Post by thread:[EE] SOT-23 P-channel MOSFET?

At 01:54 PM 9/12/2005 -0500, you wrote: >Looking for a suitable one of these. Application will be to have a PIC >output >pin drive an LED display column (up to 200mA). Something "mirror-image" to >the n-channel MMBF170 would be great. > >So far I've found the FDV304P and BSS83P, but hoping for something >lower-cost. >Any ideas? > >Cheers, >-Neil. If 100mA would do, the BSS84 is an inexpensive workhorse. Bipolar transistors (eg. MMBT4403) are also attractive, even cheaper, and can handle your 200mA with PIC-level base drive. Best regards, Spehro Pefhany --"it's the network..." "The Journey is the reward" spam_OUTspeffTakeThisOuTinterlog.com Info for manufacturers: http://www.trexon.com Embedded software/hardware/analog Info for designers: http://www.speff.com ->> Inexpensive test equipment & parts http://search.ebay.com/_W0QQsassZspeff

Nice part -- and I might squeeze squeeze by on the current -- max of 200mA is rounded up from a calc'd max of 150mA with a 20% max duty-cycle, and Fairchild's datasheet for this claims 130mA max continuous and 520mA max pulsed. It's hard to resist the cost difference, so I'll have to try it. I'm sure adjusting to 130mA won't have any noticeable difference in the LED brightness. I'm actually going the other way -- currently working well with MMBT2222's and MMBT3906's, but I need to add some extra components to the next version of this app and have no more board space. Using MOSFETs, I can eliminate the base/gate resistors to free up some more space. Yes, the board is that tight. Cheers, -Neil. On Monday 12 September 2005 03:16 pm, Spehro Pefhany scribbled: > If 100mA would do, the BSS84 is an inexpensive workhorse. Bipolar > transistors (eg. MMBT4403) are also attractive, even cheaper, and can > handle your 200mA with PIC-level base drive.

At 04:04 PM 9/12/2005 -0500, you wrote: >I'm actually going the other way -- currently working well with MMBT2222's >and >MMBT3906's, but I need to add some extra components to the next version of >this app and have no more board space. Using MOSFETs, I can eliminate the >base/gate resistors to free up some more space. Yes, the board is that >tight. You could always use dual 'digital' transistors with built-in bias resistors. Made by Rohm, Panasonic and others, and a dual is smaller than an SOT-23, but I suspect you'll find them somewhat less attractive than the MOSFETs. I tend to use jellybean BJTs and resistor arrays, partly out of ESD considerations. Best regards, Spehro Pefhany --"it's the network..." "The Journey is the reward" .....speffKILLspam@spam@interlog.com Info for manufacturers: http://www.trexon.com Embedded software/hardware/analog Info for designers: http://www.speff.com ->> Inexpensive test equipment & parts http://search.ebay.com/_W0QQsassZspeff

On Monday 12 September 2005 04:58 pm, Spehro Pefhany scribbled: > You could always use dual 'digital' transistors with built-in bias > resistors. Made by Rohm, Panasonic and others, and a dual is smaller than > an SOT-23, but I suspect you'll find them somewhat less attractive than the > MOSFETs. I saw some of these earlier (first time I ever saw that term), but they did not offer any cost advantage. I did look for arrays in SO-8 packages, but still not noticeable cost advantage (though they would be easier to solder than SOT-23's). > I tend to use jellybean BJTs and resistor arrays, partly out of ESD > considerations. For exactly that reason, I still intend to keep the resistor/MMBT2222 combination for input signals (on/off), but I also found some FETs today that included an ESD-protection zener built in -- I'm still not sure if to try it for the external signals though. Cheers, -Neil.

On Mon, 12 Sep 2005 17:22:26 -0500, you wrote: {Quote hidden}>On Monday 12 September 2005 04:58 pm, Spehro Pefhany scribbled: >> You could always use dual 'digital' transistors with built-in bias >> resistors. Made by Rohm, Panasonic and others, and a dual is smaller than >> an SOT-23, but I suspect you'll find them somewhat less attractive than the >> MOSFETs. > >I saw some of these earlier (first time I ever saw that term), but they did >not offer any cost advantage. I did look for arrays in SO-8 packages, but >still not noticeable cost advantage (though they would be easier to solder >than SOT-23's). > > >> I tend to use jellybean BJTs and resistor arrays, partly out of ESD >> considerations. > >For exactly that reason, I still intend to keep the resistor/MMBT2222 >combination for input signals (on/off), but I also found some FETs today that >included an ESD-protection zener built in -- I'm still not sure if to try it >for the external signals though. > >Cheers, >-Neil. I really like the Zetex ZXM61P03 and N03 parts - cheap and very good speed and rdson performance. Zetex also do some duals in a SOT23-6.

