DIY Electric Car Forums banner

1 - 20 of 118 Posts

·
Registered
Joined
·
730 Posts
Discussion Starter #1 (Edited)
Recently high power silicon carbide mosfet modules have become available.
The advantages are obvious. Higher frequency (no audible noise), better efficiency (smaller battery pack), no need for liquid cooling (weight saving) and lower overall cost.
The only disadvantage I've come across so far are gate voltages that differ from IGBT.
But that's an easy fix (simple voltage converter).

Overdimensioning an IPM power stage as a protection measure might work for low power (test) puposes.
For higher output power levels a more advanced protection mechanism is needed.

Desaturation detection is as far as I know the only reliable method for on-state overload protection of IPMs that are used near the load limits.

Here are links to Avago (formerly HP) desat optocoupler driver technical documents:
http://www.avagotech.com/docs/AV02-0803EN
Broadcom Inc. | Connecting Everything

The layout for a printed circuit board for the SiC gate drive circuit is attached. First layout for testing purposes.
The circuit diagrams and final pcb layouts will follow when the testing is done.

Components have been selected for high reliabilty and a temperature range of -40 °C to 85 °C as a minimum requirement. Automotive grade if available.

And should be OK for DIY. No super tiny SMT. Smallest spacing about 1/40” (0,6 mm). Solderable with a 0,3/0,4 mm tip, a steady hand and maybe a common magnifying glass.


UPDATE (2020). DELETED IMAGE: CASCODE SiC JFETS will be used when the build starts. Eight microseconds short circuit rated.
MICROCONTROLLER UPDATE: Automotive grade ATMEL D51, great development tools, E version of the ARM based chip has CAN. Suitable for FOC.
 

·
Registered
Joined
·
3,219 Posts
Your post is, unfortunately, one of those that can't be quoted (I bet you can't edit it, either), so I copied the relevant portion of text I want to respond to and will generically quote it.

...The advantages [of SiC modules] are obvious. Higher frequency (no audible noise), better efficiency (smaller battery pack), no need for liquid cooling (weight saving) and lower overall cost.
Higher frequency is not necessarily an advantage in motor drives. The lamination thickness of the armature is invariably determined by the "rotational" frequency it operates at, which is far lower than the PWM frequency. However, iron losses in the armature are definitely proportional to both armature (ie - rotational) frequency *and* PWM frequency. So, you really want to use the lowest PWM frequency possible to maximize motor efficiency.

Note that "armature" refers to the part of a motor that experiences alternating flux swings; this is usually the "stator" of an AC motor and the "rotor" of a DC motor.

As for efficiency, SiC is a so-called "wide bandgap" semiconductor and so it has a higher blocking voltage for a given conductivity, but it is still a strictly "majority carrier" device and therefore will always have higher conduction loss than a bipolar device of the same voltage/current rating (ie - IGBT). There are only two advantages to SiC MOSFETs and only one of those is relevant to motor drives: 1) they can switch at a much higher frequency as a result of not having to suffer minority carrier recombination; 2) they can tolerate a higher operating temperature (approximately 150C continuous, versus 100C for conventional Si MOSFETs and IGBTs).

Desaturation detection is as far as I know the only reliable method for on-state overload protection of IPMs that are used near the load limits.
Desaturation is a unique phenomenon of bipolar devices like IGBTs and does not apply to unipolar devices like MOSFETs. Without getting too bogged down in technical details that will just make the eyes of most people here glaze over, the voltage drop across a MOSFET is linear with current (though exponentially proportional to temperature; roughly to the 1.6 power) while the voltage drop across a bipolar device like an IGBT is relatively constant except for an ohmic loss which is linearly proportional to current (just like a MOSFET) and a diode-like increase of 60mV/decade of current up until the current is so high that there are insufficient charge carriers available. It is at this point that the voltage drop abruptly increases (e.g., from 1.5V to 6V or more) and the bipolar device is then said to have come out of saturation (ie - it has desaturated).

