Castle - Make A Special Edition Monster

[lutach]

A question for the engineers out there . Would it make sense to have a lot of MOSFET (Lets say a FET driver already was chosen and powerful enough to drive close to 300 FETs) on a controller to make each MOSFET work less to provide the power some of us need? Lets say I have a design for a controller with 240 MOSFET total, 80 per board, would it actually run cooler because there will be less load for each MOSFET? The reason for this is I see a few controllers with less FET that actually run a bit hot, but a controller with the same spec (AMP rating wise) with more FETs will actually run cooler. One example is one of my 30A old Kontronik controller runs cooler then some of the latest 35A controllers.

Well, I would think that the more surface area you have to dissipate a set amount of wattage, the better. More FETs translates into more surface area to handle the same wattage (heat we need to dissipate). But with that said...

The only thing I would be conecerned with is if there would be some sort of inefficiency with driving that many FETs? I mean, each FET is only efficient to a certain point, and what about switching losses? More FETs means more wattage lost due to switching loss no?

Interesting problem posed lutach. I'm going to keep an eye on this thread to see what the general concesus is from Patrick and Artur (and anyone else who's an EE- which I am not).

[lutach]

Originally Posted by Five-oh-joe

Well, I would think that the more surface area you have to dissipate a set amount of wattage, the better. More FETs translates into more surface area to handle the same wattage (heat we need to dissipate). But with that said...

The only thing I would be conecerned with is if there would be some sort of inefficiency with driving that many FETs? I mean, each FET is only efficient to a certain point, and what about switching losses? More FETs means more wattage lost due to switching loss no?

Interesting problem posed lutach. I'm going to keep an eye on this thread to see what the general concesus is from Patrick and Artur (and anyone else who's an EE- which I am not).

I've seen some DC/DC converter that was small and packed a lot of FETs. This was in one of the many electronic magazines I receive and the company was claiming they don't require a heat sink. They were saying something about a dense PCB, which in my point of view means a lot of components placed as close as possible. The Tekin R1pro for example, packs a lot of FET in a small package. That's why I'm asking this question with the switching part covered. I would think the more the FETs, the less heat will be generated to make a certain amount of power. I attached a couple of pics that shows some controllers that are proven to be really powerful and here is a link of another one http://www.rc-monster.com/forum/show...6&postcount=16.

[GriffinRU]

Originally Posted by lutach

A question for the engineers out there . Would it make sense to have a lot of MOSFET (Lets say a FET driver already was chosen and powerful enough to drive close to 300 FETs) on a controller to make each MOSFET work less to provide the power some of us need? Lets say I have a design for a controller with 240 MOSFET total, 80 per board, would it actually run cooler because there will be less load for each MOSFET? The reason for this is I see a few controllers with less FET that actually run a bit hot, but a controller with the same spec (AMP rating wise) with more FETs will actually run cooler. One example is one of my 30A old Kontronik controller runs cooler then some of the latest 35A controllers.

Very nice post, Five-oh-joe

More fet's bigger foot print more surface area to dissipate heat, more overall losses...
Fet's driver need to be very nice or need to be on each FET's board, but try to sync them...possible.

Let's do quick theoretical calc, if you can drive fet's as fast as fet can switch (non-real) then 300 fet's you meant total, so 50 fet's per leg/100 fet's per phase. Dynamic losses would be ~10W on fet and ~10W on diode, check attached image. With 3 phases your loss would be ~60W just for switching at 15kHz PWM, that would cook ESC pretty fast without proper heatsinking, by the way can be a nice heater :)

It would be nice to keep dynamic losses matched to pcb heat dissipation capacity.

I am pretty sure, Patrick can add/correct my post if required.

[lutach]

Originally Posted by GriffinRU

Very nice post, Five-oh-joe

More fet's bigger foot print more surface area to dissipate heat, more overall losses...
Fet's driver need to be very nice or need to be on each FET's board, but try to sync them...possible.

Let's do quick theoretical calc, if you can drive fet's as fast as fet can switch (non-real) then 300 fet's you meant total, so 50 fet's per leg/100 fet's per phase. Dynamic losses would be ~10W on fet and ~10W on diode, check attached image. With 3 phases your loss would be ~60W just for switching at 15kHz PWM, that would cook ESC pretty fast without proper heatsinking, by the way can be a nice heater :)

It would be nice to keep dynamic losses matched to pcb heat dissipation capacity.

