Check my *cough* logic - H-bridge design.

[samco]

This is a combination of a few circuits floating around the zone.. I am in need of a simple current limited h-bridge that will take PWM UP/DOWN from emc. I have built a non-current limited one that has been working great for testing.. http://www.electronicsam.com/images/...art/almost.JPG and http://www.electronicsam.com/images/...rt/ampmess.JPG
using IR2111 gate drivers..

Mainly looking for feedback on the current tripping. The reason I am using the logic I am - is I have it in hand. (in case there is a - you could do it better using....)

annother note.. parts of this was taken from the uhu iirc.. it was outputing pwm - notpwm. With emc - there is never any signal on one pin while the other is outputing pwm.. (up/down) so I needed to OR the 2 pwm signals to the flip flop so I was sure it was reset. (hope I am making sense)

Thanks for looking.. (sorry about the mess)
sam

[samco]

Oh - and it works out that if I put 2 drives on one board - all the logic gets used up perfectly...

[epineh]

I will be watching to see how your current limit goes, I am the same, I have working H Bridges running nicely on my router with no current limit at all I know that is bad but I don't have the time/knowledge to build my own circuit. I will be working on it but for the meantime that is why I went with the Elm Chan servo drive, it does a torque limit that works quite well as a current limit.

Once I get a real current limit done then its off with the Elm Chan drive and onto a closed loop to PC setup with EMC, probably with a pluto-p or Mesa Hostmot card.

Russell.

[neilw20]

Hi Sam. Russell, I think you will find this of interest also.

Some changes and the reason.

Increase R1,R2,R3 and R4 to something like 220R or 330R, or even 470R.
Reason: To reduce or stop 'shoot through' currents in bridge.
Consider the condition where all of Q1,2,3 and 4 are all off and inductive flywheel current is flowing through the load connected to PAD6 and PAD7 and through diodes U$2 and U$3 (and also through the diodes in the MOSFETS).
If transistor Q1 or Q4 are commanded to go on which is (normal part of the logic, which functionality I choose to ignore) Q1 has to reverse the current through U$3. U$3 has a reverse recovery time of just say 100ns (I dunno what it actually is) now Q1 sees a short to ground through the current sense resistor for the reverse recovery time of U$3. The load is irrelevant at this point.
What you will see for the reverse recovery time is the FULL SUPPLY across the current sense resistor R9 which could be upward of 1000Amps given that the ESR of your filter might be quite low, and there was very little inductance in your wiring. I practice it would be more like 100-500A !

At the end of the pulse now you will get a big negative spike across R9 as it in practice does have some inductance, (and you might have used a wirewound one !) This negative spike will cause U$3 to conduct again and Q1 will see a short circuit again. You have just created a high power oscilatory circuit !

To overcome this we allow the MOSFETs to turn on slowly so that the reverse recovery current is limited by themelves dissipating the energy.
We lose a small amount of efficiency, which in most cases is negligible.

2 rules:
1). DON'T TURN ON MOSFETS TOO FAST as this creates big spikes (and RFI). Make R1/Gate capacitance time 20-50% greater than the reverse recovery time of diode (not forgetting the one in the MOSFET. It might be worse).
2). Turning them off too quickly encourages the inductive load to make BIGGER SPIKES too. The resistor helps here too.

As supply voltages get higher and currents through MOSFETs are lower expect to use a lower value of resistance. The data sheets usually refer to situations where there is a supply voltage near the ratings of the device.
When used at low voltages the higher currents create higher spikes.

It is quite often very problematic measuring or seeing these spikes, but they will be there, even if fleetingly.

Hence the resistors. You can increase them until MOSFETS get a bit warm before you really need to worry, but typically 200-500 ohms is OK.

Not trying to change any of your circuit, but have a look at UC3843 PWM driver IC. It can disappear most of your logic.
It has high output drive, current sensing, a reference voltage, amplifier and draws very little current. It offer an elegant solution to many problems at a low cost. It has +/- high current output so can drive MOSFETS directly. Makes a great negative switcher supply with just a few parts too.

[samco]

epineh: stay tuned..

neilw20: Thanks for the info.
Looking at the ir2184 datasheet - it does all the shoot thru protection and also has on/off delays.

Ton/Toff is 680/270 ns
dead time of 500ns

it also has some other cool protection. (like I say - I have been using the ir2111 in a similar circuit and it has been working great).

