More Castle Creations Re-Calls!!!

[dani_r]

Nice to see a rather open and honest response from Castle. I've just bought the 75A ICE for my 500 and hope it's ok.

CC have a full reputation recovery close, just keep doing what they're doing! I guess only time will tell if they've done the right fix this time.

[Mark_T]

CC have a full reputation recovery close, just keep doing what they're doing! I guess only time will tell if they've done the right fix this time.

I think they are some distance from a 'full reputation recovery', they are doing the right thing now but only after months of denial and blaming their customers.

Follow that up with a recall that didn't fix the problem then a recall for all the ICE devices and no news on actual turnaround times with around 50,000 potential returns ...

I'm sure that some people will forgive and forget if they really do fix it this time but a lot are going to be very unlikely to touch CC for some time to come.

[dani_r]

I'm sure that some people will forgive and forget if they really do fix it this time

That's what I mean. If this fix does sort the issue then most people will move on. They have a good product (most of the time ) and have sold a fair few so people know how good it can be when it works.

Like I say, only time will tell. Let's just hope mine doesn't decide to set itself on fire

[xygax]

I think i mentioned the drive to the fets being one of the problems most likly to cause trouble. Its not turning fets on its sucking the energy out of the miller capacitor in a timely manner particularly when there are signifacant inductive loads.

and now we know its a drive problem ....

Hope you guys get it fixed quickly
STeve

[dani_r]

I think i mentioned the drive to the fets being one of the problems most likly to cause trouble. Its not turning fets on its sucking the energy out of the miller capacitor in a timely manner particularly when there are signifacant inductive loads.

and now we know its a drive problem ....

Hope you guys get it fixed quickly
STeve

What does it mean when you say not turning the fets on? 'scuse my ignorance

[xygax]

Firstly typo... should be 'off'.

To turn a mosfet on you take its gate above the Vgs threshold this charges the gate and it turns on. to turn it off basicly you need to connect its gate to its source. however there is a significant capacitor that exists between the drain and the gate and it takes a reasonable drive current to discharge it in a timely manner. if not its still partly or fully turned on when the oposite arm of the bridge comes on causing shoot through and some of the fires we have been seeing....

steve

[xygax]

Opps

very simple description of a speed controller...

to make a three phase signal for a motor it needs to be chopped up with 6 switches. (fets) These all work in sequence to make a 3 phase variable frequency output to drive a motor from DC. This all controlled by a little micro from a servo signal....

Steve

[Rctintin]

In lay terms what Steve has said is:-

The ugemaflip thingy is not allowing the how's your father to dump it's load at the right time, so it all goes Michael Barrymore! (Poof)

[rachel]

In lay terms what Steve has said is:-

The ugemaflip thingy is not allowing the how's your father to dump it's load at the right time, so it all goes Michael Barrymore! (Poof)

PMSL honestly I have never heard the MB quote before top post have arep point on me !!!!!!!!!

Beats the boaring trash CC rubbish anyday.

Thanks Rachel

[alioth]

What does it mean when you say not turning the fets on? 'scuse my ignorance

To try to break it down:

The FETs are field effect transistors. Basically they connect your LiPoly batteries to the motor, and work as electronically controlled switches to sequence the power from the batteries to the motor windings. Like any switch they can be off or on (actaully there's a region where they are sort of "not fully on" and have significant resistance).

Basically, the FET like any switch has an input and an output. The actual turning on and off of the switch is done at something called the gate. To turn the FET on, the gate voltage must exceed a certain voltage, called the threshold.

In an ideal world, it would be as simple as that, just raise the voltage well above the threshold (so the FET isn't partially conducting) to turn it on, and bring it down to 0 volts to turn it off, and it would happen instantly, and no current would need to flow into the gate. However, we are in the real world, and the gate forms what's known as a "parasitic capacitor", effectively a capacitor we don't want right there in our switch, the FET. For the gate voltage to rise to the right level to turn on, or fall back to a level to turn off, this capacitance must be charged and discharged too, which slows the speed at which the controlling voltage can rise and fall. It can also take - for a brief period of time - a significant amount of current to do this. Therefore, unwanted resistances in the rest of the system, for example, if the power rails and ground return for the chip that drives the FET's gate are not ideal, this current spike can lift the voltage at the ground pin (which is called "ground bounce", or sag the voltage at the supply pin when going the other way), and this will just make it worse.

