[rlwoodjr]

Hello,
I am working on building a servo driver and was trying to use the FAN73832 as the mosfet driver. Does anyone used this chip successfully?

I plan on using the Arduino for the micro-controller. I have had limited success,, but I am getting closer.

The problem I am having with it is that once the motor power is hooked up (even without a motor), the Arduino crashes. The only pins that are connected to the Arduino is the ground and the inputs to the FAN73832.


Any help or comments are appreciated. I have attached a schematic for review.

[rlwoodjr]

I was thinking about using a hex inverter and only using one PWM channel on the Arduino. I am not sure, but it seems that the inverter might filter out the noise that locks up the Arduino.

Again any comments or direction would be appreciated....

[bharbour]

Hi,
I looked at your schematic and I think it would be a good thing to add some .1uF and maybe some .001uF ceramic caps on the motor voltage to ground. Make sure that the ones on the motor voltage go to your motor power ground near the power FETs. For the microcontroller power supply, put some as close to the mircorcontroller power pins as you can. Make sure that your motor ground current from the low side of the FETS has a separate way back to the power supply from the logic ground current. The inductance on the motor ground wire will allow quite a bit of noise that should not be seen by the microcontroller.

The larger caps are good to have, but electrolytic caps are only good for low frequencies (10's to 100's of KHz) while good ceramic caps are good well up into high frequencies.

BobH

[rlwoodjr]

Hi,
I looked at your schematic and I think it would be a good thing to add some .1uF and maybe some .001uF ceramic caps on the motor voltage to ground. Make sure that the ones on the motor voltage go to your motor power ground near the power FETs. For the microcontroller power supply, put some as close to the mircorcontroller power pins as you can. Make sure that your motor ground current from the low side of the FETS has a separate way back to the power supply from the logic ground current. The inductance on the motor ground wire will allow quite a bit of noise that should not be seen by the microcontroller.

The larger caps are good to have, but electrolytic caps are only good for low frequencies (10's to 100's of KHz) while good ceramic caps are good well up into high frequencies.

BobH

Thanks. I worked on it a bit this week end and found that the Dell usb ports were the reason the Arduino serial would freeze, but when hooked to another computer, the motor voltage noise is still a problem. I will add some capacitors as you have suggested. The encoder/step input code works without motor voltage, but once the motor voltage is applied, the Arduino interrupts count incorrectly and rapidly.

I do have a path from the mosfet ground back to the power supply, but this ground goes to the Arduino ground too. What did you mean by "a separate way back from the logic circuit"? How can I remove this inductance noise?

I really appreciate your comments...

[rlwoodjr]

Here is the board I have so far....just designed I am using a perf board for the testing.

[bharbour]

What I mean by a separate path back to the power supply for motor ground current and logic ground current is that I would run a separate, heavy gauge wire from your Motor power connector to the power supply negative. I would also have a separate power and ground connector over by the LM7805 that just powers the regulator and logic. The ground wire from the logic power connector would connect to the motor ground wire at the power supply negative pin. From looking at your board layout, it looks like you have set up for this.

If you get reliability problems in the processor, you might try cutting the ground trace between the 220uF cap and the LM7805 and putting a 10 ohm resistor in. I would move the connector for the 15V supply to the space between the caps and the lower gate drive chip. That way, the ground and the +15V to the gate drive chips and the ground and +15V to the regulator don't have to share traces. I might even use a separate connector for 15V and ground to the gate drive chips with the ground meeting the other two grounds at the negative supply terminal and the +15V meeting the other +15V wires at the positive supply terminal.

What kind of motor currents are you planning to drive? If the motor currents are over 5A or so, I would go to a double sided board and have much larger motor power traces. Also, for heavier motor currents, I would probably use optical isolators instead of the7404 for isolation. Opto's would allow you to separate the power and ground to the gate driver chips from the logic power and ground until they get back to the power supply.

If your motor currents are modest, it should work the way it is.

The power and ground traces from the regulator to the controller look pretty long and thin to me. That processor probably does not pull much current average, but the high frequency noise current may be pretty substantial. Short, wide, power and ground traces and a few decoupling caps (0.1uF ceramic) near the chip are always good. I can't say it will not work the way it is, but I have been bitten enough times with power and ground problems that I tend to route them first and work around them. I would call the decoupling caps essential.

