L6506/IRL540 Unipolar Chopper Drive

[JonLed]

Hello forum,
I've been working on a unipolar chopper drive based on the L6506 chopper chip and IRL540s.

When using "freewheel diodes" there is no motor torque, but if I remove them the motor runs great, so I'm not using any for the time being. I'm currently waiting for some 24V zeners to put between the "freewheel diodes" and Vmotor as suggested in another forum. What do you think about this suggestion?

My Fets are running a bit too hot I believe (135*F), I've done some experimenting with fet drivers to make sure that they were driven to full enhancement, but that did not reduce the heat any. I think this heat is caused by not using external "freewheel diodes" and allowing the internal diodes to absorb the kickback. Is this correct?

Schmatics:
http://www.8052.com/users/jonled/L6506-1.GIF
http://www.8052.com/users/jonled/L6506-2.GIF

Datasheets:
http://www.st.com/stonline/products/...1392/l6506.pdf
http://www.diodes.com/products/inact...ta/ds26003.pdf
http://www.irf.com/product-info/data...ta/irl540n.pdf
http://www.orientalmotor.com/product...XTA/StPk26.pdf

Thanks

Jon

[pastera]

135°F is not hot - assuming they are not heat sinked, the fets are only dissipating about 0.45W. Using the Rdson (0.063ohms) the current would be about 2.7 amps

The no torque problem is related to your D10's - Example
When you turn on phase 1 - motor lead 1 is pulled to ground and current starts flowing - at the same time the current induces a voltage into the other half of the coil that raises lead 3 to 5.7 volts, turning on D10. The two magnetic feilds fight each other effectively canceling each other out.

Remove the D10's and you will get your torque, you also don't need D9 as the body diode of the IRF540 can easily handle the freewheel current.

Aaron

[JonLed]

Aaron,
Thanks for the reply. The IRLs are heatsinked. The temp of the heatsink is about (150*F)

Jon

[pastera]

Looking at the datsheet for the L6506 - the output impedance is about 3.1K (1.25v drop @400uA). Your circuit cannot turn on the FETs fast enough to meet the 30KHz switching frequency.
Try using a TC4468 to drive the FETS
http://rocky.digikey.com/WebLib/Telc...4467,68,69.pdf

[JonLed]

I did try the TC4468, it made no difference in FET temperature or anything else for that matter. Sorry about the schematic, its now chopping at 20KHz.
One thing that bothers me is that on every stepper drive I've ever seen, they all use "free wheel diodes". Is this not the recommended way?

You suggested not using any diodes at all. I've not yet tried this, but it seems to contradict everything I've ever read. I will try it tomorrow.

Everything is working good with the drive except the temperature of the FETs, I've been told (and read) that they should not get warm at all. That's basically my concern at this point.

Jon

[pastera]

Do you have access to an O-scope? If so, look at Vds and Vgs.
Take a look at Figure 6 of the FET datasheet, this shows the charge needed to turn the FET on. Your circuit simply can't turn the FET on fast because it does not have a high enough voltage to drive the gate. The majority of the charge must be delivered with the gate at 3 volts - this leaves 2 volts to push the current through the driver impedance and the gate resistor leaving you with a current of about 600uA. Under this condition it will take 61uS to turn on the fet, or longer than your switching cycle. Plus, the .33ohm sense resistor works against turning the FET on! Use the TC4468 with a Vcc of 12V and you will see a large improvement in your switching times.

Also, make sure you are not turning on both phase 1&2 or 3&4 at the same time. It looks like you are using a small microcontroller to drive the phase lines - makes sure you wait at least the calulated turn off time between turning one phase off and the other on.

Aaron

[JonLed]

Ok, so what I'll do tomorrow is remove all my diodes, put the TC4468 back in with Vcc of 12V (I used 8V before, I was told not to exceed 10V because it was a logic level FET).

I'm sure that phases 1&2 or 3&4 are not being turned on at the same time.

I'll have to check on the timing (calculated turn off time?) is this the turn off delay time of the FET (39nS)?

Thanks
Jon

P.S. Yes I have a scope

[pastera]

Originally Posted by JonLed Ok, so what I'll do tomorrow is remove all my diodes, put the TC4468 back in with Vcc of 12V (I used 8V before, I was told not to exceed 10V because it was a logic level FET).