I am pretty sure, Patrick can add/correct my post if required.

Artur,

What I was getting at is basically looking at the Tekin R1 controllers. They are small and pack a lot of FET for the space so if we were to make them a little bigger and add more FETs, would that make the FETs work less to produce the same power and thus having less or the same amount of heat? I own a Etti 200A controller and I ran it a few times. This controller is only rated for 5S lipos and I ran 5S on it and the controller didn't even get warm. This controller doesn't have a heat sink like most car controllers and I was surprised on how cool it was. My set up at the time only pulled 98A spikes, but I ran my truggy for 7 minutes. The one picture in my previous post shows a 6 power board controller that puts out a lot of AMPs. Something like that would be great for the surface side of the hobby.

[GriffinRU]

Originally Posted by lutach

Artur,

What I was getting at is basically looking at the Tekin R1 controllers. They are small and pack a lot of FET for the space so if we were to make them a little bigger and add more FETs, would that make the FETs work less to produce the same power and thus having less or the same amount of heat? I own a Etti 200A controller and I ran it a few times. This controller is only rated for 5S lipos and I ran 5S on it and the controller didn't even get warm. This controller doesn't have a heat sink like most car controllers and I was surprised on how cool it was. My set up at the time only pulled 98A spikes, but I ran my truggy for 7 minutes. The one picture in my previous post shows a 6 power board controller that puts out a lot of AMPs. Something like that would be great for the surface side of the hobby.

Luciano,
From 50 fet's (300 total) to 8 fet's (48 total, Tekin R1PRO) your dynamic losses falling from ~20W to 3.2W, pcb can absorb that. Conduction loss at 100Amps @25C with 8 fet's ~2W -> as you can see great combo.
Dynamic losses are small with less fet's and becoming dominant with more, so balance should be found in-between on design stage. Where entire ESC been layout and verified step-by-step, you just cannot add fet's boards without penalty.

[lutach]

Originally Posted by GriffinRU

Luciano,
From 50 fet's (300 total) to 8 fet's (48 total, Tekin R1PRO) your dynamic losses falling from ~20W to 3.2W, pcb can absorb that. Conduction loss at 100Amps @25C with 8 fet's ~2W -> as you can see great combo.
Dynamic losses are small with less fet's and becoming dominant with more, so balance should be found in-between on design stage. Where entire ESC been layout and verified step-by-step, you just cannot add fet's boards without penalty.

Lets assume you have a quad power board R1, would that make less then the 3.2W of dynamic loss? The controller will have more FETs and each one would be working less to make the same amount of power that a single or double power board controller would be. Another example which is completely off topic would be in car audio. I had a 4 woofer set up in my car and it was loud (Load enough to have neighbors complain that things were falling from their furniture's ), but then I went up 4 more for a total of 8 woofer wired for the same load to the same amp and it was blistering loud. Each woofer was getting less power from the amp, but they were putting out higher db at the same time. The design that I have in mind (Artur I will send you 2 examples via e-mail) will be a total of 240 small FETs with a datasheet rating of 10A each and they come in 30V and 60V. Each power board will have 80 FETs so 40 in the H-Bridge you mentioned in another post. I'm just throwing out ideas basically to see a rock solid controller that is better then a Schulze 40.160, but a tad smaller for around the same price.

[lutach]

Artur,

I sent you the e-mail with the 2 photos. Let me (us) know what you think of those controllers.

[GriffinRU]

Originally Posted by lutach

Lets assume you have a quad power board R1, would that make less then the 3.2W of dynamic loss? The controller will have more FETs and each one would be working less to make the same amount of power that a single or double power board controller would be. Another example which is completely off topic would be in car audio. I had a 4 woofer set up in my car and it was loud (Load enough to have neighbors complain that things were falling from their furniture's ), but then I went up 4 more for a total of 8 woofer wired for the same load to the same amp and it was blistering loud. Each woofer was getting less power from the amp, but they were putting out higher db at the same time. The design that I have in mind (Artur I will send you 2 examples via e-mail) will be a total of 240 small FETs with a datasheet rating of 10A each and they come in 30V and 60V. Each power board will have 80 FETs so 40 in the H-Bridge you mentioned in another post. I'm just throwing out ideas basically to see a rock solid controller that is better then a Schulze 40.160, but a tad smaller for around the same price.

