Selecting MOSFET

[permal]

Hi,

During the weekend I got some great help re. transistors and MOSFETS. I hope I've now learned enough to dare select a MOSFET and use that in my design.

My requirements are as follows

* Controllable at 0-3.3V (via MCP2307 I2C IO Expander)
* 500mA at V_DS 5V continuous load, so 2.5W
* V_DS up to max 12V i.e. ~210mA.
* Preferably no heat-sink needed.
* SMD
* Low-side switching.

I'm currently looking at the IRLL110TRPBF, N-Channel for low-side switching.

Alternatively the SQ1431EH for high-side switching (would of course need extra control circuitry as shown in this post by @Spec.

V_gs(th) is 1V, so the 3.3V ought to be sufficient.

If I'm reading the SOA graph correctly, I'm well within the safe area for both.

Could I please get a second opinion on the choice of these MOSFET based on the above requirements?

[aheid]

My requirements are as follows

* Controllable at 0-3.3V (via MCP2307 I2C IO Expander)
* 500mA at V_DS 5V continuous load, so 2.5W

V_gs(th) is 1V, so the 3.3V ought to be sufficient.

If you look at figure 1, you'll see that the IRLL110TRPBF barely be able to do 500mA at 3.3V with that transistor. Looking at the V_GS(th), it lists 1V min, and 2V max...

Compare this to say the AO3402, you can see that V_GS(th) is 1.5V max, and from figure it'll be more or less fully on at 3.3V.

However your specification of "500mA at V_DS 5V continuous load" is a bit ambiguous to me.

Do you mean you want to turn on/off a load with an effective resistance of about 5V / 500mA = 10 Ohm, or that you want the transistor to have an effective resistance of 10 Ohm, so that it has 5V over it when 500mA is running through it?

If the former, the IRLL110TRPBF won't really cut it. If the latter, then the SOA for the IRLL110TRPBF doesn't mention DC at all. While the SOA for the AO3402 does have a DC line, I wouldn't particularly trust it, given it's a trench construction.

[Audioguru]

The request is very confusing.
First the Mosfet has a current of 500mA and a Vds of 5V is to be used as a 2.5W heater. Then the Mosfet is a low side or high side switch when it barely gets warm.
The tiny Mosfet shown will burn up if it must heat with 2.5W.


The threshold voltage range of a Mosfet is the Vgs when it begins to conduct a very small current.

[permal]

Yes, I can see now that the question is a bit confusing (very much, actually). Sorry about that, totally mixed things up. It is not V_ds at 5 & /12V, but V_cc.

So:

5V and 12 rails.
500 and 210mA continuous load at the respective voltages.
3.3 V_gs

As @aheid points out, IRLL110TRPBF isn't suitable for those numbers.

[spec]

Hi,

During the weekend I got some great help re. transistors and MOSFETS. I hope I've now learned enough to dare select a MOSFET and use that in my design.

My requirements are as follows

* Controllable at 0-3.3V (via MCP2307 I2C IO Expander)
* 500mA at V_DS 5V continuous load, so 2.5W
* V_DS up to max 12V i.e. ~210mA.
* Preferably no heat-sink needed.
* SMD
* Low-side switching.

I'm currently looking at the IRLL110TRPBF, N-Channel for low-side switching.

Alternatively the SQ1431EH for high-side switching (would of course need extra control circuitry as shown in this post by @Spec.

V_gs(th) is 1V, so the 3.3V ought to be sufficient.

If I'm reading the SOA graph correctly, I'm well within the safe area for both.

Could I please get a second opinion on the choice of these MOSFET based on the above requirements?

Hi permal

Attached is a list of NMOSFETs that I put together for a similar application. The $ sign indicates merit.

When switching current with a MOSFET it is important to make sure that the RDS (at the VGS that you are supplying) is as low as possible to minimize the voltage drop and also to limit the power dissipation in the MOSFET. You should be able to achieve around 50mR with a 2V5, or greater, G/S voltage.