In this app, a single column output from the PIC will be connected to the gates of 2 mosfets -- one P-channel and one N-channel. The appropriate LED digit will be selected by the high or low state on that output pin. Cheers, -Neil. On Monday 12 September 2005 10:51 pm, William "Chops" Westfield scribbled: {Quote hidden}> On Sep 12, 2005, at 1:16 PM, Spehro Pefhany wrote: > >> Looking for a suitable one of these. Application will be to have a > >> PIC > >> output pin drive an LED display column (up to 200mA). > > Interestingly, the LED display example on piclist James pointed me at in > reply to MY question had N-channel MOSFETs in that position (yeah, even > with common anode LEDs.) I guess you can more-or-less guarantee Vsource > (LEDs being non-resistive), so you can ensure enough Vgs for that to > work? > > http://www.piclist.com/techref/io/led/8x7s8pin.htm > > BillW

>>> Looking for a suitable one of these. Application will be to have >>> a PIC output pin drive an LED display column (up to 200mA). > Interestingly, the LED display example on piclist James pointed me > at in > reply to MY question had N-channel MOSFETs in that position (yeah, > even > with common anode LEDs.) I guess you can more-or-less guarantee > Vsource > (LEDs being non-resistive), so you can ensure enough Vgs for that to > work? > > http://www.piclist.com/techref/io/led/8x7s8pin.htm As you infer, that's a rather "naughty" way to use those FETs. The end result is probably acceptable in many cases but is not "tightly designable" . It will work, of course, but it leads to a rather undefined LED state. The basic design is quite clever. With one I/O pin low and one high and the rest floating only one LED is driven. No problem with that. BUT the FET is driven by the difference between its gate - driven to Vdd (presumably 5 volts) and its source which is "on top" of the relevant LED. The FET needs Vth minimum to turn on and somewhat more than this to provide LED current. How much more than this will vary with FET, batch, temperature, current and more. Iled ~= (Vdd - Vled - Vgs)/ 220r But Vled will vary with current, batch, temperature etc A more controlled result could be produced using the same scheme with the FET drains grounded and common cathode LEDs or possibly common anode LEDs still and a bit of head scratching. I'll leave that to others as i should be working. RM

At 08:51 PM 9/12/2005 -0700, you wrote: >On Sep 12, 2005, at 1:16 PM, Spehro Pefhany wrote: > >>>Looking for a suitable one of these. Application will be to have a PIC >>>output pin drive an LED display column (up to 200mA). >Interestingly, the LED display example on piclist James pointed me at in >reply to MY question had N-channel MOSFETs in that position (yeah, even >with common anode LEDs.) I guess you can more-or-less guarantee Vsource >(LEDs being non-resistive), so you can ensure enough Vgs for that to work? > >http://www.piclist.com/techref/io/led/8x7s8pin.htm Ugh, that's a terrible idea. To be fair, it does clearly say "untested". You need only glance at the Vgs(th) min/max/typ figures and typical on- region characteristics to see how bad it really is. http://www.fairchildsemi.com/ds/2N/2N7000.pdf >Best regards, Spehro Pefhany --"it's the network..." "The Journey is the reward" speffKILLspaminterlog.com Info for manufacturers: http://www.trexon.com Embedded software/hardware/analog Info for designers: http://www.speff.com ->> Inexpensive test equipment & parts http://search.ebay.com/_W0QQsassZspeff

At 02:04 AM 9/13/2005 -0700, you wrote: >On Sep 13, 2005, at 12:34 AM, Russell McMahon wrote: > >>Iled ~= (Vdd - Vled - Vgs)/ 220r > >Huh? Vgs doesn't enter into the LED current. Vds does, but the FET >is basically ohmic if Vgs is past threshold: > Iled ~= (Vdd - Vled) / (220 + Rfet) Yabbut, Vgs(th) for only 1mA can be anything from 0.8V to 3.0V, and *typically* to get 100mA anode current you need at least 4V (see curves on datasheet). Some current will flow with a red display, but the display will be very dim and probably very uneven. >But we're in violent agreement. My initial reaction was that it was >just wrong, and then on further thought I decided it would probably >work anyway (but don't put a blue display in there!) I suspect that a blue LED display will give much more consistent results. ;-) Best regards, Spehro Pefhany --"it's the network..." "The Journey is the reward" .....speffKILLspam.....interlog.com Info for manufacturers: http://www.trexon.com Embedded software/hardware/analog Info for designers: http://www.speff.com ->> Inexpensive test equipment & parts http://search.ebay.com/_W0QQsassZspeff