Bob

[rlwoodjr]

I am planning to use 48 VDC Pittman motor and 500 CPR encoder. I bought some of them on ebay:

http://cgi.ebay.com/ws/eBayISAPI.dll...T#ht_500wt_928

at 50 vdc they have a stall current of around 5 amps. I have some 2 oz copper boards. I was hoping to keep it single sided.

I have thought about opto couplers, but thought they were too slow. Looks like you can get them faster than I realized.

Is there an easy calculation to figure out what capacitors to use or it it just trial and error? I do have a scope to see the noise...

[bharbour]

With a stall current of 5 Amps, you are probably OK with this layout for the power section.

There is no good method of calculating decoupling caps that I have seen. I try to put a 0.1uF cap per power pin, but on parts with many power/ground pins that gets impractical.

Most modern processors are fabricated with CMOS technology. CMOS does not require much current to speak of when it is sitting in one state. The current is required when you change states (1->0 or 0->1). What this means is that you get very short, fast current drain spikes. The inductance of the power supply network makes it difficult to supply these spikes directly from the power supply, so you put small capacitors close to the chips to supply these current spikes.

Unless you have a really fast scope, you can get an impression of the noise level on the power supply, but it is tough to measure the noise. I would add a .1uF cap or two to your power traces as close to the power and ground pins as you can and try it. If you get noise problems, you can go back in and solder a few more caps on the back side of the board or add jumper wires on the back side to fatten up the power and ground distribution.

With slower circuits like power supply filter caps, you can estimate your cap requirements reasonably well, because the speeds are low enough that real caps act enough like ideal caps to get by.

Bob

[neilw20]

With the circuit you have, you need to pulse the high side drive on a regular basis to keep ensure the bootstrap capacitor has adequate charge so that it all stays digital and never goes linear.
Do you intend to have a current limit in the circuit?
What are you doing about shoot through protection? Inductive loads can do strange things, and you must ensure the mosfets are always protected.

Another thing I have found useful is to slow the switching down considerably with gate resistors even up to 470 ohms or so.
This will increase the dissipation as the frequency goes up, but being very slow switching makes the inductive load/spikes no longer a problem.

There is also a shoot through problem with a device turning on very fast, causing high currents to charge the opposite device capacitance.
When it turns off fast, the interruption of current through the inductance makes big spikes.

Getting a nice compromise takes a bit of effort.

Opto couple the output from the CPU, at least during initial testing, then get the earthing and spikes right later.

Without a fast storage CRO it is not easy to get this right.

[rlwoodjr]

With the circuit you have, you need to pulse the high side drive on a regular basis to keep ensure the bootstrap capacitor has adequate charge so that it all stays digital and never goes linear.
Do you intend to have a current limit in the circuit?
What are you doing about shoot through protection? Inductive loads can do strange things, and you must ensure the mosfets are always protected.

Another thing I have found useful is to slow the switching down considerably with gate resistors even up to 470 ohms or so.
This will increase the dissipation as the frequency goes up, but being very slow switching makes the inductive load/spikes no longer a problem.

There is also a shoot through problem with a device turning on very fast, causing high currents to charge the opposite device capacitance.
When it turns off fast, the interruption of current through the inductance makes big spikes.

Getting a nice compromise takes a bit of effort.

Opto couple the output from the CPU, at least during initial testing, then get the earthing and spikes right later.

Without a fast storage CRO it is not easy to get this right.

I am using 15% to 85% duty cycle to keep the boot strap voltage. If I go much higher/lower than that the fan73832 shuts down (it has under voltage lock out). As soon as you pulse the boot strap it starts up again with seemingly no damage.

I have thought about current limiting, but one step at a time. I am a novice to h-bridges and micro-processors. I have looked at some circuits such as the uhu and elm and learned a lot from them, but I have a long way to go.

The fan73832 has shoot through protection built in. I am using a 100k so my delay for shoot through is around 1 microsecond. I have considered using the uhu version for extra protection, but I have not lost a mosfet (yet) with this arrangement. On a test bed I have run full 3500 rpm one direction then 3500 rpm the other direction, 3 seconds each way. I did this for 15 to 20 minutes with no problems.

I was hoping to make this simple and cost effective. I started with bjt's and opto-couplers form a web page I found, Learned about shoot through and advantages of mosfets. My first mosfet attempt was with p channel fets. Then high side drivers, the complexity and cost keep going up......bummer.

[rlwoodjr]

I never could get to work for more than an hour or two so I have abandon this project and moved to a simpler one. I will post it shortly.