• Floating channel designed for bootstrap operation
Fully operational to +600V
Tolerant to negative transient voltage
dV/dt immune
• Gate drive supply range from 10 to 20V
• Undervoltage lockout for both channels
• 3.3V and 5V input logic compatible
• Matched propagation delay for both channels
• Logic and power ground +/- 5V offset.
• Lower di/dt gate driver for better noise immunity
• Output source/sink current capability 1.4A/1.8A

http://www.datasheetcatalog.org/data...irf/ir2184.pdf

The current sense resistor I am using claims to have no inductance.. It is .015R at 7w

thanks again.. sam

[neilw20]

Yes it protects against shoot through protection by turning off the MOSFET if it does not saturate, by measuring the voltage across the MOSFET, then skips a drive cycle. Do you really want these almost invisible hiccups, because of the reverse recovery time of some other diode which the chip does not know about?
Sure it protects it, but IMHO if you want it to work nice you still need to turn on the MOSFET a whisker slower than the RR time.
Just delaying the turn on wont stop the current flowing through the diode.
Something has to change the direction of the current. Just because you delay for 680/270nS that will not stop flywheel current from flowing. It will continue to flow while there is inductive energy in the coil which quite often will be longer than 680nS.

Neil

Russell. Any comments?

[epineh]

Originally Posted by neilw20 Russell. Any comments?

Not really, most of this stuff goes over my head, I am currently running much the same configuration as Sam's schematic using 2184's, I have used 270ohm gate resistors with a fast recovery diode in parallel, (reverse) and it seems to be working OK so far, FETS run only slightly warmer than ambient under rapid accel/decel conditions.

My bridges don't have current limit (yet) but the torque limit with the Elm Chan is doing the job surprisingly well, I had about 12 crashes while jogging around last night and the drive faults out nicely without catching fire, always a bonus

Sam back to your logic, I assume that you are disabling the gate drivers upon overcurrent which in turn produces a following error from the drive or in your case EMC ? I was thinking of doing the same or possibly having the fault signal latching to make sure the bridge doesn't get turned back on if the overcurrent detection resets before the following error picks up the fault... or is that completely unnecessary ?

Another thought I had and Neil you may have ideas about this, is why not use a small micro, say an AVRtiny2313 (I have dozens of them in a drawer ) reading either the sense resistor voltage directly and setting a trip point or reading an amplified linear analogue voltage (I have another circuit in mind, I will swap PC's and add link) say from 0-5v for simplicity ?

Other thoughts I had for doing this is that the micro could be used for a couple of handy things, like handling all logic, enable, fault and if the drive/axis was to be disabled without bringing up an error state to the controller eg. a 4th axis that gets added and removed frequently. Also a little 6 or so LED bar graph showing servo current would be nice.

The current limit could be set using rs232 maybe, might be a little more simple to setup than trimpots, dunno just thinking out aloud. Would a micro running at 20MHz have any problem detecting overcurrent in a 20KHz servo ?

I know cost is usually the drawback for this kind of thing but I can get a tiny 2313 for $1.70 on my doorstep, so it is hardly a high price for some nice features...

Russell.

Here, found link, thought this might be a good way to send 0-5V to micro...

http://www.cnczone.com/forums/showpo...0&postcount=95

[samco]

This is just cycle by cycle current limiting.. I am doing it to help protect the h-bridge and servo. Yes - if something happens and the amp goes into current limit - emc will throw a following error. (if it can't get to the commanded positon because of it.)

If everything is tuned right in emc - I really don't think the amp should go into current limit.

sam

[samco]

anyone have info on a good layout for Kelvin configuration. (sense resistor)

[samco]

Here is an fyi - the tl082 doesn't like the -supply grounded. (comparing Millivolts)

(glad I breadboarded it)

So - I need to find a comparator that works at near 0v or use an extra supply

sam

[neilw20]

The LM339 / LM393 will work because it IS a comparator.
You will need a 6K8 or 10K pullup on the outputs because they are open collector.
The TL082 can cause a few weird effects during powerup when the internal biasing is too low. During power up the inverting input will work like a non-inverting input which can cause strange side effects.
Using a traditional LM358 is never a drama, and it's input common mode voltage includes ground which makes it easy to use with single sided supplies.
IN this case you need the speed of the LM339 family.
You can run it close to it's maximum supply voltage so that it still responds to spikes correctly. Use at least 1K or higher for R10 and use a fast diodes (1n4148 ok) to protect the input from exceeding the supplies.
Beware that a generic 4013 requires clock rise/fall times < 15uS.
If you cant achieve this easily then the Phillips HEF4013 have schmitt trigger clock inputs. Never get caught replacing a HEF4013 with a generic one. HEF's are often used for this characteristic and circuits can start doing erratic things.

[samco]

Thank you - that is what I needed to know. (getting there slowly)


thanks again
sam