The threshold voltage, at its extremes needs to be low enough so that the 3V3 logic turns the MOSFET on and high enough so that the 3V3 logic can turn the MOSFET off. A good compromise is to choose a MOSFET with a maximum threshold voltage in the region of 1V to 1V5.

Provided that you are turning the MOSFET on and off relatively infrequently, and/or are turning the MOSFET on and off relatively fast, you do not have to worry too much about the SOA.

(your link above does not work)

[permal]

Hi permal

Attached is a list of NMOSFETs that I put together for a similar application. The $ signs indicate merit.

Oh my gosh, you're an angel 

How did you find these components? Is there an easier way than reading the data sheets for each one? The search tool at, e.g. Mouser only goes so far in what parameters you can search with. ('m probably missing something obvious)

[spec]

Hi permal

Attached is a list of NMOSFETs that I put together for a similar application. The $ signs indicate merit.

Oh my gosh, you're an angel 

How did you find these components? Is there an easier way than reading the data sheets for each one? The search tool at, e.g. Mouser only goes so far in what parameters you can search with. ('m probably missing something obvious)

There are two main ways:

Go to the websites of the manufacturers of the devices that you are interested in, and enter the device type (MOSFET) in the high level search box. Then further home in on the device that you want. Finally you will probably be presented with a table with a list of likely components and many columns of parameters: VDS, ID etc.

You can refine that list by editing the values for the various columns in the table. For example in your case only N type MOSFETs.

Having narrowed the list of components down, you then click on the link for the data sheet for the device of interest. Then go through the datasheet and check all the parameters that interest you.

For a search of many manufacturer's suitable parts, do the same thing with the electronic component distributor sites: DigiKey, Mouser, Element14, etc. In my opinion DigiKey has the best search facilities and Mouser is pretty good too.

In my case, I store a huge number of datasheets for all components of interest and any application reports etc: Resistors, Capacitors, Inductors, Transistors, and so on.  The PDF of NMOSFETs was extracted from the directory listing for, NMOSFETS_0-10A.

One of the advantages of searching a component distributors data base is that you can get the price of a component and you will know that the component is available.

Here is a snapshot of a DigiKey table for MOSFETS: https://www.digikey.co.uk/products/en/discrete-semiconductor-products/transistors-fets-mosfets-single/278?k=MOSFET

Attached below is a snapshot of my directory for PMOSFETs up to 10A, which you may be interested in:

[permal]

...

Right. So what I lack is simply experience and having done this for X years to accumulate such lists.

[spec]

...

Right. So what I lack is simply experience and having done this for X years to accumulate such lists.

Yes, that is exactly it. The whole thing seems pretty daunting at first, but it is like all fields, you soon get the hang of it. I can remember when the first opamps came out. I looked at the data sheets and thought that I would never be able to understand all that, but now I look at the same data sheets and think how basic and simple they are. And when the 7400 logic chip came out, the data sheet was a mystery.

But, I think I mentioned this before, variety reduction (VR) is the big secret. There are literally millions of component types available, so what you need to do is just focus on your area of electronics, and at the moment I would guess that you would be looking for NMOFETS and PMOSFETS to suit your 3V3 logic system. So the best thing to do is just chose one NMOSFET and one PMOSFET that will do all the tasks you will require in your 3V3 system. So for example, instead of choosing MOSFETs that will handle 500mA, go for a 6A capability. There are two suitable N & P MOSFETs from the lists I posted.

Then get to know those preferred devices well. After that, when you do a new design, you do not need to search for suitable devices. Instead you design your new circuit around the devices you know. Also, this approach radically reduces the number of components you need to stock.

[permal]

...

Right. So what I lack is simply experience and having done this for X years to accumulate such lists.

