Hello. I need help about a strange problem I'm facing with the DRV8844.
In some applications is possible that the controlled DC motor gets mechanically turned, while power is not supplied to the control board. This happen, for instance, in electric actuators equipped with reversible gearbox. In such applications, the gearbox chain is normally moved by the motor, but is also present a mechanical bypass through which the user can manually operate the actuator. Obviously this is permitted only while the system is not powered.
For the control board, this means that the output connected to the motor can be "back-powersupplied" by the motor itself, which behaves as a dinamo: the induced Back-EMF will build a current flow through the substrate diodes of the H-bridge Mosfets (freewheeling diodes) thus actually powering the board.
To discern between a proper power supply or this "backward powered" situation, the designer must add circuitry to detect the power origin, and act accordingly keeping the driver totally disabled in presence of such "backward power". Furthermore, the designer must verify the maximum amount of power and voltage that will be "injected" in the system, and find a way to limit their values to levels that will not harm the driver and/or other board components. In some cases a bypass of the freewheeling diodes, usually by means of external Schottky diodes, can also be necessary as well as dummy loads across the supply rails to dissipate the power entering the system.
That said, I've just finished to develop a control board for one of this applications, based on the DRV8844, intended to renew a lineup of devices designed back in 2004 and based on the Freescale MC33887. Everything worked fine and really satisfactory, until we get to test the manual operation of the actuator. During this tests, we found out that manual actuation was almost impossible because of the motor being very "stiff", i.e. very hard braked.
After some investigations, I've observed a strange behavior of the DRV8844 when its outputs are energized. The same behavior has been seen also on the "DRV8844EVM Rev.A", so it can be easily reproduced on the Evaluation Kit.
To simulate a DC motor manually moved, I use a bench power supply, equipped with both voltage and current controls. Then I took the DRV8844EVM Evaluation Board, with JP2 closed and ALL the 10 JP3 jumpers OPEN. Since all inputs are equipped with pull-down, in this way the driver will be totally disabled. No connections have to be done to the EVboard.
Then, follow this steps:
- Set the power supply to 10V output (not critical) and limited to 0.1A.
- Connect two wires to the power supply, but also a 3rd wire to keep the output shorted (thus starting the test at 0V real).
- Connect the power supply to a pair of the "OUTx" testpoints of the EVboard, e.g. OUT1 & OUT2.
- Take off the short at the power supply.
At this point, the voltage should rise to 10V and the board itself should be supplied with that voltage (just 2*VF less). But instead, what can be seen is that the current is "swallowed" by the DRV8844 with the voltage limited to about 1.8V, while at the VM testpoints only 0.6V are seen. Even worst, if we slowly rise the current sourced by the supply, the voltage does not change while all the added power gets dissipated by the DRV8844 itself.
In this state it seems that a very high level of power dissipation can be reached, eventually damaging the silicon itself, while voltage at the VM remains at 0.6V, as if a directly-polarized diode were present with Anode to +VM and Cathode to -VM.
I've tried to bypass the substrate diodes (freewheeling diodes) with 4 Schottky diodes type BYV10-40, but no difference.
By the other hand, if I properly power up the driver before the test, this behavior doesn't shows up and the DRV8844 act exactly as expected (provided that all inputs are in the "OFF state", of course). After this test, to return to the state where the problem is seen, one must ensure that VM has drop under a certain threshold, about 1.5V/2V.
Since in my application I use the DRV8844 outputs paralleled 2-by-2, at first I thought the problem could be related to this aspect, but this doesn't seem the case, since the behavior can be seen, as described, on just a single pair of totem-poles.
This problem is endangering the development of our product, so I hope that someone will find out what the problem is and how it can eventually be solved.
Please note that nor the Freescale MC33887 neither the Allegro A3959 shows a similar behavior.
Thanks for your kind attention,
Riccardo.