Imagine it like this: imagine one of those big industrial lever switches. Imagine that you pull it down to turn the lights on, and instead of being a simple on-off, while the switch is between fully off and fully on, there is a big variable resistor. You don't want to stay in this region with the variable resistor for very long, because you get lots of heat build up from the resistor if you do, effectively you want to turn the switch on and off as fast as possible. Now imagine we put a hydraulic damper on this lever switch. This is our parasitic capacitance. However hard you pull on the switch you'll always have some transistion time between fully off and fully on as you have to force the oil around the system. Now imagine you pull on it really really hard because you want it to switch fast...your connection to ground might not be enough, and your feet lift off the ground, and that's the fastest you'll be able to make it move. Imagine also the ground is also a rubber sheet, so when you want to push it the other way, if you push hard enough you won't move the switch any faster, instead the rubber sheet bends under your feet. Now imagine you have to do this while being synchronized with other people doing the same thing with their switches lest you cause a short circuit...

The other thing about FETs is that if the gate is not connected to anything - let's say the driver chip's output fails - the gate effectively just floats at some level you can't necessarily predict. Effectively, a FET with an unconnected gate can just slowly (in terms of many milliseconds) turn on, and if this happens at the wrong time you've got a short circuit across a LiPoly pack that can quite happily discharge at many hundreds of amps, and the FET will turn into a very spectacular fuse. The parasitic capacitance can do other nasty things, too, for instance unwanted voltages can become capacitatively coupled from the power side of the circuit back into the driver.

[R1stu]

On a side note. Would BMFA insurance cover damage to property? say if you flew your heli with a known CC product and the heli caught fire and crashed into someone's hse/car garden and set fire to their property would the cough up?

[dani_r]

In lay terms what Steve has said is:-

The ugemaflip thingy is not allowing the how's your father to dump it's load at the right time, so it all goes Michael Barrymore! (Poof)

Thanks for the explanation! Coz I basically had no idea what most of that meant :P

[dani_r]

To try to break it down:

The FETs are field effect transistors. Basically they connect your LiPoly batteries to the motor, and work as electronically controlled switches to sequence the power from the batteries to the motor windings. Like any switch they can be off or on (actaully there's a region where they are sort of "not fully on" and have significant resistance).

Basically, the FET like any switch has an input and an output. The actual turning on and off of the switch is done at something called the gate. To turn the FET on, the gate voltage must exceed a certain voltage, called the threshold.

In an ideal world, it would be as simple as that, just raise the voltage well above the threshold (so the FET isn't partially conducting) to turn it on, and bring it down to 0 volts to turn it off, and it would happen instantly, and no current would need to flow into the gate. However, we are in the real world, and the gate forms what's known as a "parasitic capacitor", effectively a capacitor we don't want right there in our switch, the FET. For the gate voltage to rise to the right level to turn on, or fall back to a level to turn off, this capacitance must be charged and discharged too, which slows the speed at which the controlling voltage can rise and fall. It can also take - for a brief period of time - a significant amount of current to do this. Therefore, unwanted resistances in the rest of the system, for example, if the power rails and ground return for the chip that drives the FET's gate are not ideal, this current spike can lift the voltage at the ground pin (which is called "ground bounce", or sag the voltage at the supply pin when going the other way), and this will just make it worse.

Imagine it like this: imagine one of those big industrial lever switches. Imagine that you pull it down to turn the lights on, and instead of being a simple on-off, while the switch is between fully off and fully on, there is a big variable resistor. You don't want to stay in this region with the variable resistor for very long, because you get lots of heat build up from the resistor if you do, effectively you want to turn the switch on and off as fast as possible. Now imagine we put a hydraulic damper on this lever switch. This is our parasitic capacitance. However hard you pull on the switch you'll always have some transistion time between fully off and fully on as you have to force the oil around the system. Now imagine you pull on it really really hard because you want it to switch fast...your connection to ground might not be enough, and your feet lift off the ground, and that's the fastest you'll be able to make it move. Imagine also the ground is also a rubber sheet, so when you want to push it the other way, if you push hard enough you won't move the switch any faster, instead the rubber sheet bends under your feet. Now imagine you have to do this while being synchronized with other people doing the same thing with their switches lest you cause a short circuit...

The other thing about FETs is that if the gate is not connected to anything - let's say the driver chip's output fails - the gate effectively just floats at some level you can't necessarily predict. Effectively, a FET with an unconnected gate can just slowly (in terms of many milliseconds) turn on, and if this happens at the wrong time you've got a short circuit across a LiPoly pack that can quite happily discharge at many hundreds of amps, and the FET will turn into a very spectacular fuse. The parasitic capacitance can do other nasty things, too, for instance unwanted voltages can become capacitatively coupled from the power side of the circuit back into the driver.

Many thanks for your explanation! Much clearer now

[coolice]

Hey.