Long story short, you can't use a typical "desaturation protection" circuit with SiC MOSFETs; rather, overcurrent protection will have to depend on high speed "cycle by cycle" limiting.
 

·
Registered
Joined
·
33 Posts
SiC has much higher thermal conductivity than Si and is less temperature sensitive which is a big plus in an EV.

WBG transistors would allow the magnetics for a boost converter to shrink to allow for smaller packaging.

Easy to parallel since they are more or less self regulating.

Less turn on/off losses

Easier to make a pure/very low THD sine wave motor inverter because of higher switching frequency. (Smaller filter) And output is spike free. This also improves motor efficiency

I've been reading more about GaN than SiC but I believe they share similar characteristics.
 

·
Registered
Joined
·
730 Posts
Discussion Starter #4
Desaturation is a term that is used for bipolar devices. when the so called linear region is entered. The voltage increases rapidly with increasing current. At high loads it leads to device destruction (overheating). That's the key property. Mosfets show a similar behaviour. I'll keep using the term desaturation because that is what manufacturers still use in datasheets and white papers.

Iron losses: if the flux variations are kept low, the loss increase due to a higher PWM frequency should not be substantial.

Conduction losses: those losses are about the same, but switching losses are much, much lower.
 

·
Registered
Joined
·
730 Posts
Discussion Starter #5
The PCB is ready and the first test is done. The results confirm that the Avago ACPL-332J can be used as the primary desat driver.

Driver board Q&A:

Q: What is the maximum repetitive peak working isolation voltage?
A: The ACPL-332J optocoupler's isolation voltage specification is 891V peak.

Q: How many SiC devices can be driven by one board?
A: Two Cree CAS300M SiC mosfets can be driven in parallel.

Q: The Cree CAS300M SiC needs a very low gate resistor for the highest efficiency. Can the board handle the high peak currents?
A: Yes. Each mosfet has its own IXYS IXD*609* high current driver ic.

Q: How is the IXYS driver turned off when the SiCmosfet desaturates?
A: An IXYS IXDD driver ic has an enable input. A signal from the 332J optocoupler drives a transistor that pullls the enable input low.

Q: Can the fast switching of SiC mosfets cause interference on the board?
A: The board is shielded. The optocoupler and other isolation circuits have been designed for inverter application.
The CMR value is closer to 10kV per microsecond than the usual design goal of 5kV per microsecond for inverters.
 

·
Registered
Joined
·
33 Posts
Desaturation is a term that is used for bipolar devices. when the so called linear region is entered. The voltage increases rapidly with increasing current. At high loads it leads to device destruction (overheating). That's the key property. Mosfets show a similar behaviour. I'll keep using the term desaturation because that is what manufacturers still use in datasheets and white papers.

Iron losses: if the flux variations are kept low, the loss increase due to a higher PWM frequency should not be substantial.

Conduction losses: those losses are about the same, but switching losses are much, much lower.
According to the datasheet for this device, CAS300M12BM2, there is no "linear" or "desaturation" region so you don't need desaturation detection. The current applied is almost linearly proportional to the voltage drop across S-D (Ron) That would simplify driver design as all you would need is a way to measure current. Pretty much what Tesseract said.

Why such an overkill device though? I don't see any EVs over about 3-400V. You can sacrifice some of that blocking capability for higher current capacity and a much less expensive device...
 

·
Registered
Joined
·
730 Posts
Discussion Starter #7 (Edited)
There's a linear region allright for the CAS300M12 in the on state. For every mosfet.

As soon as the drain current of the mosfet exceeds a certain level the mosfet no longer acts like a resistor but as a (gate) voltage controlled current source.

The desat is a extremely fast protection mechanism (few microseconds) that no current measurement system can match. It needs to be that fast for short circuits and low gate voltages. In general: fault conditions.

Why expensive SiC? I'm not buying them now. The price has to drop first.
For instance size. Phase output filters are much bulkier for cheaper devices and so is the cooling.
 