We will continue next week-end, need to go.
Dynamic loss is fixed with fet, each fet has a capacitance in the gate (value dependant on temp, load...) to charge and discharge this capacitor requires energy, so more fet's more energy.
Conduction loss would be related to Rdon and more fet's lower value, so good thing, but as you can see you need to keep balance...

I will check e-mail later, and let you know.

Keep in mind TO-220's and D-Pak's are huge packages and can easily absorb ~1W without heatsink, but they are not as fast as smaller, tuned fet's.

Your Amp has a pretty sized heatsink, right :)

[lutach]

Originally Posted by GriffinRU

We will continue next week-end, need to go.
Dynamic loss is fixed with fet, each fet has a capacitance in the gate (value dependant on temp, load...) to charge and discharge this capacitor requires energy, so more fet's more energy.
Conduction loss would be related to Rdon and more fet's lower value, so good thing, but as you can see you need to keep balance...

I will check e-mail later, and let you know.

Keep in mind TO-220's and D-Pak's are huge packages and can easily absorb ~1W without heatsink, but they are not as fast as smaller, tuned fet's.

Your Amp has a pretty sized heatsink, right :)

Heat sink issues will be no problem as the design I have in mind will allow for a very good heat sink to be used without sacrificing size (Compared to much larger ESCs). Look at this article in Power Electronics Technology http://powerelectronics.com/thermal_...osfet_cooling/. The LFPAKs that I have been mentioning are being compared to the other package found on controllers such as the Schulze 40.160, Actronic 120, Jeti SPIN170-300 and the ones I'm having made. This article is saying things like, "This fact also opens the possibility of further reducing the pc board space occupied by the power MOSFETs by using physically smaller devices and placing those devices closer together. To investigate this possibility, another simulation was carried out where each pair of D2PAK MOSFETs in the H-bridge circuit was replaced with three LFPAK devices.

The LFPAK package is much smaller than the D2PAK, occupying the same pc board footprint as the familiar SO-8 package. However, unlike the SO-8, the LFPAK is a true power package that incorporates a bottom metal contact, which provides an effective heat path out of the device. There is an additional thermal pathway between the top of the device silicon and ambient through the top part of the encapsulation.

Although the LFPAK solution increases the total number of power packages used, the total board area occupied by this solution is significantly less than for the D2PAK case because the LFPAK package is much smaller than the D2PAK." and "In the last few years, there have been significant advances in the packaging of MOSFETs, including the introduction of the power SO-8 package. Bottom-side cooling can now be used successfully to transfer heat through the pc board, even when smaller power packages like the LFPAK are used in place of the D2PAK. The package on-resistance and inductances for these smaller package types are also significantly lower. The total losses in a system caused by these sources are therefore reduced significantly, even with the additional devices needed when using the smaller package types in place of larger MOSFETS.

The greatest advantage when switching from D2PAK to LFPAK is the resulting reduction of board space occupied by the MOSFETs, since pc board top copper is not needed to radiate heat. The smaller MOSFETs can be placed closer together, and the previously occupied board space is made available for other components." The components that I have in mind though are much smaller measuring 3mm Length x 2mm Width X 0.8mm height. If this works, it should make a very powerful controller.

Originally Posted by GriffinRU

Very nice post, Five-oh-joe

More fet's bigger foot print more surface area to dissipate heat, more overall losses...
Fet's driver need to be very nice or need to be on each FET's board, but try to sync them...possible.

Let's do quick theoretical calc, if you can drive fet's as fast as fet can switch (non-real) then 300 fet's you meant total, so 50 fet's per leg/100 fet's per phase. Dynamic losses would be ~10W on fet and ~10W on diode, check attached image. With 3 phases your loss would be ~60W just for switching at 15kHz PWM, that would cook ESC pretty fast without proper heatsinking, by the way can be a nice heater :)

It would be nice to keep dynamic losses matched to pcb heat dissipation capacity.

I am pretty sure, Patrick can add/correct my post if required.

Thanks Artur!

Lutach- sounds like you got some serious stuff planned. I haven't read through all of your posts yet, so I'm a bit behind. It's nice to have consumer ideas thrown into the mix though!

So what's the difference between the SO8 FETs and the DPAKs? I'm guessing DPAKs have a metal type housing whereas SO8s are still that resin/plastic?