bit confusing. but it seems to show the fet gate drive only reaching a max of vsupply
maybe use something like an hip4081 a bit pricy but very good at this kind of thing

ps use physically larger fets if you can because it's easier to get heat out of them, paralleling them also works very well as your I2R losses overall are halved and the power is spread across 2 devices. much easyer.

one psu I made had 24 fets in parallel at high frequencies and it works just fine ;)
 
Chemelc, Yes, this circuit is for a power steering assist and at rest and normal straight driving, the motor is turned into a dynamic brake to resist bump steering.
Another tidbit of information is probably of little concern is that when the FETs were destroyed, I was testing them to see if they would fail, the OHV has 4 wheel drive and the differentials can be locked and when turning, there is a terrific bind and this causes the steering to try to go straight and requires a lot of power to counter act. Turning the wheels to full left or right is virtually impossible on pavement, but on sand or dirt, just hard.
richardb, the 555 in the schematic is driven by the oscillator via the trigger input and out drives a boost circuit for 25v for the upper FETs, it has a feed back to CV to control the voltage.
While checking the spec sheets of the LM2901n and LT1017, I noticed this application,(see attached) and I will modify my circuit to include it.
 

Attachments

  • Untitled.png
    Untitled.png
    16.6 KB · Views: 178
Last edited:
Thank you everyone, after I get my changes done and the OHV runing again, I'm replacing a few things and adjust toe in, I'll put a temp sensor on it and give it another test and post the results.
 
Chemelc, Yes, this circuit is for a power steering assist and at rest and normal straight driving, the motor is turned into a dynamic brake to resist bump steering.
Another tidbit of information is probably of little concern is that when the FETs were destroyed, I was testing them to see if they would fail, the OHV has 4 wheel drive and the differentials can be locked and when turning, there is a terrific bind and this causes the steering to try to go straight and requires a lot of power to counter act. Turning the wheels to full left or right is virtually impossible on pavement, but on sand or dirt, just hard.
richardb, the 555 in the schematic is driven by the oscillator via the trigger input and out drives a boost circuit for 25v for the upper FETs, it has a feed back to CV to control the voltage.
While checking the spec sheets of the LM2901n and LT1017, I noticed this application,(see attached) and I will modify my circuit to include it.

OK, But V+ Need to be Higher than 10 Volts.
Transistors,typically have a 0.6 volt drop.
So 11 or 12 Volts would be Better.
 
Gate Current is NOT an Issue for Mosfets.
Typical Gate Current on Mosfets is MUCH LESS than 1 uA.
But a GOOD Waveform, with a good Rise and Fall is ESSENTIAL.

DOES he Actually Have a GOOD SQUARE WAVE, Or is he driving with a TRIANGLE Wave?
A Triangle Wave will Create a LOT OF HEAT in the Mosfets.

I usually just drive my Mosfets wih a 555 and Never have any Problems.
Even at Quite High Frequencies.

While it is true that the static current into a mosfet gate is near zero, you do need current to dynamically drive a square wave into a capacitive load such as the gate of a mosfet.

I setup a simulation using a 1uA current source driving a mosfet with a Qg similar to what kinarfi is using. It took over 25 milliSeconds just to reach the miller plateau, and over 60mS to reach it's fully on state. At the 11-20KHz that kinarfi is running, the mosfet would never even think about turning on.

As for the 555, it is actually a pretty good gate driver. It's output is rated to 200mA DC, but I expect it's actual transient current ability is much higher.
 
Last edited:
**broken link removed** box reference deleted--Post#42---Some screen shots and a photo, added the circuit shown in post 30 to one comparator driver but had to keep the 1K resistor to feed the base, I will also put schottky diodes in. There add information on the pictures.
my frequency is 43KHz
I have to stay with a 10v V+ because of the hall effect device, but the pull up resistors are tied to 14 v or 25 v
The content of the dropbox was up loaded to ETO in post #42
 

Attachments

  • 2014-01-13 PWR STEERING PHOTO + SCHEMATIC.png
    2014-01-13 PWR STEERING PHOTO + SCHEMATIC.png
    3.9 MB · Views: 184
Last edited:
How many times do we have to say it, you cannot drive a mosfet fast using a pullup resistor. You need a proper driver to get this working without overheating, even a simple bjt complementary buffer will be better than this.