·
Registered
Joined
·
33 Posts
How much current are you running through the device? The Cree's graph tops out at 600A. With a decent driver, it looks like it can tolerate the rated current plus a few current spikes here and there. Then again, it doesn't take much to fry a FET.



I guess I'm thinking the cycle-by-cycle current limiting should throw a red flag once a certain threshold is passed. Or am I missing something there?



I was still speaking of the WBG semiconductor as I love the potential, but more so some of the other devices on the market.
 

·
Registered
Joined
·
730 Posts
Discussion Starter #9 (Edited)
It's much easier to fry an IGBT. Latch up. I've even used avalanche rated MOSFETS as a IGBT turn off protection in a design.

But glad you asked. Desaturation detection protects against FAULT conditions in the on state. At high loads.
Basically two cases: overcurrent (short circuit) and faulty gate drive voltages (low). When rapid intervention is a must.

The nominal current load for the CAS300 is 300A. But it takes quite a bit of math to calculate the maximum current for a specific application.

Technical background: The linear region starts above 600A. Up to 1500A (device maximum).
During a fault condition it's power (heat) that destroys the mosfet. But the thermal properties of the CAS300 are impressive.
A dynamic junction to case thermal resistance of about 0.0002 K/W for a single pulse event.
That means the CAS300 can take more than 200kW heat. But only for a very short time. In the microsecond region.
 

·
Registered
Joined
·
3,219 Posts
There's a linear region allright for the CAS300M12 in the on state. For every mosfet.
Correct. MOSFETs - including SiC types - act like a resistor with some temperature dependence when fully turned on.

As soon as the drain current of the mosfet exceeds a certain level the mosfet no longer acts like a resistor but as a (gate) voltage controlled current source.
Incorrect. You are describing how a bipolar device - such as an IGBT - behaves. The voltage drop across the collector-emitter junction of a bipolar device is almost linear with gate/base voltage once the gate/base voltage is brought above conduction threshold [and then the CE drop increases approx. 60mV per decade of current, just like a BJT, plus a minor contribution from Ohmic losses]. In contrast, the drain-source resistance in a standard Si MOSFET might halve when going from 5V to 10V, and change not at all when going from 10V to 15V. SiC MOSFETs tend to require a higher voltage to fully turn on (20V vs. 10V for a standard MOSFET), and thus have a somewhat wider "square law Ohmic region" above the threshold voltage, but otherwise behave the same.

There is such thing as a SiC BJT (and JFET), but I haven't personally used either.

The desat is a extremely fast protection mechanism (few microseconds) that no current measurement system can match. It needs to be that fast for short circuits and low gate voltages. In general: fault conditions.
Desaturation occurs much faster than a "few microseconds"; more like a few nanoseconds depending on the die size. Usually bipolar devices are rated to withstand desaturation for 6-10us, however, which might be what you are referring to. Nevertheless, it is entirely possible to make a current limiting circuit that reacts nearly as quickly as desaturation; the problem is that a majority carrier device does not limit short-circuit current whereas a bipolar device in desaturation does. In other words, bipolar devices will protect themselves to some extent during a short. It's one of the primary reasons they are more rugged than unipolar (majority carrier only) devices.

Why expensive SiC? I'm not buying them now. The price has to drop first.
For instance size. Phase output filters are much bulkier for cheaper devices and so is the cooling.
"Phase output filters" are rarely used on motor drives, and only then when the cables between the inverter and motor are longer than 10's of meters. Occasionally there is a light "dV/dt" output filter and/or a common mode choke around all three phase cables, but those are not really for integrating the PWM'ed output voltage waveform into more of a sine wave; just for softening the switching transitions.
 

·
Registered
Joined
·
33 Posts
Fault protection from over current situations and gate drive errors can be observed by a microcontroller.



Would a sine wave drive really increase motor efficiency? Or is that just marketing talk?
 

