100kW AC Inverter Project (Opensource)

[Mesuge]

The following thread contains important discussion
and source files about ~100kW AC inverter (opensource project)
from canadian university:

viewtopic.php?t=161&start=1

----
OK, I'll repost the summary here, not to go offtopic on that thread..
----

"The rated continuous output current from the inverter is 200A (RMS). This allows up to approximately 96kW output when running from a 230 to 480VDC voltage in a maximum 30°C (85F) ambient temperature environment."

There seems to be fully documented opensource 100kW AC inverter project(for traction application) on the web. They won the prize from the CircuitCellar/Microchip for 2007. It has PID functions, regen, CAN bus communication. Several people tried to contact the authors for more details, but so far no replication/verification has been published after their original release. In any case you can download both the abstract and documentation here:
http://www.circuitcellar.com/microchip2 ... T2291.html

Few more details from the press:

http://www.lookoutnewspaper.com/archive ... 03/2.shtml


Running on E – for electricity/OakBay News


This is the university dept. mentioned in manual:

http://www.elex.camosun.bc.ca/home/index.htm

Some discussion has been ongoing on this topic at the following forums as well:
http://www.diyelectriccar.com/forums

I think that mere mortals might eventually crack that holy grail aka diy 50-100kW AC inverter, such a project would certainly help the conversion scene by a huge margin (the AC Tri-Zilla won't be out for years)!

But leave those <200kW bidirectional marvels to ACP's Alan Cocconi and the young Tesla guys, so keep it simple and affordable.. Thanks!

[Benonymous]

Excellent post Mesuge! Always interested in this sort of thing

[Mesuge]

Just to bump up or refine this important thread, hah.

There seems to be at least two independent replication projects of that CircuitCellar/Microchip AC drive design currently under way. They are trying to refine it also from the angle of safety and better user control options (usb comm., regen, driver controls, bms, etc.)

These two guys go under the following IDs: "Madmac" and "Wirecutter" :
http://www.diyelectriccar.com/forums/member.php?u=52
http://www.diyelectriccar.com/forums/member.php?u=842

--
It has been discussed in a bit disorganized fashion in these two threads (the important talk starts aprox. from pages):


http://www.diyelectriccar.com/forums/sh ... 402p5.html

http://www.diyelectriccar.com/forums/sh ... =microchip

[Mesuge]

The opensource AC 100kw drive design originaly published in the
CircuitCellar mag has been updated for their latest August 2008 issue
(available online), also thanks to popular demand and questioning of
the authors by the EV community.

The article is here:
http://www.circuitcellar.com/archives/v ... index.html

Info/code package downloadable:
ftp://ftp.circuitcellar.com/pub/Circuit ... er-217.zip

[AMPrentice]

This is a great project and if there was a step by step it will surely put many EVs on the road.

Now we need a group buy for someone to make them for us. Maybe a New Venture could start for someone, a name like King Ghidorah (the alien beast with 3 heads for 3 phases) would rival a Zilla

[TimF]

Definitely interesting, something I'm going to look at. I like the idea of modularity, using the same controls for different size power silicon to allow for units to be built to suit individual needs without a complete redesign.

[iwyze]

This sounds like a project with my name on it.
Currently I am comparing this "DYI" against modifying a industrial VFD to EV.

Right now I am busy working on a GSM/3G telemetry system, so should have time to start this over Christmas. Initial trials on a test bench will be required until I can no longer extract smoke from silicon.

The goal is to build a consolidated unit that handles most of the standard EV requirements:
· Variable Frequency Drive VFD
· Battery Charger
· DC to DC converter
· Liquid cooling to keep the size down and seal the system.

Other desirable ideas are:
· Telemetry
· Battery Management BMS
· Communication to a separate HMI like the ‘Toyota Prius’

Have I missed any typical features required on a EV conversion that are useful to keep a modern car running? Also would anyone be interested in pooling know how?

[Benonymous]

Most EV inverter/controllers have the CAN bus. I believe this open source project has CAN. Anyway, that's the only thing I think you left off your list. The big cost with this project is going to be the Powerex IGBT modules. The last time I checked, the ones used in the documentation cost around $495 a pop and you need three. You also need three of the bespoke driver boards but these are a minuscule cost compared to the IGBT's.

