2

I'm working on a project where low cost and small BOM is just as important as optimal functionality. The goal is to drive a cheap DC motor (aka 'yellow motor'), in two directions, optionally with PWM.

We drive both a microcontroller and this motor from a regulated 3.3V source, which is able to provide more than enough current. The MCU (obviously) controls the motor direction. The motor consumes ±60mA when free running, which I think is pretty close to what it will be doing in the actual product (light load).

If I'd take the 'safe' route, I'd use the TI DRV8837 H-bridge IC and get a really small BOM and tiny footprint and be done with it. The unit price on a 3K reel for this on Digi-Key is €0,33. This is about 7% of the total product cost (± €5,-).

So IMO there is a case to look into a basic H-bridge made from transistors. For a BJT solution, I'd need the following:

  • 2x NPN
  • 2x PNP
  • 4x back EMF Schottky diode
  • 4x base R

When I choose the cheapest parts on Digi-Key for 3K reels, I get SOT-23 3904/6 transistors, some SOD-123, MINIMELF or SOT-23 Schottky diodes and a bag full of 0402 1K resistors, which would add up to only €0,16. That's less than half the price. Using GP 4148 type diodes would save me another 2 cents.

Using 2 P/N MOSFETs pairs, I could drop the base resistors, altough when I search for the optimal solution, I would ideally need gate resistors AND pull-downs. Also, there many questions on this board regarding the need for back-EMF Schottkies in parallel with the MOSFETs, when there are body-diodes in the MOSFET, too. My question is, would the body-diode inside the MOSFETs be good enough for this particular situation? Let's assume not and look for the cheapest parts again:

  • 2x 2N7002 2 * €0,024
  • 2x BSS84 2 * €0,038
  • 4x Schottky 2 * €0,022

This adds up to €0,17 even with some resistors thrown in. So I guess using MOSFETs is just about as economically attractive as using BJTs. But again, dropping the back-EMF diodes would save a lot.

Say we would drop the (optional) requirement to PWM the drive signals, then the amount of switching would be extremely low. How much an impact would back-EMF have considering the circumstances?

-- update

Using a pair of dual Schottky common anode/cathode diodes in SOT-23-3, I can shave another 4 cents off, but adds another unique part to the BOM.

Genoil
  • 258
  • 2
  • 13
  • Is your question « what is the cheapest » or « what is the effect of back-emf »? Please edit your question appropriately. Also, is there a labour overhead for numbers of components? – Solar Mike Jan 10 '19 at 10:23
  • Thanks @SolarMike I tried to rephrase the topic. Also good point on the labour overhead. The DRV8837 option as 2 unique parts, whereas a transistor solution has 4 or 5. I was checking with the assembly partner and they charge about $38 per unique part at qty. 3000. This is about 1.3 cents per part per unit. – Genoil Jan 10 '19 at 11:06

1 Answers1

3

I would look at small/cheap NMOS/PMOS FETs, for example: AO3400 / AO3401

These can handle a very large Id (more than 4 A) which is overkill for your application but means that the current that the Body diodes can handle is also increased (compared to low current MOSFETs). The body diodes in the AO340x can handle 2 A so in my view they can take over the task of the back EMF diodes easily.

I would ideally need gate resistors AND pull-downs.

That depends entirely on how your H-bridge controls the MOSFETs. If you're controlling the MOSFETs directly from a microController then you only need a high value resistor between gate and source of the MOSFETs. That is needed to keep the MOSFETs off when the microController initializes and its outputs are high impedance / undefined.

Bimpelrekkie
  • 80,139
  • 2
  • 93
  • 183
  • Thanks. But these cost € 0,077 each, which runs pretty close to the DRV8837 option, esp. with more expensive assembly. – Genoil Jan 10 '19 at 11:07
  • Those AO340x are just an **example**. I have **no doubt** that there are cheaper but similar MOSFETs for sale. Just search for MOSFETs with Id_max > 1 A. There are **many** types having very different prices. I do not understand how assembly is more expensive as these MOSFETs come in very standard SMD packages. – Bimpelrekkie Jan 10 '19 at 11:10
  • Ok if I search digikey for 1A+ MOSFETs, I get some 4 cent parts, which is pretty good indeed. But when I look at the maximum body diode current of the 2N7002 that I selected earlier, It also has a maximum body diode If of 2A. What is the difference? – Genoil Jan 10 '19 at 11:25
  • Assembly is more expensive because they count the number of unique parts on your BOM. – Genoil Jan 10 '19 at 11:27
  • *It also has a maximum body diode If of 2A.* All I see in: https://www.onsemi.com/pub/Collateral/NDS7002A-D.PDF is a maximum **pulsed** current of 0.8 A. I am not saying that 0.8 A or 2 A is not enough. It probably is enough. **You** will have to do the calculations summing up all the costs. I just suggest **a possible** solution, then you use that to make a cheap as possible solution using that idea. Only then can you fairly compare it to the other solutions. So calculate the **total cost**, include assembly and stop saying "it has more components so assembly is more expensive". – Bimpelrekkie Jan 10 '19 at 11:42
  • I understand and I really value your feedback, sorry if I didn't really get that accross. If I use the cheap 2 cents mosfets, i may have to use 4 identical schottkies of 2 cents each and each unique components costs me 1 cent in assembly, it adds up to 4*2 + 4*2 + 3 = 19 cents. if i use your suggestion and use 4 cents 1A+ mosfets and no schottkies, it adds up to 4*4 + 2 = 18 cents. So in the end it will come down to finding a good deal and working out the exact details, because the example given is of course not fine-grained enough.. – Genoil Jan 10 '19 at 14:00
  • as shown in my other question (https://electronics.stackexchange.com/questions/417082/freewheel-diodes-and-shoot-through-prevention-required-on-this-low-voltage-h-bri) there's another case for using 1A MOSFETs. The yellow motor requires a peak current of almost 800mA to start spinning. I have now built a version of the H-bridge with the cheapest MOSFETs I could find (2N7002+BSS84), but it fails to get the motor spinning. – Genoil Jan 16 '19 at 13:22