The gate capacitance of that mosfet is 3.3nF and for purposes of gate switching it will act even a little bit worse.
 
Gate Current is NOT an Issue for Mosfets.
Typical Gate Current on Mosfets is MUCH LESS than 1 uA.
But a GOOD Waveform, with a good Rise and Fall is ESSENTIAL.

DOES he Actually Have a GOOD SQUARE WAVE, Or is he driving with a TRIANGLE Wave?
A Triangle Wave will Create a LOT OF HEAT in the Mosfets.

I usually just drive my Mosfets wih a 555 and Never have any Problems.
Even at Quite High Frequencies.

Hi,


"Gate Current is NOT an Issue for Mosfets."
"Typical Gate Current on Mosfets is MUCH LESS than 1 uA."

Sorry to say but that's not entirely true. While it is true that the STATIC gate current is not usually an issue for MOSFET's (typically 1ua or less as you state) the DYNAMIC peak gate current can be 1 amp or more. That's a big difference, between 1ua and 1 amp.

The reason we need 1 amp or more is because the equivalent gate capacitance takes time to charge, and capacitors need lots of current to charge fast. That's why we need higher current, but only during the transition time from on to off or from off to on. We need it during the off to on transition too because we also want the MOSFET to turn off fast too.

With a low current like 10ua or even 1ma the input capacitance takes too long to charge and so the MOSFET turns on or off too slow, and that's when we see the triangle wave rather than a nice square pulsed gate voltage.


For example for an N channel device what happens turning turn on is the drain starts to fall as the gate voltage starts to rise and that couples back to the gate through the drain to gate capacitance. That makes the gate voltage stop rising for s time because that is negative feedback. Only after the drain gate capacitance charges does the gate rises all the way up. To get this to happen fast we need a very high gate current. It does not last as long as the normal gate current, but the driver has to be capable of doing this for at least a short time. The time the current reaches 1 amp (or more) can be mush shorter than the total gate pulse time but it depends on the total pulse width and the input capacitance.

The gate current requirement also depends on switching speed because it's a matter of how long it dwells in the transient state compared to how long it dwells in the fully on or fully off state. If we have a very slow on/off pulse time we may get away with a much lower gate drive current because it only has to change state once in a while. This is more typical with say a relay driver which only turns a relay on or off once in a while.

The power for resistive loads during the transition period can be estimated as one-sixth of the peak current times the peak voltage times the duty cycle. So for example if we have a 10 volt system and 60 amps peak, that would mean each switch period (on to off or off to on) takes 100 watts times the duty cycle of the transition period. If the transition period takes up the whole half cycle (a very fast on and off pulse) that means the transistor dissipates another 100 watts. If the transition period takes 10 percent of the time that means it only dissipates another 10 watts. So the faster we get this transition period to end the less power the MOSFET uses.
 
Last edited:
How many times do we have to say it, you cannot drive a mosfet fast using a pullup resistor. You need a proper driver to get this working without overheating, even a simple bjt complementary buffer will be better than this.

The gate capacitance of that mosfet is 3.3nF and for purposes of gate switching it will act even a little bit worse.

I'm not sure what your saying here, I'm changing the drivers from simple pull up to a transistor with a schottky "pull down" as shown in the uploaded file.

The LM2901N, National, has a response time of 1.3us which is much faster than the 43KHz of this circuit

Question for administration, is it proper to use photo sites like drop box for photos and drawing which I will delete after a while or should they be uploaded here for future references.
 

Attachments

  • Untitled.png
    Untitled.png
    5.9 KB · Views: 154
Sorry I was looking at the schematic you shown in the last post, and saw the same comparator configuration as before, just with 1k instead of 2k.
This way it looks it should work reasonably well.
 
It is my understanding pictures are to be uploaded to ETO and not hosted on third party sites. Due to the possibility of corrupt links or deletion.
 
It is my understanding pictures are to be uploaded to ETO and not hosted on third party sites. Due to the possibility of corrupt links or deletion.
Thank you, I agree with your reasoning, but was concerned about the space required, I shall upload my files to ETO.
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top