·
Registered
Joined
·
730 Posts
Discussion Starter #12 (Edited)
Posted by Tesseract:

Quote:
Originally Posted by Tony Bogs
As soon as the drain current of the mosfet exceeds a certain level the mosfet no longer acts like a resistor but as a (gate) voltage controlled current source. End quote.

Tesseract:
Incorrect. You are describing how a bipolar device - such as an IGBT - behaves. The voltage drop across the collector-emitter junction of a bipolar device is almost linear with gate/base voltage once the gate/base voltage is brought above conduction threshold [and then the CE drop increases approx. 60mV per decade of current, just like a BJT, plus a minor contribution from Ohmic losses]. In contrast, the drain-source resistance in a standard Si MOSFET might halve when going from 5V to 10V, and change not at all when going from 10V to 15V. SiC MOSFETs tend to require a higher voltage to fully turn on (20V vs. 10V for a standard MOSFET), and thus have a somewhat wider "square law Ohmic region" above the threshold voltage, but otherwise behave the same.
Maybe you're more experienced with conventional (non punch through, very light punch through) IGBTs with 10 microsecond short circuit ratings than fast punch through types and mosfets.

I've mostly used mosfets for designs.
Mosfets can definitely enter linear mode in the on state.
In fact, the Cree CAS300 has a very constant transconductance over working temperature: gfs=94.
Drain current = gfs * gate voltage.
Clarification: this formula sets the threshold current value. Below it the SiC mosfet acts like a resistor, above it as a gate voltage controlled current source.
Probably not a good (drain-source voltage independant) current source, but who cares? It is a fault condition in inverters.

Desaturation protection: common practice for fast IGBT (punch through) and mosfets.
If I remember correctly, Cree supplies a gate driver with desat protection for the CAS300. And IXYS drive ics. Sold seperately.

Phase filters /pure sine input:
The main reason for using filters is EMI/RFI. Keeps the major part of the high frequency components in the shielded inverter housing.
And (less important):
The motor will have less iron losses. But the filter adds losses.
 

·
Registered
Joined
·
730 Posts
Discussion Starter #13
I've compared the 300A Cree CAS300 with a “reference design” for a 300V DC bus Siemens Azure motor that uses fast IXYS PT IGBTs. The motor is well known on this site. Maximum phase peak current 225A at nominal power, 450A at short period peak power.

One CAS300 outperforms two 200A PT IGBTs on the switching losses.
Numbers at peak power (450A), based on extrapolation of datasheet information:
Cree @20 kHz 80W
Two IGBTs @16kHz in parallel 540W

But since the SiC is used above the nominal current rating (450 to 300) the conductance losses are higher. About 300W. Net result 540 – 380 = 160 W. Per mosfet.
So in total for a full inverter about 1kW loss saving at max peak power.
 

·
Registered
Joined
·
3,219 Posts
...Mosfets can definitely enter linear mode in the on state.
In fact, the Cree CAS300 has a very constant transconductance over working temperature: gfs=94.
Drain current = gfs * gate voltage.
Clarification: this formula sets the threshold current value. Below it the SiC mosfet acts like a resistor, above it as a gate voltage controlled current source.
Probably not a good (drain-source voltage independant) current source, but who cares? It is a fault condition in inverters.
I was mostly objecting to you calling the behavior of any kind of MOSFET experiencing overcurrent as "desaturation", as that is something that can only occur in bipolar devices.

That said, I actually looked at the datasheet for this module and it does seem that because of the relatively low transconductance you could, indeed, use a standard IGBT drive circuit with desaturation type overcurrent protection, so any argument over proper nomenclature is academic at best. The maximum fault current through a CAS300M12BM2 module with 20V of gate drive will be around 1900A and with 15V of drive (ie - the standard for IGBTs) it will limit to around 1400A, which it can take for 200us, at least. That's pretty impressive, actually.

The switching energy specs are also impressive, but I would still limit switching frequency to the 12-16kHz range.

And one other note, the Cree gate driver uses an isolation IC from Infineon that is often in short supply (1ED020I12-F2).
 