Yes, that is exactly it. The whole thing seems pretty daunting at first, but it is like all fields, you soon get the hang of it. I can remember when the first opamps came out. I looked at the data sheets and thought that I would never be able to understand all that, but now I look at the same data sheets and think how basic and simple they are. And when the 7400 logic chip came out, the data sheet seemed so complicated that I gave up.

But, I think I mentioned this before, variety reduction (VR) is the big secret. There are literally millions of component types available, so what you need to do is just focus on your area of electronics, and at the moment I would guess that you would be looking for NMOFETS and PMOSFETS to suit your 3V3 logic system. So the best thing to do is just chose one NMOSFET and one PMOSFET that will do all the tasks you will require in your 3V3 system. So for example, instead of choosing MOSFETs that will handle 500mA, go for a 6A capability. There are two suitable N & P MOSFETs from the lists I posted.

Then get to know those preferred devices well. After that, when you do a new design, you do not need to search for suitable devices. Instead you design your new circuit around the devices you know. Also, this approach radically reduces the number of components you need to stock.

Excellent advice that! Thank you.

[spec]

No sweat

[Zero999]

Another thing to consider is the switching speed. The gate of a MOSFET looks like a capacitor who's value increases when the MOSFET starts to turn on. This is due to the Miller effect. A certain charge or number of electrons is required to fully turn on a MOSFET at a given starting drain-source voltage, a parameter known as the gate charge and is measured in nano Coulombs nC. The lower the gate charge, the better, as it's less charge is required from the driver circuit. This is not much of a problem, if the MOSFET doesn't need to be switched quickly, but at high speeds it becomes signifcant.

[exe]

Attached is a list of NMOSFETs that I put together for a similar application. The $ sign indicates merit.

Love to see AO fets which I have as well because of their good performance / price ratio (but I haven't put into application yet). Question, was Rds@2.5V actually measured or inferred from the datasheet?

[spec]

Attached is a list of NMOSFETs that I put together for a similar application. The $ sign indicates merit.

Love to see AO fets which I have as well because of their good performance / price ratio (but I haven't put into application yet). Question, was Rds@2.5V actually measured or inferred from the datasheet?

Alpha and Omega (AO) Yes, they make great MOSFETs. At one time they seemed to be in the lead with performance but other manufacturers have now caught up. As you say they are still a good price though.

The figures that you see in the table are all worst-case and taken from the data sheet where possible. Where the figures are not given directly on the data sheet they are extrapolated from the closest figures. To check what is what, always read the datasheet- that is the bible.

As far as RDS goes, yes they are all worst-case with a VGS of 2V5. Don't forget though that the RDS is for a junction temperature of 25 degC and RDS increases with temperature, in most cases. So, to account for possible temperature increases, I tend to double the RDS to get a working value.

[permal]

Attached below is a snapshot of my directory for PMOSFETs up to 10A, which you may be interested in:

You should put them up on Github or some such, I'm sure other people like me also struggle with this.

[permal]

So based on @Spec's list, I'm looking at Si2374DS for an N-channel version. Searching for similar chips, but P-channel, I found SI2323CDS which looks nearly identical to my untrained mind, except the mode.

* Any objections to these?
* Both datasheets specify "Continuous Drain Current" in their Abs Max section, with the caveat that t=5S. How is that "continuous"?
* As I'm going to need eight of each and they'll all be placed on a row (at the I/O connectors), I'm just wondering if another model in dual/quad package is a better option?

[spec]

So based on @Spec's list, I'm looking at Si2374DS for an N-channel version. Searching for similar chips, but P-channel, I found SI2323CDS which looks nearly identical to my untrained mind, except the mode.

* Any objections to these?
* Both datasheets specify "Continuous Drain Current" in their Abs Max section, with the caveat that t=5S. How is that "continuous"?
* As I'm going to need eight of each and they'll all be placed on a row (at the I/O connectors), I'm just wondering if another model in dual/quad package is a better option?