[Johny]

Hi iwyze. I went a fair way down the path of a DIY controller. Lots of parts prices, annoyed lots of sales guys worldwide and gained a good understanding of where the smoke was going to come into it. Vector space control became my sticking point - I realised that I was never going to get the time to understand, let alone code, it (I am in embedded software by trade). The 100K open source project has some fairly large holes in it mainly in the area of the DC Bus, and layout instability.

Anyway - I ended up buying a slightly undersized industrial VFD (Lenze 16kW) that I intend to modify if it doesn't give me enough torque.
See http://electricvogue.blogspot.com/
My current path is to try it as is then (an idea about 1 day old) switch from star to delta at some particular RPM so I get lots of torque at low revs without the requirement of a large current controller. The Lenze has built in editable 'function blocks' so lots of EV specific stuff can be done. Eg. Brake lights when generator current exceeds a certain value.

In my DIY, I was only targeting 100 Amp continuous but will be happy to give whatever info may be helpful. Powerex was always the MOST expensive silicon - there are others but it depends hugely on current required. I found Gate drivers to be quite significant in the cost.

[bga]

Hi All,

Some thoughts on building AC inverters:

Johnny:
The problem with upgrading the power semis is that the gate drive reqirement increases, so the controller may not be able to drive the IGBTs. Also, the bus filter capacitors won't be up to the incresased current. I would worry about the controller being thick and not very useful for an EV.

New design notes:
I downloaded the source code from the Circuit cellar 100KW controller and ACIM source code from Mircochip. The two are about 95% identical.

I set up the DsPIC30F compilers on my PC and compiled the source code. It produced a plausible object with only a few compiler warnings.
The source code looked to be in fair condition and relatively easy to extend.
This sounds to be useable.

I have been designing a new main controller PCB, using the Circuit Cellar project and the motor controller development board from Microchip as a base. These are nearly identical and have extra hardware, not needed in an EV controller, but lack some stuff needed for an EV project.

I want the controller logic to run from 12V, with a 10 to 40V range. Sufficient hold up to allow the controller to shut down safely.
Probably no braking resistor, if the batteries can't take it, use the hydraulic brakes.

Ignition, enable and reverse switches.
Throttle and brake pedal position analog inputs.
Brake light output.
reversing light output
Various dash lights for battery and motor
Serial I/O and CAN buss (this'll be an adventure) for additional functions.
Hope the DSP doesn't run out of computrons.


So far:
Powerex 200A/1200V/20KHz IGBTs are approximately USD$125.00 each from Richardson Electronics.
The 300A variety are only $10.00 more, but need a bigger gate drive, so may be the way to go. (This would guarantee a 100kW capability)

The Gate driver bits are approximately $20USD per IGBT (2 sets per module) from Richardson

Bus/Phase Hall effect current sensors (LEM HAIS 100-P) are available from Farnell for approximately $40 each these can measure up to 300Amps at 100KHz.

The link capacitor remains a problem.
I was considering an array of small aluminium electros (30 x 470uF/400V)
on each of the +300 and -300V legs. This should handle 60A RMS within the rating of the caps and be overload tolerant to 100A or a more. 400V sounds OK, the maximum bus voltage should be just short of 400V per side.
Cheap and reasonably compact if Altronics caps are usable.

Using 300Amp IGBTS and 600V buss should result in a 100KW controller.


cheers
BGA

[antiscab]

Hi BGA,

i doubt 300A rated IGBTs would do 100kw, even at 600v on the input.
you have to derate the IGBTs away from their maximums by at least a factor of 3.

300A is the absolute limit of a 300A rated IGBT.
so at the peak of the cycle, 300A is the limit = 212A RMS.
but 300A at the peak assumes the duty cycle is 100% (it wont be)

then theres di/dt and dv/dt to worry about.
How far you can push this depends on how much motor inductance you have to work with.
for a 100kw controller, you would be looking at 1000A or more components.

Matt

[antiscab]

hang on....i may have made an error here:
100A (141A peak) RMS on each phase would give around 72kw at 600v input (415v RMS out).

so not as far out of this world is i previously suggested.
i still think 300A rated IGBTs would be too small.

a danfoss 5042 uses 150A IGBTs, and its only rated to 40kva.

Matt

[acmotor]

Matt,
Just to throw in here, my VLT5042 uses 2MBI 150SC-120 Fuji IGBT. (three in 3 phase bridge)

2MBI 150SC-120 IGBT

IGBT app. notes

These are 150A (300A 1ms) devices.
The 5042 is rated at 97.6A peak current(69A RMS)(60 sec) per phase motor current so these are derated to 60%. (~48kW at motor).
Peaks are still less than 150A and well less than 300A.