The way I've understood the problem, after someone took apart their 120HV and fixed it themselves, is that while the components are of good quality (better than other ESC makers it's claimed) the gate controller chip is just too far away from the FET's. I think a PCB track length of 5cm was mentioned.
Apart from the obvious fix of moving components to fix the problem, another option was to use lesser quality FET's.

Here is a post of a linked topic somewhere on this forum, this guy knows his beans it seems;

RC Groups - View Single Post - Ice120HV exploded in flight

"
My Story...

i have had 2 of 3 120HV catch fire. Ive also had 1 of 2 80HVs blow up.

I am Power Electronics Engineer, and have designed both firmware and hardware for these types of controllers.
In the past few years Ive designed various brushless ESC for both commercial and military use, including high power, high voltage ESCs up to 20kw, and regard myself an expert in the field.

Sooo...
On day 1 of owning my 120HV and 80HV, i had the 120HV lose sync of the motor in the first flight. It seemed like it had trouble powering one of the phases. Swapped it out for the 80HV, and all was well.

I thought.. This is either one of two things. A Software issue, or Hardware issue. It had to be a hardware issue because some controllers work fine for months, and some fail first flight, and they all have the same firmware.

I decided to open and have a look at the controller as i couldn't be bothered sending it in and waiting for warranty.

Anyways, a quick test of the FETs showed they are all working fine. I tested the Gate to Source resistance, and the Drain to Source resistance. Even applied signal to the gates and re-tested the Drain-Source resistance. All was well. One thing i noticed here was that the the slew rate on signal generator was pretty low, why?? it because the gates draw alot of current at high frequency. This is because we have heaps of gate capacitance, due to the Low Rds FETs (which have a high gate charge), and the fact that we have Many of them in Parallel. But anyways they were working fine.

Sooo.. What is the Issue..
I tested the voltage-divider/low-pass filter circuits on each phase of the PCB. These circuits provide phase voltage feedback to the MCU so the controller knows the stator angle, and when to commutate. These were also fine.

One more thing to test... The FET drivers...
Ah ha.. When powered up, one of these units was heating up under no-load. This is not normal. The drivers used in the design are the Intersil ISL6700 (http://www.intersil.com/data/fn/fn9077.pdf). They are a solid driver (when used correctly) and I've used them in previous low-power designs. A quick trip down the road to element14 and purchased a new set of 3 drivers (http://au.element14.com/intersil/isl...33?Ntt=isl6700). Some hot-air soldering and all was well. Controller worked perfectly and is still going strong.

One thing that worried me was the fact that each single driver is driving 12 FETs. Not only is it stressed from the extra capacitance, but also the long signal path to the gate from the driver adds a inductive load as well. The drivers are located far from the FETs on the upper PCB. The signal path to the further most FET is nearly 5cm, and even further when we take into considering the connector between the 2 PCBs. No good, because in some cases the combination of components and PCB layout (causes parasitics) can cause excess voltages on the driver pins and damage the units.

This Technical Brief slightly explains what is happening.. http://www.intersil.com/data/tb/tb447.pdf

This worried me alot. Why??? Because the driver that failed, was linked to the further most MOSFET Half-Bridge/Phase.

This has led me to believe that if the ISL6700 driver is excessively stressed, or is manufactured on the lower side of the manufacturing tolerance (because every batch is different), it could fail and cause the controller to fire. If the FET gates are left floating after the driver fails, you will have a fire because the parasitic voltages inherent in the gates will caused both the lower and upper side FETs to conduct, leading to a direct short circuit across your LiPOs.

Solutions:
1) Replace/Redesign the current driver for a more robust unit with added protection. The current unit is driving too large of a gate capacitance via a extra long path.
2) Take more consideration in the PCB Layout.

I have been using castle ESCs for over 10 years for there service, quality and price. I love them and to be honest I'm a bit of a castle fanboy, and will continue to be. However, in this particular case (12S LiPO ICE_HV), there is a flaw in the design, and I'm glad it getting rectified.

Regards,
Tony "

Complete topic link; Ice120HV exploded in flight - Page 7 - RC Groups
.

[Mark_T]

On a side note. Would BMFA insurance cover damage to property? say if you flew your heli with a known CC product and the heli caught fire and crashed into someone's hse/car garden and set fire to their property would the cough up?

Prior to the notice issued by CC, if what you describe happened and you were found liable then yes, the insurance would pay out.

After the notice from CC and if you could be shown to have received or been made aware of it, then I'd suggest you were not complying with this part of the insurance policy if you were still flying with one of the affected devices :

'(b) It is the duty of the insured to exercise reasonable care to see that their models are sound and in proper order and that all reasonable safeguards and precautions against accidents are provided and used.'

Even in the first case, you have to be found liable, it is not just a case of crash, burn and payout.