Your choice is fine, but I would recommend the Si2312CD (NMOSFET), and Si2323CDS (PMOSFET) (as you chose) because they are the latest MOSFETs from Vishay.

You can get dual MOSFETs in one case, but I would advise sticking to singles. They are in an SOT23 case which is pretty compact, but bigger than most cases these days, so easier to handle.

[spec]

Attached below is a snapshot of my directory for PMOSFETs up to 10A, which you may be interested in:

You should put them up on Github or some such, I'm sure other people like me also struggle with this.

Hmm that's an idea. The thing is that the list changes frequently as new devices become available. Also I have multiple lists for most components so where would it end.

(The NMOSFET table took 12 hours to do, so it is not a quick task.)

[permal]

Your choice is fine, but I would recommend the Si2312CD (NMOSFET), and Si2323CDS (PMOSFET) (as you chose) because they are the latest MOSFETs from Vishay.

You can get dual MOSFETs in one case, but I would advise sticking to singles. They are in an SOT23 case which is pretty compact, but bigger than most cases these days, so easier to handle.

Ok, I'll go with your recommendation Finally I can make some progress on this project

Hmm that's an idea. The thing is that the list changes frequently as new devices become available. Also I have multiple lists for most components so where would it end.

(The NMOSFET table took 12 hours to do, so it is not a quick task.)

I have no idea how you do it, but I keep all my documents, code and such in Git. It not only gives you revision history, it also acts as an extra backup. I've found this to be helpful even for schematics although they are not as easy to compare as source code, but it is doable. I've lost count of the times where I've reverted back to a previous version to start over. Perhaps commercial tools have this built in, KiCad does not.

I'm not surprised it takes a lot  time to compile the list, god knows I've spent hours just trying to select these and I had good help from you.

[spec]

Your choice is fine, but I would recommend the Si2312CD (NMOSFET), and Si2323CDS (PMOSFET) (as you chose) because they are the latest MOSFETs from Vishay.

You can get dual MOSFETs in one case, but I would advise sticking to singles. They are in an SOT23 case which is pretty compact, but bigger than most cases these days, so easier to handle.

Ok, I'll go with your recommendation Finally I can make some progress on this project

Hmm that's an idea. The thing is that the list changes frequently as new devices become available. Also I have multiple lists for most components so where would it end.

(The NMOSFET table took 12 hours to do, so it is not a quick task.)

I have no idea how you do it, but I keep all my documents, code and such in Git. It not only gives you revision history, it also acts as an extra backup. I've found this to be helpful even for schematics although they are not as easy to compare as source code, but it is doable. I've lost count of the times where I've reverted back to a previous version to start over. Perhaps commercial tools have this built in, KiCad does not.

I have been burnt a couple of times with using the cloud, either when the vendor puts their price up to ridiculous levels, or throttles their service, so I now use a NAS, with dual redundancy.

Configuration control is not that difficult: just give all your items a date and issue, and anything that you think may be vaguely useful just save it and start a new issue.

I'm not surprised it takes a lot  time to compile the list, god knows I've spent hours just trying to select these and I had good help from you.

Glad the list was a help.

I have kept lists of useful components since before home computers or the internet were even thought of. The way it works is that if I see a component that looks useful, I just store the datasheet. For example many EEV members mention components that are interesting.

On a new project I, or someone else, generates a list of preferred components for that project, so that everyone working on the project chooses components from the project preferred list first. For example, all decoupling capacitors may be 100nF, 35V working, X7R.

Then there are component lists of potential components for a particular design, as in the list of 3V3 logic level NMOSFETs.

All this may sound pretty heavy, but it pays big dividends and saves a lot of time in the long run. And if you are just keeping lists for home electronics, there should not be that many components anyway.

It can be quite interesting, at least I think so,  learning about new components, some of which, like the OPA192 series of opamps, depletion mode NMOSFETs, and the relatively new GaN MOSFETs, which are becoming available at a reasonable price, are game-changers and open up new design avenues.

