This question is about a personal project and related to my previous question but focusing on different issues or questions. By increasing the shunt resistor from 1 mOhm to 330 mOhm I wanted to avoid using a special amplifier like in my previous question.

A uC's PWM output will control 2A to 6A current through a heating-wire and since the resistance might change, the current value roughly will be fed back to ADC of the uC for regulation ect. The circuit will be soldered on a perfboard.

For minimalism, I decided to use a single-supply op amp and low-side sensing to avoid common-mode-voltage related issues.

For clarity I will write down the questions one by one in detail:

1-) This article examines the use of op amp, difference-amplifier and instrumentation amplifier for this purpose. What I first understand is, if I use a difference-amplifier or an instrumentation amplifier I have to use a split-supply(or maybe some reference voltage techniques which will reduce the range I don't know actually). Since the only option left for me is an op amp option, the article mentions the following issue for that case:

The drawbacks to low-side sensing are disturbances to the system load’s ground potential and the inability to detect load shorts. Figure 2 depicts a typical low-side sensing scenario.

And below Rp or Vp illustrates this:

enter image description here

My question is, in my application would that be an issue as well? I mean I can live with 100mA error. I couldn't find a topology where I can use a single difference amplifier with a single-supply. Is that possible? Or how can I minimise the error introduced by ground?

2-) I used LM358 as an active filter with unity gain and it follows another LM358 as an amplifier with a gain of two(to match 0 to 5V range to the ADC).

Green is the swept PWM voltage across Rsense; blue is the output of the active filter with LM358 and the red is the final output. Is LM358 proper for this application? And is 0.33 Ohm shunt resistor is high enough?

3-) Basically the question is do I need a gate driver for this MOSFET(for power dissipation concerns)? According to the data sheet this is a logic MOSFET. But still some in my previous question told me I still need a driver for since the MOSFET has capacitance which will drive excessive current.

But one of the commentators told me that I need a special driver such as this one. But the rest didn't find it as an issue. So I'm confused how to drive this MOSFET properly. Here is what LTspice shows the MOSFET's power at %95 PWM: enter image description here

Edit: enter image description here enter image description here

  • \$\begingroup\$ Soliciting opinions is off topic. This is a q and a site. Try being specific and clear about your questions. \$\endgroup\$ – Andy aka Jun 10 '17 at 13:22
  • \$\begingroup\$ Omg I spent an hour to be clear, if you still think my question is unclear I think thats my capacity. Im sorry. \$\endgroup\$ – floppy380 Jun 10 '17 at 13:24
  • 1
    \$\begingroup\$ I jumped to the bottom and read your request so it's worth deleting the opinions and suggestion bit. I'll read the rest! \$\endgroup\$ – Andy aka Jun 10 '17 at 13:25
  • 1
    \$\begingroup\$ Alright I deleted that part maybe better format now. \$\endgroup\$ – floppy380 Jun 10 '17 at 13:26
  • 1
    \$\begingroup\$ Also don't forget I^2R, (6)^2 *0.33 = 10.8 Watts... \$\endgroup\$ – sstobbe Jun 10 '17 at 13:37

1) in my application would that be an issue as well?

Maybe yes, maybe no. If you want good accuracy you can use this approach, but you have to be very careful about routing your signals. If you make your power connection right next to the return end of your load, and also put the ground connection of your sense resistor very close to that point, you should be OK. 6 amps is (by hobby standards) a pretty hefty current, but it's not outrageous.

2a) Is LM358 proper for this application?

Yes, and it's a good choice as long as you're using cheap, readily available op amps. It's input configuration allows inputs right down to ground, and that is important in this circuit. There simply aren't a lot of this-generation op amps which will do the job.

2b) And is 0.33 Ohm shunt resistor is high enough?

Actually, .33 is probably a bit too large. As I mentioned in your previous question, at 1 mohm 6 amps produces 6 mV, which does not appreciably affect the MOSFET. At .33 ohms, full on current is about 2 volts, as you recognize. The thing is, assuming you're using a 3.3 volt MCU to provide your PWM, assuming 3 volts for the output to the gate of the FET means that the gate-source voltage is only 1 volt when on. This is the nominal turn-on voltage for your FET (Vgs(th), or gate-source turn-on voltage threshold - look it up in your data sheet), and the threshold is usually set at a very low current, such as 250 uA or 1 mA. So you need more voltage to drive the FET to handle 6 amps. If you drop the sense resistor you'll have less sense voltage which will give you more gate voltage margin, but you'll also need more gain in your conditioning circuitry, and the circuit will be more sensitive to stray resistances in the ground path.

This leads to

3) do I need a gate driver for this MOSFET(for power dissipation concerns)?