On this basis 300A IGBTS should handle 100kW for 60 sec.
Industrially VFD manufacturers would not choose this as a continuous operating level though as ambient / heatsink temperature would be an issue. Motor would cook as well !

The difference between single and parallel devices is that single devices can work a lot closer to design load and parallel devices must de-rate considerably. For 10 in parallel I would de-rate to 40% or less if you wish it to be reliable.

[antiscab]

Tuarn,
does danfoss have a suggested minimum and maximum motor size for their drives?
do they mention a minimum motor side inductance?
the derating would probably have to be increased if there isnt enough motor inductance for continuous current on the motor side at the switching frequency.

the only ev intended 3-phase controller ive ever opened up is my scooters one, and that uses 200A hexfets and only puts out 40A RMS/phase (57A peak). Im guessing the massive overrating has to do with the lack of motor inductance due to the relatively few number of turns in the field coils.

why would the motor cook aswell?
does the harmonic distortion on the output waveform become more significant at higher loading?
or is it because the motor is operating beyond rated (depends on motor?)
both perhaps?

BGA - i stand corrected.
would 400v caps be sufficient?
assuming a 110 lifepo4 cells on each side, just after a full charge, you would be seeing 418v. i know from my scooter controller that caps will deal with short term repetitive over-voltage fairly well (i use 63v caps at 57-60v underload, 63v under no load, and 70-73v just of a charge for the past 10'000km).
im guessing your not aiming to be able to run 415v motors at higher than rated voltage?

Matt

[acmotor]

Danfoss suggests to choose a VLT ... "the rated output current Ivlt,n must be equal to or higher than the required motor current" ... and "one frame size larger than nominal motor for VLT model provided that Ivlt,n is not continuously exceeded"
Inductance being very much a function of the frame size. The point here as you say being the inductance has a minimum requirement, you can't just rely on current limiting. ( Applies to DC systems as well - picture a certain curtis on an 11" )

I have tried oversized motors on controllers and they will not do automatic motor adaption (they report motor is oversized when they try to test reactance). If you run it anyway, you will get VLT overload alarms. Probably di/dt generated.

This is probably of no interest to EVs as it is usually the other way around - big controller - little motor.
This combination works without limits.
I have run a 300W 3PIM off my 5042. It works fine and so it should. AMA does not work below 5.5kW on this 30kW drive. It says "what motor ?" but runs Ok.
The danger is that you cannot set the drive up in calculated thermal overloads etc if the motor is too small, so you need to be careful as it will push the load dependent v/f and saturate the motor, overheating it. You can set torque limits from 1 to hundreds of percent and also set current limits so you can make it fairly safe.

When a 3PIM is running at less than its shaft fan design speed or at greater than nominal torque there is risk of overheating. This is why many VFD driven 3PIMs run external fans (as I do). There is plenty of thermal mass (80kg of copper, iron and ali in my case) so overloads can be for 1 to 10 minmutes without heat problems. Industrial is very conservative in the overload area.

Harmonic distortion is less at higher switching frequencies (I use 14kHz) as the waveform approaches sinusoidal, however switching losses are greater. (1% ?) So heat loss in controller rather than motor.

Danfoss runs 4 x high ripple 450V 3300uF caps in a 2 x 2 in series arrangement with 15k equalising resistors. DC bus is speced to go to 850V on regen.

[Johny]

My 2 pennies worth here. My understanding is that running electolytic caps over their rated voltage reduces their life, in fact the further down from rated voltage you get, the longer the life. The Lenze VFD only has 400 VDC caps making a total of 800V. With Lithium batteries it will be tight to get operating voltage then keep it under 800V during regen braking (assuming max of 4.2V per cell). I would planb for 450 V caps.
BTW bga, the Lenze VFD I have is in the same case as the 2 higher power rated versions and the gate drivers appear to be standard for the 16, 22 and 30kW - I may not need to touch it anyway as I am happy enough with woody's figures on my projected performance.

Edit: DC bus is speced to 760 V on regen.

[bga]

Hi All,

The bl***dy keyboard. I've lost this post 3 times now.
I've resorted to typing it in a text editor and CPing to the browser.

