1

I will be driving ~1600 linear resonant actuators (LRA) in parallel (each requiring 0.6V @ 19 mA AC). As the buck regulator, I will be using TPS546C23 synchronous buck converter from TI to convert 5V 4A DC supply from battery bank to 0.6V 20A DC. To control so many LRA using only 4 pins of an MK20DX256 ARM Cortex 4 MCU, I will be using several daisy-chained TLC5940 LED Driver to control gates of 2N7000 MOSFETs (each LRA will have its own MOSFET) via PWM signals. I tested the setup with a smaller number of ERM motors (which are similar to LRA but works with DC voltage) and it works just fine.

The problem is converting the 0.6V 20A DC to AC for the LRA. I did not find any suitable IC or circuit diagram to do DC/AC conversion of such low voltage and high current. I tried to look for individual chips that could possibly do DC/AC conversion of 0.6V 19mA DC to AC but failed to find anything. Most inverter circuits use 555 timers and I don't think those circuits can work with such low voltages. Also, LRAs are normally driven with haptic drivers but it severely limits how many LRAs I can connect to the MCU, unlike the setup I designed.

Any suggestion would be appreciated.

Swapnil Saha
  • 157
  • 1
  • 7
  • My first thoughts would be to look at a small audio amplifier such as the LM386. They're robust and cheap. Can you generate a master "tone" to feed the amplifiers and somehow apply a mute to turn them off as required? – Transistor Jul 14 '20 at 08:18
  • I suspect your problem is down-converting your 5V to 0.6V DC, and then trying to convert that to AC. – Simon B Jul 14 '20 at 08:21
  • ^ @SimonB partially yes. It's very easy to down-convert 5V to 0.6V DC at very high current using synchronous buck regulators or switching regulators (lots of ICs are there). The challenge is to convert that small voltage to AC. – Swapnil Saha Jul 14 '20 at 08:48
  • @SwapnilSaha the honest truth is that at such a high current, having many small inverters is easier than having one large inverter. Also, if you're going the single inverter route, the intuitive way would be to invert *down* from your 5V DC supply to AC directly, not first convert that down to DC (your DC low voltage value is wrong anyway, see: effective voltage of AC) – Marcus Müller Jul 14 '20 at 08:50
  • by the way, what are the requirements to your AC? Is a square wave OK, or do you need a sine? if the latter, how much total harmonic distortion is OK? – Marcus Müller Jul 14 '20 at 08:54
  • @Marcus Muller sorry yes it should be around 0.12V DC instead of AC. Square waves are fine to drive LRAs. The only requirement is the resonant frequency of 240 Hz, which shouldn't be too difficult to achieve. – Swapnil Saha Jul 14 '20 at 09:08
  • I am open to going through multiple inverter route. Let me know if you have any suggestions for it. – Swapnil Saha Jul 14 '20 at 09:10
  • 1
    err, no, if you want an AC effective voltage of 0.6 V, you shouldn't be using 1/5 of that as DC voltage. Again, I think your whole approach of going low on the DC side is a bad idea. – Marcus Müller Jul 14 '20 at 09:10
  • @SwapnilSaha Draw a block diagram of what you propose please. – Andy aka Jul 14 '20 at 09:11
  • Hence, my *suggestion* is: let go of the 5V -> 0.6V DC/DC idea and design an inverter based on a H-Bridge (or many H-Bridges) feeding a transformer (many transformer). – Marcus Müller Jul 14 '20 at 09:11
  • @Marcus Muller all right. Sounds good. – Swapnil Saha Jul 14 '20 at 09:12
  • Maybe you can salvage an induction oven circuit for your needs. Just check the different shopping sites for induction heating module. – Christian B. Jul 14 '20 at 10:17

2 Answers2

4

Let go of the 5V -> 0.6V DC/DC idea and design an inverter based on a H-Bridge (or many H-Bridges) feeding a transformer (many transformers).

You could then let the inversion step also involve the voltage conversion step (which is a good idea, because DC/DC voltage conversion is harder than AC/AC, which can be done trivially with a transformer).

Also, 5V 4A is a lot of power for such a low supply voltage - I'd recommend you don't go that low. A thrown out laptop power supply typically delivers around 20 V. Working down from a higher voltage is easier, here, because you get more "headroom" for regulation.

Then: I have no idea what a "linear resonant actuator" is, but it darn much sounds like what a speaker does.

So, idea: that "H-Bridge feeding into a transformer design" is something you shouldn't be designing yourself. That's just a class-D amplifier in the end, using the properties of a transformer instead of output inductors to do the lowpass filtering / smoothing of the output voltage. So, class D-amplifier, use a higher input voltage than 5V, use an appropriate transformer to convert down to your desired AC voltage. Feed whatever sine waveform you can generate into the amplifier, calibrate output amplitude once.

If you want the AC amplitude at your actuators to be very exact, you can pick an amplifier with external feedback. I doubt you'll need it.

In fact, if your class-D amplifier is beefy enough, you don't need the transformer at all – and can directly work with sufficiently sized inductors as you'll find in the application circuits in your amplifier's datasheet. I find the transformer alternative nice, because it shifts the cost from a lot of low-resistance semiconductor switches to a simple relatively low-power transformer.

Marcus Müller
  • 88,280
  • 5
  • 131
  • 237
  • 1
    Sounds like a classic case of XY problem :-) – winny Jul 14 '20 at 11:21
  • Sorry for the late reply. What I finally did was instead use ultra-low power and tiny PWM haptic driver (DRV8601) from TI and supplied proper PWM frequency to it via TLC5940 ouput @235Hz (resonance frequency of LRA). The driver chip consumes only 1-1.5mW of power at most. With the setup I can actually run the LRA at half the rated power without losing vibrational force and I can run as many LRAs as I want. – Swapnil Saha Aug 09 '20 at 00:39
0

~1600 LRAs does sound like a lot of fun. However, I believe how you control them depends a lot on what exactly you want to acchieve.

I'm by no means an expert, but what I've read is that the resonant frequency of LRAs is very narrow (high Q). Meaning, that if you want them to be efficient, you have to excite them exactly at resonance frequency. This frequency however may be different for each of your LRAs. This is the reason for chips like the DRV2603. On the other hand, if you mount all of them on the same carrier (or very close to each other on a mechanical system) they will likely be behave as coupled oscillators. Electrically this would look something like this:

schematic

simulate this circuit – Schematic created using CircuitLab

This means, if the coupling is strong enough (think: all mounted in the same direction on a free floating board) your whole system will have a single resonance frequency and will vibrate at just that frequency. You could have observed something similar with your ERMs.

If you're fine with the system behaving like that, you can likely switch the whole array using a single (or maybe several MOSFET H-bridges). The frequency is quite low (easily controlable from a Microcontroller) and switching 20A with a MOSFET is not hard. I don't think you will break your LRAs if you use the 5V directly, as long as you keep your duty cycle low enough (RMS is what matters here). If you do implement a H-bridge, keep in mind to switch one MOSFET off before you turn on the other.

michi7x7
  • 325
  • 2
  • 9
  • I ultimately went for the DRV route. However, instead of 2603, I used 8601. 8601 is much less flexible than 2603 (lacks I2C, SPI etc.) and instead takes in PWM signals at the resonant frequency of LRA, which is easy to generate with TI TLC5940. – Swapnil Saha Aug 09 '20 at 00:42