ESP8266 Community Forum

- Sat May 16, 2015 5:09 am #17541 Yep, the FDS8896 looked to be a nice fit here. Looking at the datasheet, at 3.3V gate->source, the Rds(on) seems to be around 7.5mOhm. I've not measured how much current it can actually handle, but here's some theory:

Assuming there is no heatsinking done on the print board whatsoever, the chip heats up 125 degrees celsius for every Watt of power it has to dissipate. At 7.5mOhm Rds, that means you can run 8 Amps through it to make it heat up by 60 degrees more than ambient temperature (which would make it definitely hot to the touch, but not hot enough to do any harm).

Here's a graph of the above scenario (Y is temperature rise in degrees celsius, X is the amount of Amps): http://www.wolframalpha.com/input/?i=y+ ... %3D0+to+10

Now, this design does have some heatsinking built into it, but I can't say for sure how effective it's going to be. So the heat up is probably even less than the graph shows, we'll have to test that.

My LED strips are around 11 Watt per meter I think, so if you have comparable ones 5 meter would be 55 Watt, or around 4.6 Amps. This should be well within the boundaries of what this module can handle.

The programming lanes on the back are just for emergencies, I intend to write the software in such a way that we can just reprogram them over wifi. I only use the lanes once at the moment, for initial flashing. They are spaced just like dupont wires without the connectors on, so it's easy to temporarily press wires onto the lanes, or solder them on if you want permanent access.

- Sat May 16, 2015 7:40 am #17552

Yep, the FDS8896 looked to be a nice fit here. Looking at the datasheet, at 3.3V gate->source, the Rds(on) seems to be around 7.5mOhm. I've not measured how much current it can actually handle, but here's some theory:

Assuming there is no heatsinking done on the print board whatsoever, the chip heats up 125 degrees celsius for every Watt of power it has to dissipate. At 7.5mOhm Rds, that means you can run 8 Amps through it to make it heat up by 60 degrees more than ambient temperature (which would make it definitely hot to the touch, but not hot enough to do any harm).

Here's a graph of the above scenario (Y is temperature rise in degrees celsius, X is the amount of Amps): http://www.wolframalpha.com/input/?i=y+ ... %3D0+to+10

Now, this design does have some heatsinking built into it, but I can't say for sure how effective it's going to be. So the heat up is probably even less than the graph shows, we'll have to test that.

My LED strips are around 11 Watt per meter I think, so if you have comparable ones 5 meter would be 55 Watt, or around 4.6 Amps. This should be well within the boundaries of what this module can handle.

The programming lanes on the back are just for emergencies, I intend to write the software in such a way that we can just reprogram them over wifi. I only use the lanes once at the moment, for initial flashing. They are spaced just like dupont wires without the connectors on, so it's easy to temporarily press wires onto the lanes, or solder them on if you want permanent access.

Yes, this FET is a very interesting and good choice. I agree with your thermal estimates, but think they are valid only for very low duty cycle; Figure 8 in the datasheet says "Pulse duration 80 µs / Duty cycle = 0.5% max". Fig 9 shows the correction factor for increased junction temperature.

An other way to do a conservative estimate would be to assume minimum mounting pad, and use the figure of Note 2b), which gives junction temperature 125°C when dissipating 1W. Assuming Rds_on 7.5 mohm at Vgs=3.5V and low duty cycle (=low Tj), using the correction factor from Figure 9, we should expect Rds=10,1 mohm. This gives Imax=sqrt(1/0.0101) = 9.9 A Which is quite impressive! When using this in a real.world circuit I would derate that 50% and say max 5 A. Still impressive!

And I agree with you that it can power a 5 meter LED strip at full blast.

- Sat May 16, 2015 5:08 pm #17596

Yes, this FET is a very interesting and good choice. I agree with your thermal estimates, but think they are valid only for very low duty cycle; Figure 8 in the datasheet says "Pulse duration 80 µs / Duty cycle = 0.5% max". Fig 9 shows the correction factor for increased junction temperature.

An other way to do a conservative estimate would be to assume minimum mounting pad, and use the figure of Note 2b), which gives junction temperature 125°C when dissipating 1W. Assuming Rds_on 7.5 mohm at Vgs=3.5V and low duty cycle (=low Tj), using the correction factor from Figure 9, we should expect Rds=10,1 mohm. This gives Imax=sqrt(1/0.0101) = 9.9 A Which is quite impressive! When using this in a real.world circuit I would derate that 50% and say max 5 A. Still impressive!

And I agree with you that it can power a 5 meter LED strip at full blast.

Wow, someone who can actually make a calculation with these values. I've been trying for a while, but try as I might, I'm not able to fully figure it out. In the end I chose a different module then tjclement because I would like them to support 40v instead of 30v max.

If you would be so kind, what values would the FDS8447 hit when using them in the same scenario as tjclements board? I would want to use them at 12v for regular LED strips and 38.5v for COB LED's. But I have no clue what kind of thermals I need to look at.

Sorry to hijack your topic tjclement.

- Sun May 17, 2015 1:52 am #17616

If you would be so kind, what values would the FDS8447 hit when using them in the same scenario as tjclements board? I would want to use them at 12v for regular LED strips and 38.5v for COB LED's. But I have no clue what kind of thermals I need to look at.

Sorry to hijack your topic tjclement.

Yes, datasheets can be useful.
The FDS8447 is specced slitghtly differently, but the data are there. You can a worst case Rds_on of 20 mohm, which will enable you to run 7 A with the chip at 125 deg (hot but technically safe). Personally, I would use it for up to around 3.5 A current steady state (somewhat cowardishly derating 50%).