[permal]

Ok, so taking inspiration from @Spec's schematic in an earlier thread I've ended up with this. Just added a pull down to the left mosfet and the 5/12V switch so what acould possibly go wrong?

[spec]

The switching circuit was designed for 3V3 and 5V systems (max 7V), so it will not be suitable for 12V switching because the VGSmax of the PMOSFET of 8V will be exceeded. For the same reason the input control voltage must not exceed +-7V.

The circuit can be modified to handle 12V or another PMOSFET can be used with a VGSmax of 13V or greater (20V most likely).

The same goes for the NMOSFET if the input voltage will exceed +-7V.

Taking about preferred component lists, a 6V8 zener diode type CZRQR52C6V8-HF is always handy to have for gate protection (in general, get a range of zener diodes).

Another consideration is speed. The circuit, as is, will be reasonably fast but for higher speeds the 120R gate stopper resistors could be reduced to 10R and the 4k7 pull-up resistor value could be reduced.

If you post a specification for a switching circuit to suit all your requirements, we can have a look at that, or do you want to do the design yourself? We can talk you through it if necessary.

http://www.vishay.com/docs/72024/72024.pdf

http://www.vishay.com/docs/65900/SI2312CD.pdf

http://www.comchiptech.com/admin/files/product/CZRQR52C2-HF%20THRU%20CZRQR52C39-HF-RevB.pdf

[spec]

Here is one way to do it (R3/D1 and R7/D2 must be mounted on their respective MOSFET terminals, or as close as possible, using as short leads as possible):

[permal]

Wow, do I feel the fool now. 

Thank you for your patience, @Spec.

The circuit as I drew it is what I need - 5 and 12V output so your latest schematic will do the trick.

Questions:

I suppose D2 strictly needed since I'll be controlling that only via the MCP23107 at 3.3V?

...as close as possible, using as short leads as possible

Can please you elaborate on the reason behind that?

[spec]

Wow, do I feel the fool now. 

Not at all. I wont tell you about the mistakes I have made.

Thank you for you patience, @Spec.

No sweat

The circuit as I drew it is what I need - 5 and 12V output so your latest schematic will do the trick.

That's good. By the way you can always trade speed for lower current consumption.

I suppose D2 strictly needed since I'll be controlling that only via the MCP23107 at 3.3V?.

Yes, and you can drop the input resistor value back to 120R.

...as close as possible, using as short leads as possible

Can please you elaborate on the reason behind that?

Sure

MOSFETs are odd. They are as fast as hell, but they have huge parasitic capacitances from drain to source, drain to gate, and gate to source. These capacitances are not in the tens of pF like for BJTs (normal transistors). Instead they are in the nF range- take a look at the data sheets. The consequence of this high frequency response and high parasitic capacitances is that MOSFETs tend to oscillate, typically from 500kHz to 20mHz, if their physical layout in not compact and you do not include a gate stopper resistor. They may not oscillate continuously, but only at certain voltages, temperatures, or currents.

Because traces/wires have inductance and capacitance this also adds to the chance of parasitic oscillations, so you must make traces/wires as short as possible.

I have said this many times before, but in general, the nice circuit that you see depicted in schematics is nothing like the actual physical circuit that you build, because there are parasitic resistances, capacitances, and inductances all over the place. This is especially the case for prototype breadboards with sockets and jumper leads.

There is a further complication: skin effect. At DC, current flows through the whole cross-section of a trace/wire, but as the frequency increases, say from 1kHz upward, skin effect causes the current to bunch towards the outer surfaces of the traces/wires, and the skin effect increases with frequency. The net result is that the nice conductor you see on the schematic is an even less nice resistor at high frequencies. The skin effect is the reason why Litz (insulated multi-strand wire) is used to wind high-frequency transformers.

https://en.wikipedia.org/wiki/Litz_wire