OPA548EVM: Using OPA548EVM to drive Thorlabs VCFL35 voice coil flexure scanner

Hi Ilan,

Thanks for all of the details here regarding your setup!

Ilan Kleinman said:

1. Is the OPA548EVM suitable for driving this type of low-resistance inductive voice coil load in voltage-drive mode?

OPA548 is a perfect pick for the range you are targeting. 

Ilan Kleinman said:

2. For this application, would TI recommend using the standard non-inverting voltage amplifier configuration, or the Improved Howland Current Pump configuration?

Non-inverting is the simple approach. Howland can be especially nice for very reactive loads, though the additional components of the HCP can add some complexity to the design + additional power loss through the current sense resistor. I think both solutions would work for your intended use-case. 

Ilan Kleinman said:

3. If using voltage-drive mode, are there recommended stability components for the 2.2 ohm + 167 µH load, such as:

   - series output isolation resistor

   - Zobel network

   - feedback capacitor

   - changes to Riso, Rhcp, Cf1, or Cf2 on the EVM?

I tried to avoid the series isolation resistor on your design, though we could add this if you want. 

I configured the amplifier in a gain of 2V/V with a 1/2 attenuator on the frontend. I use a 1nF CF and an aggressive 1uF + 2Ohm snubber. I assume you have a 50Ohm source impedance for the Fgen, though this could be made a higher resistive divider if you want to operate Hi-Z output on the generator. 

You are welcome to play around with values and make changes as you see fit.

Simulating the closed loop response gives us a very flat frequency response over frequency:

Ilan Kleinman said:

4. What supply voltage would TI recommend in order to minimize amplifier dissipation while still allowing clean output swing up to 3.1 VRMS across the load?

   Would ±12 V be reasonable?

+-12V is fine, though +-10V may be preferred if this voltage is adjustable. 

Ilan Kleinman said:

5. What is the recommended method to set the OPA548EVM current limit for a maximum load current of 1.4 A RMS?

   The EVM user guide states that the default 14.7 kohm RCL gives approximately 2.5 A for OPA548. Should RCL be increased to around 26–33 kohm  for this load, or should the value be calculated differently from the OPA548 datasheet?

Correct, somewhere in this range. Note, we should apply some guard band from your maximum expected current to account for natural errors in the current limit circuitry. For example, if I were using a current source of 1A maximum, I might set my limit to 1.2A. 

Ilan Kleinman said:

6. Are there any thermal concerns with continuous sine operation near 3.1 VRMS into 2.2 ohm?

   Is the EVM heatsink sufficient, or should a larger heatsink/fan be used?

Here is another simulation, this time showing Pd in the IC. Remember, power delivered to the load can often times be less than the power dissipated inside the IC. 

Here PD(W) represents the power dissipated inside the IC. Here, RMS current is about 10W with a +12V supply

If we reduce to +-10V supply, we now have just 7.8W RMS. 

Now we can pass this information to our thermal design. 

Here is the PN for the heatsink on the EVM: Aavid Thermalloy 530002B02500G (2.6C/W)

If we have ~5C/W total Theta JA with the EVM heatsink, we could expect ~50C increase in junction temperature at +-12V supply with the expected load. Note, very low frequency signals should not always be treated from an RMS power perspective. It is good that we have enough thermal headroom here to not be too worried for entering thermal shutdown. It all goes to say that a fan is likely not required for this setup, however an elevated ambient temperature could change this. 

Ilan Kleinman said:

7. Are there known limitations of the OPA548EVM when driving voice coils or similar inductive electromechanical loads from a function generator?

Sometimes we have some amount of crossover distortion with these devices when the current is significantly out of phase from voltage. In this case, we should not see much of this since the load is tied to GND, and the snubber helps shift the phase current to be more in line with voltage. Otherwise, no real drawback with the Fgen. Non-inverting is nice in this respect as the Fgen is effectively separated from the high power output.

Ilan Kleinman said:

8. If OPA548EVM is not the best solution, would TI recommend another TI device or EVM for this application, such as OPA549 or another power op amp?

I think OPA548 is our best option here, and it has an EVM. Something like OPA549 or OPA541 could work, though we do not have a web available EVM. Is this for prototyping work, or is there a need to design custom hardware for your implementation?

Thanks,

Jacob