[tivoidethuong]

@ JBV great desgin PCB

[kreutz]

IR2104 has 520ns dead-time, why would they have deadtime if it didnt work?

There are two dead-times, dead-time for the internal mosfet driver's totem-pole output and dead-time for the power stage high side and low side Mosfet switching.

They (International Rectifiers) don't know what Mosfets/IGBTs you are using or what PWM frequency you are going to use. So, how do you expect that a fixed dead-time will work for everything you could possibly drive?

Do you know why the designer of the UHU board used two resistors, a capacitor and a diode in the gate drive circuit? Because he knew he had to control turn off and turn on times in order to avoid cross-conduction. Dead-times for servo control power stages are often in the micro-seconds range. If you want more control over the dead-time, then you control it in the firmware by using the two output compare registers on one timer to automatically set a dead-time, that way it does not consume CPU processing time, otherwise you have to do it in hardware.

Note: I don't have any relationship with the UHU board's designer, The UHU board is used as example because the schematic is well known.

By the way, how will you control motor current?

[JBV]

why would i need to with big motors?

[kreutz]

why would i need to with big motors?

Lets forget about torque control for a moment, what happens if your motor stalls? You just let it burn itself or blow the fuses? What about overloads?, will you let your motor demagnetize itself? It does not matter how big is the motor, the bigger the motor, the bigger the potential damage is if something goes wrong.

[JBV]

You dont have to buld the driver! problem solved?

[kreutz]

You dont have to buld the driver! problem solved?

Don't get mad, it is not my intention to discourage you or anybody else, a lot of people are going to see your schematics and assume they can be built safely.

I just want you to improve it so it can be safely assembled and used by other people too, I assume you began this design in order to develop your own solution, and I always encourage that, there is a lot to learn along the way and that is also good.

Think about this: How good is our own solution if is not good enough to protect our own motors, driver boards or power supplies? If you are aware of the risks, then OK go ahead, but I feel that I should at least point them out, just in case you haven't realized yet.

Since you did not like my comments I will not post any more comments on your boards anymore. Remember that others are potentially going to create unsafe situations by trying to reproduce them, and that is your responsibility too, they are taking for granted that they are safe.

Regards,

Kreutz.

[JBV]

Sorry if i got mad! :P I know it's a minimalistic driver, i like to build small and cheap! I belive i can build one of these for ~$30

If i could program ASM i could easily add current limiter. I will have that and other features on my own driver later!

[kreutz]

Sorry if i got mad! :P I know it's a minimalistic driver, i like to build small and cheap! I belive i can build one of these for ~$30

If i could program ASM i could easily add current limiter. I will have that and other features on my own driver later!

Thanks,

Kreutz.

[Xerxes]

Kreutz,

I think 500 ns deadtime is enough for almost any modern MOSFET. I have built several drives with IR chips and mosfets without special deadtime controls and it seems that even less than a half of that deadtime would be enough.

Limiting turn-on time is a way to reduce the current spike from opposing MOSFET body diode reverse recovery and has nothing to do with deadtime duration. Turn-on transient is similar no matter if deadtime is 500ns or 10us.

[smarbaga]

hello:
a few resistors, capacitors and or diodes is a small price to pay to ensure that both positive and negetive transistor turn on/off states do not overlay.
it usually takes 2 to 4 pc board layouts to get things working the way they should ( if you are lucky ).
but we do get ancious and look at our just arived boards and say
" O my, isn't that a beatuful thing. " :-)

[kreutz]

Kreutz,

I think 500 ns deadtime is enough for almost any modern MOSFET. I have built several drives with IR chips and mosfets without special deadtime controls and it seems that even less than a half of that deadtime would be enough.

Limiting turn-on time is a way to reduce the current spike from opposing MOSFET body diode reverse recovery and has nothing to do with deadtime duration. Turn-on transient is similar no matter if deadtime is 500ns or 10us.

Dead-time has nothing to do with dV/dt limiting, as you pointed out. But everything to do with cross-conduction. It all depends on mosfet driver maximum current capability and Mosfet gate total equivalent charge including miller effect. Layout plays a role due to the inherent RLC resonance that tends to make the gate current to oscillate at high frequency (ringing effect).

Switching times have to be carefully controlled and tested at the higher operating voltages (due to the miller effect) and that is only possible if you include different paths for charge and discharge of the gate capacitance or include an adjustable dead-time (firmware or hardware) on the input PWM driving waveform for upper and lower drivers on the same leg. Otherwise you are relying on the manufacturer's specifications (bad habit), component tolerances and plain Luck.

BTW: most of the manufacturer's specs are figures dependent on the test fixture/parameters, mostly different to real life circuits. Never rely on spec figures without a minimum-maximum range specified. (In our example, IR2184, it is important to check also the matching figures for low and High side dead-times at turn on and turn off).

On the (IR2184) data-sheet, If you check the plot: dead-time vs temperature there is a range specified; typical around 400 ns, minimum around 300 ns, and maximum around 500 ns. But it is important to notice that sink and source currents depend on Vcc and temperature too, so will the switching times of the driven Mosfets.