·
Registered
Joined
·
730 Posts
Discussion Starter #15
Ah, what's in a name?

Originally posted by Sonikaccord:
Fault protection from over current situations and gate drive errors can be observed by a microcontroller.
Correct.
The driver board drives two optocouplers to signal undervoltage and desaturation FAULTs to the the microcontroller.

Originally posted by Sonikaccord:
Would a sine wave drive really increase motor efficiency?
Absolutely true. High frequencies cause added iron core losses, eddy current losses and skin effect losses. They also cause extra wear on ball bearings and stress on the wire insulation unless precautions are taken.
 

·
Registered
Joined
·
730 Posts
Discussion Starter #16
PWM frequency and sine filter

Here's a link to a comprehensive inverter output filter design guide:
www.danfoss.com/NR/rdonlyres/27F81E71-3779-4406-8EA0-849044873F59/0/Output_Filters_Design_Guide.pdf

In the guide Danfoss recommends the use of either dV/dt or sine filters in applications with frequent regenerative braking.

I've done the first calculations for the filters. The results are looking good for a low pass (sine) filter.
If the PWM frequency is high enough (say 50kHz) I'm pretty sure it won't be too difficult to make a coupled filter inductor with a relatively small and low cost 110mm iron powder toroid.
Also needed: three “DC LINK” rated 50 nF (600V / 20Arms ripple) output capacitors. Gives the filter a cut off frequency somewhere around 3kHz. Installing two filters in parallel should keep them cool enough for the 100kW peak power output of the “reference” Siemens Azure motor.
Total material cost estimation: only US$250 per filter. In comparison, dV/dt filters aren't that much smaller or cheaper.
 

·
Registered
Joined
·
33 Posts
Re: PWM frequency and sine filter

Nice article Tony!

So I see the benefits of a LPF in motor drive applications. Are the benefits the same for induction motors and PMSM? I would think that the induction motor would benefit more from this design.

How great is the tradeoff between motor efficiency and filter efficiency?
 

·
Registered
Joined
·
730 Posts
Discussion Starter #18 (Edited)
If cost isn't an issue, filters can have very low losses. But there's always a tradeoff. Weight, cost and performance.

For instance: the low cost filter needs a high PWM input frequency. Obviously, the low cost is a plus.
The tradeoff is a slight increase in switching losses in the SiC. But the filter copper losses are lower.

Estimates for the reference motor and a single filter:
Filter magnetic core losses: 200W max (at low speed)
Copper losses: 40W at nominal load, 160W at peak power
Dual filter:
Filter magnetic core losses: 100W max (at low speed)
Copper losses: 80W at nominal load, 320W at peak power

The core losses depend strongly on the magnetic properties of the toroid.
200W means that the toroid needs forced cooling with a peltier element. So I prefer the dual filter solution.

PMSM: the effects on windings, iron and ball bearings are the same.
Effects on the PM: don't know, but for DIY: you can only get it wrong once (demagnetize).
 

·
Registered
Joined
·
730 Posts
Discussion Starter #19 (Edited)
The next version of the PCB design for the gate driver is attached.
With component values and device identification.
If all goes well, I'll post a picture of the assembled PCB next week.

UPDATE (2020): IMAGE DELETED, cascode SiC JFETS will be used
 

·
Registered
Joined
·
730 Posts
Discussion Starter #20 (Edited)
Here's the picture of the assembled sic gate driver board as an attachment.
The components are in DIY size. Made with a cheap phone, so it is not very sharp.

The next step is the filter.
A measurement must be done to know the effect of the core losses of the toroid.
Amidon material 26 is the reference material. It's iron powder for choke and filter applications.

As a reminder:
Recently EV regulations include EMC tests.
No wonder, the advancing technology in power electronics leads to ever faster switching and higher frequencies.
The filter priority is easier EMI/RFI compliance. The other effects are bonuses.
 
1 - 20 of 118 Posts
Top