> On Sep 13, 2005, at 12:34 AM, Russell McMahon wrote: >> Iled ~= (Vdd - Vled - Vgs)/ 220r > Huh? Vgs doesn't enter into the LED current. Vds does, but the FET > is basically ohmic if Vgs is past threshold: > Iled ~= (Vdd - Vled) / (220 + Rfet) 'Fraid not. IF a FET is driven from an unrelated voltage source, as is usually the case, you would be totally correct. However, in this case you need to consider how the circuit works. The FETs are driven in "common drain" mode - effectively 'emitter followers'. As the gates are at V+ the sources must be > Vth below them. I say ">" as Vth is specd for typically <<1 mA and we want 10's to 100's of mA here so the FEt needs a bit more drive. Vth varies depending on whose data sheet you read, and the current you spec it at, and the die temperature - all of which is my point. And the difference between typical and max is non trivial. So the "available" voltage is Vdd less whatever the FEt needs across gs to drive it at the requisite current. You then drop the LED voltage, which cane asily vary 10% between various devices, and the balance is dropped across the current limiting resistor - which isn't going to have a very energetic job in this case :-). And that doesn't count any driver drop (and we have source and drain to consider). All up Vgs may well be 2v and could be 2.5v (depends on spec sheet used) (but could be only 1v. Rdson is several ohms for this FET - probably not too important here. Vled depends on LED but lets say 1.3v. That gives 5 - 1.3 - 2 = 1.7v to share amongst current limiter, drivers, wiring, Rdson and Murphy. Use a LED with higher voltage drop than 1.3v and your headroom diminishes. BUT drive the FETS in common source mode and it all comes right. ie use PFETS here with some drive signal changes or rearrange with FETS at the bottom. >> But Vled will vary with current, batch, temperature etc >> > Not by much... Agree. But by enough to matter ;-(. > But we're in violent agreement. My initial reaction was that it was > just wrong, and then on further thought I decided it would probably > work anyway (but don't put a blue display in there!) Or white, or green? or ... :-) > And "charlieplexing" is a neat hack... Indeed. it's a great idea. but it needs work. Most great ideas do :-) Talking off the top of my head, so this may be wrong, but: Someone else can draw the diagrams ;-). Using PIC ports as direct drivers is going to severely limit your multiplexer current. having the FET in one leg to supply doesn't help if the return is to a PIC pin. What you "really" want is a row/column matrix of drivers that the PIC drives in similar manner to this but the actual display current goes via drivers alone and not the PIC pins. You then only need ONE current limiting resistor ;-) as there is only 1 segment lit at a time. RM

At 11:03 PM 9/13/2005 +1200, you wrote: >Talking off the top of my head, so this may be wrong, but: Someone else >can draw the diagrams ;-). Using PIC ports as direct drivers is going to >severely limit your multiplexer current. having the FET in one leg to >supply doesn't help if the return is to a PIC pin. What you "really" want >is a row/column matrix of drivers that the PIC drives in similar manner to >this but the actual display current goes via drivers alone and not the PIC >pins. You then only need ONE current limiting resistor ;-) as there is >only 1 segment lit at a time. Bad juju to have too low duty cycle. >Best regards, Spehro Pefhany --"it's the network..." "The Journey is the reward" EraseMEspeffspam_OUTTakeThisOuTinterlog.com Info for manufacturers: http://www.trexon.com Embedded software/hardware/analog Info for designers: http://www.speff.com ->> Inexpensive test equipment & parts http://search.ebay.com/_W0QQsassZspeff

At 11:56 AM 9/13/2005 +0400, you wrote: > > Ugh, that's a terrible idea. To be fair, it does clearly say "untested". > > > > You need only glance at the Vgs(th) min/max/typ figures and typical on- > > region characteristics to see how bad it really is. > > > > http://www.fairchildsemi.com/ds/2N/2N7000.pdf > >BSS138 instead ? Better, but p-channel would be so much better again. Or NPN emitter follower, or PNP BJT with base resistors. Best regards, Spehro Pefhany --"it's the network..." "The Journey is the reward" speffspam_OUTinterlog.com Info for manufacturers: http://www.trexon.com Embedded software/hardware/analog Info for designers: http://www.speff.com ->> Inexpensive test equipment & parts http://search.ebay.com/_W0QQsassZspeff