There is also the issue of component change. Most of the people, without deep electronics knowledgement, think that it is safe to replace components when they find another with similar specs and higher voltage/current capability (for example), and that makes the later solution (rely on manufacturer's specs and Luck) more risky.

[Xerxes]

Kreuz, you are absolutely correct that gate drive shouldn't be overlooked. I have the first hand experience in MOSFET drive difficulties If interested, see: http://www.cnczone.com/forums/showthread.php?t=23216

But I still don't think that deadtime is the worst problem as long as it is long enough. If 300 ns is enough for a certain half-bridge with adequate safety margin, then IR chip internal deadtime will do. I think as short as 100 ns would work for fast mosfets (of course that should be verified, not assumed). Some 200 VDC class D audio amps use switching freq. of about 1 MHz. Anything near 500 ns deadtime is not possible there.

[kreutz]

Kreuz, you are absolutely correct that gate drive shouldn't be overlooked. I have the first hand experience in MOSFET drive difficulties If interested, see: http://www.cnczone.com/forums/showthread.php?t=23216

But I still don't think that deadtime is the worst problem as long as it is long enough. If 300 ns is enough for a certain half-bridge with adequate safety margin, then IR chip internal deadtime will do. I think as short as 100 ns would work for fast mosfets (of course that should be verified, not assumed). Some 200 VDC class D audio amps use switching freq. of about 1 MHz. Anything near 500 ns deadtime is not possible there.

When using an output stage like this one, 50% duty cycle corresponds with null while higher than 50% duty cycle will make rotation in one direction and less than 50% will make the motor rotate in the reverse direction. A dead-band in the nano seconds will not avoid cross-conduction completely due to the fact than even at null, the power amplifier is sourcing motor current.

There is a trade in: power amplifier efficiency (better cross-conduction protection) against non-linearity in the gain function. It is not uncommon to see dead-bands in the order of 3 to 5 microseconds at a PWM frequency of 30 KHz.

Setting the dead-band to a minimum will provide maximum holding torque at the expense of higher power losses (due to some cross-conduction).

The problem will probably manifest itself with over-current and/or overheating of the output stage when using higher power motors, driving low current low voltage motors will allow for more cross-conduction (less efficiency but higher linearity) without excessive heating of the mosfets because most of the mosfets being used here are over-rated.

Thanks for the link!

[cbcb128]

if{0,0}=>{0,0}
if{0,1}=>{0,1}
if{1,1}=>{0,0}
if{1,0}=>{1,0}
?

[Russo1]

Hi Guys

For weeks I've been trying to get the right fuse setting.

My problem is that PonyProg does not "see" any of my programmers. They were all homebuilt, so I went and bought a STK500. Now AVR Studio is programming the chips fine, but I get nothing only the error LED blink, even if there is no chip on the board. I made no changes to the source files and just write them as is.

Reading the chip in AVRS 4 everyting is OK. Problem is that I don't know the fuse settings. I'll try and post a bitmap of the settings in AVR 4, currently it seems I will have to write the chips in AVR 4

If any of you guys have a working chip and AVR Studio 4 (latest build), I would geratly appreciate if you can just help me with the fuses.

Regards

Ruanb

[jameslee]

Who the person who succeeds?
I failed.
It probably is not above in source, it wants knowing.

[jameslee]

I failed.
It probably is not above in source, it wants knowing.

[kreutz]

Hi Guys

For weeks I've been trying to get the right fuse setting.

My problem is that PonyProg does not "see" any of my programmers. They were all homebuilt, so I went and bought a STK500. Now AVR Studio is programming the chips fine, but I get nothing only the error LED blink, even if there is no chip on the board. I made no changes to the source files and just write them as is.

Reading the chip in AVRS 4 everyting is OK. Problem is that I don't know the fuse settings. I'll try and post a bitmap of the settings in AVR 4, currently it seems I will have to write the chips in AVR 4

If any of you guys have a working chip and AVR Studio 4 (latest build), I would geratly appreciate if you can just help me with the fuses.

Regards

Ruanb

It seems that the fuses are OK. Check the Xtal you are using is the proper Xtal cut for that micro.

[dewayne]

I have for some time been keeping an eye on this thread and it is very interesting ihow the ELM trees are groing and sprouting out here.

Anyhow I am interested in building a servo system now to replace my 1200 oz stepper motor on my 4 x 8 router table. I really want speed lots of speed greased lighting type 300+ inches per minute in g0 travel.

Can anyone brake down as to iwhich Elm system is which based on perhaps MaxCurrent etc.

The original Elm is not big enough in power for me but I notice that there appears to b some spin off from the original design.

Are there any Eagle files anyoine is willing to share.

Thanks alot Guys,

DeWayne (UK)

[GuntisK]

Hi guys! ELM SMC servo controller looks quite good. Have already bought atmels, but only problem is high current driver stage. You see- I have DINAMO-SLIVEN servo motors which is rated 108V@18A. Someone mentiioned here to use IGBTs instead of FET transistors. But as I remember, there is a little differencies in controlling IGBT and FET. Or everything is the same? Have never worked with IGBT, so I would appreciate any help from You!
Guntis