Well, if you give the gate enough voltage, I'd guess that you'll be OK in terms of dissipation. But a driver is a very good idea, especially if you keep the sense resistor at .33 ohms. Like, a very, very good idea. If you're using a 5 volt MCU, I'd guess you don't need a driver.

  • \$\begingroup\$ +1 for the effect that too high a source resistor will start to turn off the mosfet on high currents. \$\endgroup\$ – Andy aka Jun 10 '17 at 14:11
  • \$\begingroup\$ Thanks I updated the circuit. Please see my edit. PWM freq. will be 500Hz and 0/5V pulses not 3.3V. But for a better Vgs I listen your suggestion and reduced to Rsense to 50mOhm which at 6A dissipates 2Watt at 100%PWM. So it make sense to reduce Resense. I wanted to replace LM358 with an opamp with much better input offset voltage but the best result for the active filter for some reason is with LM358 so I will stick with it. I also increased R_nichrome to 1.9Ohm to set the current amplitude at 6A. So as far as I understand at this freq (500Hz)and this Vgs setting I dont need any driver right? \$\endgroup\$ – floppy380 Jun 10 '17 at 16:24
  • \$\begingroup\$ I think we are missing a point here about freq. Freq. of the uC can be low like 500Hz but when I adjust the duty cycle very low like %1 or very high I think that will be a problem. Triggering MOSFET at 500Hz with %50 duty is not the same thing triggering 500Hz with %1 duty. But maybe still harmless 25mA. Just wanted to point out \$\endgroup\$ – floppy380 Jun 10 '17 at 20:09
  • \$\begingroup\$ @doncarlos - I don't understand your concern, but reread my original response. The way you drive the MOSFET gate will cause inaccuracy at low duty cycle. The simple solution is a MOSFET driver IC. \$\endgroup\$ – WhatRoughBeast Jun 10 '17 at 20:24
  • \$\begingroup\$ @WhatRoughBeast I really want to keep it simple so if not needed I dont want to use an extra IC. I meant my application will not use PMW at duties less than %30 so I dont think it would be problem. Btw I want to replace LM358 opamps with LMC6482. LMC6482 has less input offset voltage and it is rail to rail. \$\endgroup\$ – floppy380 Jun 10 '17 at 22:12

1) try and keep the opamp that senses the resistor close to that resistor. Tee off all circuit 0 volts from the bottom of that resistor i.e. use star pointing techniques where you can.

2) you can use a lower value resistor with more gain in the second opamp stage of course. Going higher than 3.3 ohms might mean it starts to burn or get too hot. Don't go too low or the input offset voltage errors from the LM358 will start to be significant.

3) if you are only running at a few kHz max then it's unlikely you'll need a specific driver for the mosfet but do check the data sheet and look at the graph for drain current versus DS voltage for various gate levels. This will tell you if your gate drive level is high enough. Specific gate drivers tend to be required in the hundreds of kHz in order to charge and discharge the gate capacitance fast enough.

  • \$\begingroup\$ I modified the circuit please see my edit. I lowered the Rsense to 50mOhm(Is that an okay value for offset error issue?) and increased the gain of the second stage. For the active filter any other op amp than LM358 in simulation gave worse result. Even the ones with very low input offset voltage. But regarding your 1st answer; do you mean that the ground of the Rsense should be very close to the GND of the 12V power supply? If so, in practice how long max? Maybe I dig it too much but never tried such circuit before and cannot find anymore practical information quantitatively. \$\endgroup\$ – floppy380 Jun 10 '17 at 16:28
  • \$\begingroup\$ The 12 volt power supply can be feet away but once you decide where 0 volts is, stick to it and make that your star point. The logical place is at the low voltage end of the sense resistor. \$\endgroup\$ – Andy aka Jun 10 '17 at 16:32
  • \$\begingroup\$ What do you mean by starpoint. Could you draw a simple illustration about what you say if possible it would help a lot, Im very curious if I get it right? \$\endgroup\$ – floppy380 Jun 10 '17 at 16:34
  • \$\begingroup\$ You'll find pictures if you google it. Basically it means that any connection to 0 volts connects to the single unambiguous star point independently (as far as practical of course). Clearly those 0volt connections that carry micro amps can share a limb with other low current connections. The whole point is that you define where your 0 volt reference is and sensitive to connections do not share a wire or path with the 6 amps or pwm signals. \$\endgroup\$ – Andy aka Jun 10 '17 at 16:39
  • 1
    \$\begingroup\$ See C4 and C5 forming an island with he associated driver. That's an example of where you can take a short cut. See how the driver and fet have labels saying keep short. That could form another island of course but, until you get it intuitively you might as well just follow the basic rule and try short cuts out later. \$\endgroup\$ – Andy aka Jun 10 '17 at 19:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.