IGBTs and drive power
---------------------
Johny, I was a bit concerned after I found out how much drive these things need.
The literature suggests 1.25A of gate drive for every 100A of IGBT. Upping the IGBTs would also require the gate resistors be revised to accommodate the new devices. It looks like the gate resistors are about 2 ohms for 400A devices.

I haven't gone looking for snubbers yet.

I did look for some big IGBT modules and found Powerex CM400DY_24A, 1200V 400A 20KHz for about $185 USD each (240AUD) at Richardson Electronics. I followed the link from powerex's site (www.pwrx.com)
There are other brands that may be better value. I'm not tied to Powerex yet.


Electros and bus volts
-----------------------
Mostly electros can be surged above the rated voltage, but it can shorten their working life. I would expect that an EV with nominal 600V batteries shouldn't do this very often. Temperature is the main issue with working life in alum. electrolytics. Power capacitors exhibit heating because of resistive effects from the ripple current on the devices. Hard working caps may need active cooling to keep them happy.

I would expect that for the most part, the EV won' be drawing a lot of current. the 600V bus means that every 20 amps is 12kW. It would be likely that the bus will be in this region most of the time, with occasion excursions to 200A+.

I was suggesting using 400V caps because I can get these without thinking, although they aren't ideal, or even good. Altronics 400V/470uF have a lousy ripple current, so a large number will be needed, probably 50 or 100. I'm definitely interested in sourcing a better solution.

I located some super-duper film capacitors for link buses. They are a bit low in uFs but great otherwise (Electronic Concepts Unlytic) (300uF/1200V/2500A peak)
Also kind of expensive.

Tuarn, could you look up the manufacturer and part number for the Danfoss controller's caps, please. I'll have a go at sourcing some similar ones.


Parts sourcing
-----------
The brain (DsPIC30F6010A) has arrived, one small part of the solution. I ordered a small number of these from Microchip, they took 4 days to get here from Thailand. On with the man PCB, along the lines of the PIC motor control board and the Circuit cellar project, which are nearly identical.

I was reading the application note on the Microchip ACIM application. Running the control loop at 20,000 cycles per second, it only uses 9 Mips of the 30 available on the processor, there should be plenty available for housekeeping, and communications with peripherals etc.

The intention is to prove up the basic controller using a low current/low voltage driver initially. This should minimise the possibility of burnt fingers and smoke.

Cheers

BGA

[acmotor]

I can't see a brand on my Danfoss's caps. BIG blue and shinny ! I'll check out some other VLT units.

FWIW though, the cap bank only needs a low ESR at 5-20kHz i.e. switching frequency and not at the 50-300Hz that the cap bank is needing to deal with when running off the 50Hz mains.

The battery bank is a good solid low ESR supply (milli ohms) (a lot lower than caps) at the lower frequencies. Although the fuses, wiring and control gear add some inductance and resistance. These are bad on the DC bus side of the IGBTs.

I have seen people remove the cap bank but this is not smart as the switching back emf and regen pulses need a close at hand high frequency low ESR place to dump so as to limit voltage spikes on the IGBTs. But overall uF can be less than in 50Hz system.

I have a set of 4 large caps I had trialled as a bus double up on a site with bad power. All I got was smooth bad power.
You are welcome to these if they are of any use to you. PM me.

[Johny]

The Lenze 9327 (15kW) uses 400V 4700uF times 2. There is no way they could shoehorn any larger ones in there so it's my opinion that they also use these for the 22 and 30kW model which comes in exactly the same case size.
These caps have about a 80mm 24 volt fan especially for them - it must be thermally (or current) controlled because I haven't seen it come on with my bench set-up yet. My pricing on caps was that around $150-200 buys you a 4700uF 450V.

The Lenze main power board has slotted PCB holes for the IGBTs which currently phisically measure the same size as Fugitsu and Mitsubishi 100A and 150 Amp IGBTs. It looks like I could fit 200-250Amp ones in there but I would either have to fgiure out how the control board identifies the power board (some kind of data exchange or strapping option) or "lie" to the current sensing. I'll just leave it for now as it only effects acceleration at low speeds. The Lenze is physically about the smallest controller I have found - it only weighs 12-15kg.