Not a problem - the gate is rated to 16V, so 12V gives you a 25% margin.
But, after looking closer at the data sheet going over 5 volts gives very little for the extra gate charge. Go with the driver with VERY short leads to the gates and gate resistor

Originally Posted by JonLed I'm sure that phases 1&2 or 3&4 are not being turned on at the same time.

Just because you signal controlling the fet does not overlap does not mean the fets are not overlapping - it takes some time to turn the FET on and off.
Read an-944 from Iternational Rectifier
http://www.irf.com/technical-info/appnotes/an-944.pdf

Originally Posted by JonLed I'll have to check on the timing (calculated turn off time?) is this the turn off delay time of the FET (39nS)? Thanks Jon P.S. Yes I have a scope

You can calculate you switching times or simply measure them.
Trigger off of your P0.x signal and look at both Vgs and Vds. The Vgs will not rise to the drive voltage (IE it sits at the miller plateu) until the FET is switched on - opposite is true for turn off.

The turn off delay is not your switching time. You should shoot for <5% switching time or
20KHz = 50uS *.05 = 2.5uS switching time
This means you will need to move ~40nC of charge into the gate in 2.5uS with the gate at 3v and your drive at 5 volts.
So Amperes = Coulombs/Second Therefore:
Ig = 40nC/2.5uS = 16mA
Rg = (Vdrive - Vgate)/Ig = (5-3)/0.016 = 125 ohms

The output resistance of the TC4468 at 5V is ~20ohms so your 100ohm gate resistor is fine. The above does not take into account the total switching times but is close enough.

Aaron

[JonLed]

Ok... I removed the external diodes and replaced the TC4468 Vcc=12V. Yes the gate signal looks much better, but the FET temperature is still 125-130*F. Is there any way to get this down?

On the switching times - Is this what you mean?
I output 0110 on the 4 port lines then output 0000 and wait 2.5uS before I output 0101.

I tried this, but I couldn't tell that it made any difference.

Jon

[NC Cams]

Try using Schottky diodes for the shunting diodes in place of fast recovery type.

If you drive with 12v, remove the 5.1v zeners from the gates. Actually, 10V is plenty.

The 100 ohm gate resistor may be fine - try one of a bit less resistance. We had better luck driving higher amperage IRLZ44 fets with 10 ohm gate resistors.

Try finding a fet with a lower ON resistance in the same or larger die size having the same voltage rating. The more the fet resistance, the more heat it will have regardless of how hard and fast you charge the gate.

Try driving the fet with a honest to goodness fet driver. We solved heating problems on IRLZ44's by merely using 10v gate but a STIFF driver instead of driving directly off of a logic level pin.

Try bypassing the fet gate series resistor with a 1N914 so that the fet shutsoff off faster than it turns on. Like this:

drive -----l<------- fet

I left out the resistor but you should get the idea.

Try adding a 10K pull down from the gate pin to ground - insert between series resistor and gate pin on fet. This makes sure that the gate won't self oscillate as well as makes sure the fet stays OFF. This should NOT be needed if you drive with a real, honest to goodness fet driver.

[pminmo]

Your fets are really capacitive and the l6506 won't drive them well, result hot fets. Freewheeling problem is probably diode selection.
The TC446X is a much better match for your fets. i would pull the zeners, probably not fast enough to be of much help anyway. with the tc driver 100ohms may be unecessary or too high in value.
Make sure your physical layout isn't contributing to the problem.

[pastera]

Originally Posted by JonLed Ok... I removed the external diodes and replaced the TC4468 Vcc=12V. Yes the gate signal looks much better, but the FET temperature is still 125-130*F. Is there any way to get this down? On the switching times - Is this what you mean? I output 0110 on the 4 port lines then output 0000 and wait 2.5uS before I output 0101. I tried this, but I couldn't tell that it made any difference. Jon

Only put dead time in the phases that change 0110 > 0100 > 0101

Did you measure your switching time? This is the critical factor not knowing this everyting else is a guess. You should be able to see the same wave shape for Vgs as in AN-944 from IRF.

Also, make sure the gate charging loop is as short as possible - Q4,Rg,Rs and TC4468 mounted close together with minimum lead lengths.
Large bypass caps for the TC4468

Aaron