Anyway.
Gate Drivers. The trench IGBTs have lower saturation levels - about 1.8 volts as opposed to 2.8 for the "regular" IGBTs but the payment is that they require larger Gate drive as they have around twice the gate capacitance. From everything I've picked up on, the 2 biggest causes of smoke will be inadequate Gate drive and poor bus layout. I can send pics of the Lenze layout if you like but basically treat it like RF with the snubbers right on top of the IGBTs. Bus bar to the big caps.
I was planning for about 5-8 Amps of Gate drive for an expandable design with only 2-3 for a 100Amp IGBT. Have you noticed that IGBT ratings are always at 20 degrees C? I would go for about 50% higher than max. planned current. Pricing was about USD$75-100 for the 100 Amp CM100DU-24H or similar via Alibaba.com.

Also. When you mentioned 20 A at 600V being 12kW, that's not where the most ripple will be happening. Someone comment here but I would think most ripple will be at highest motor current which will be when running at lower than rated motor speed.

I searched long and hard for Gate drive IGBTs and the best bet for my 100Amp was from this place:

ANSWER TO ONE OF MANY EMAILS
-------------
John,

No problem.

Presumably because you have the 600V DC source you just need the IGBT chopper.

Please consider the following;-

Eupec BSM100GD120DLC, 6 pack IGBT,        Price $195.00 ex GST, ex Melb store.
Data attached. Ex stock.

CONCEPT Driver,               1 x 6SD106EI        Price $290.00 ex GST, ex Melb store.
Data attached. Two weeks delivery

Unfortunately small quantities are quite expensive because of the freight.
However, in qty pricing we can get down to around $150.00 for the 6 channel Concept Driver.

Regards,
Mike

Mike Trubridge
Fastron Technologies Pty Ltd
25 Kingsley Close
Rowville
Melbourne
Australia, 3178
Tel: 61-3-9763-5155
Fax: 61-3-9763 5166
Web: www.fastron.com.au
-----------------------------------------------

They also has a complete bridge but it was $2500 (without ripple caps).

Got to work now but I have bucket fulls of files, specs etc. on all this if it helps.

[Johny]

Lenze power pcb layout. Caps have been removed for this photo - they connect via copper strips to the large solder pads nearest the viewer.
The 3 phase input is the middle bolts on the left. DC Bus input are the outer bolts on the left with protective ground stud nearest. Motor output are the bolts on the right. Another protective ground nearest viewer. You can clearly see that only one IGBT (the far one) goes directly to the stud. The other 2 are current sensed.

The black wires are gate driver connections and you can see the 2 blue snubbers mounted above the IGBTs. They mount the IGBTs on the heatsink then mount the power pcb on the IGBTs.

The controller board stands up at the back.

[Johny]

Lenze cap. You can also see the Lenze heat sink on the left and the huge surface area they try to get with fins about 7mm apart. Two 24 VDC fans (on the other end) cool it - again they have never turned on with my test set up.
Thank you to my EV sponser (I wish).



The caps and their fan awaiting remounting. They are sitting in the cover so you can see how large they are with respect to the controller.

[acmotor]

The caps in Danfoss VLT 5042 are 4 x 450V 3300uF as here....

[Johny]

Where's the soft drink can???

Danfoss total of 6600uF per side for 37kW rated as against Lenze 15/22/30kW 4700uF per side. It all hangs together since ripple ratings vary a bit.

[bga]

Thanks for the info,

BTW The capacitor specs are:

EPCOS:
type B43564-S9478-M3
450V
4700uF
ESR (10kHz) 25milliohm
26.3 Amp ripple current (base cooled)
gets an over-rating of approx 1.4 at 10kHz (40 amps).

Nippon Chemi-con:
Looks like the numbers may have changed since it was made. the current ones are black.
A similar component from the lineup is

    ERWY451LGC332MEF5M
    (RWY450LGSN3300MCEA)
    Screw type, RWY, Polar
    450V
    3300uF
    76mm dia, 155 length, 17.7A at 85deg C,

Their 4700uF parts have a lower ripple current in the 85 dec C range.

A series-parallel network of 4 x 4700uF 450V caps will result n 4700uF at 900 volts.

My sleazy array of little electros doesn't sound all that bad, but would need a PCB and 50 of them.

BGA

[Johny]

bga
Here are the specs on the IGBT bridge from Fastron. This is all 3 devices in one package. It looks to be a really good price for all 6 transistors in one package. 160 Amps at 25 degrees, 100 Amps at 85 degrees.

uploads/293/BSM_100_GD_120_DLC.pdf

The Gate driver is a total 6 gate driver and looks good also. I couldn't upload the specs for some reason. If you are interested in the Gate driver (or the above) let me know. The guy as Fastron is pretty easy going and happy to help.