How too Drive j111 from a PCF8574

Discussion in 'General Electronics Chat' started by RickH, Nov 3, 2011.

  1. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    I am trying to figure out how to make the 3.2V output of my PCF8574 I/O Expander chip drive a -15V to controle a bunch of J111 JFET switches.
    I'm looking for cheap and small parts count for this.

    Would I just use an opto-coupler or just a transistor and if so whats the best way to wire it up.
     
  2. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Not sure exactly what you expect with so little information, but you might look at digital isolators. They're a relatively new technology, and they are very fast compared to optocouplers.

    Here's a design guide from TI:
    http://www.ti.com/lit/an/slla284/slla284.pdf
     
  3. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    The ISO72xx family is way overkill for what I am looking to do.
    Not sure what info you need but I'll describe it in detail.

    I have a Propeller uC with an I2C buss and 8 PCF8574 I/O Expander chips, all but 4 of the I/O are to be used as outputs to turn on/off the j111 jfet switches. The jfet switches require a -15V signal to turn them off and a 0V signal to leave them on, thier Depletion mode jfets. What I would like is a cheap array of switches that can be controlled by a 3.2V output from the PCF8574 I/O Expander chips to switch -15V to the j111's to turn them off.
     
  4. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
  5. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    Its kinda what I was thinking but I was looking more for an array of transistors in one package with 8 or more transistors.
    How would I bias the 3.2V out? I need 0 to -12V.
    [​IMG]
     
  6. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    This is basically what I ended up with after biasing and such, its touchy though, small changes in resistance through it off.[​IMG]
     
  7. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    That's not legal; you have the transistor inverted. You're breaking down the base-emitter junction. With real components, that would permanently reduce the gain of the transistor.

    Besides, run it over temperature, and you'll see how unstable it is.

    Here's a way you can do it using two transistors, two resistors:

    [​IMG]

    You obviously want to use the left schematic.
     
  8. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    OK thanks, only problem is now I'm at 120 Transistors and 120 Resistors lmao, gonna be a big switching board:)

    Not sure how much heat I would be generating driving JFETS, I am under the imprecision their Voltage driven with little to no current draw.
     
  9. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    If you can come up with a design that uses less than 3 resistors and 2 transistors AND that is a conventional circuit which is well within the specifications of all components, use it!

    Note that I said 3 resistors; I realized that I didn't have a resistor on the input emitter. That would have lead to very high currents when your output was high. Here is a revised circuit:

    [​IMG]

    You may need to experiment a bit with R3; increasing it to ~2.7k will kill the output. I don't know for certain what your output voltage will be under various load currents. The lower the output voltage, the lower resistance you will need. 1k Ohms will probably work under most circumstances. 2k Ohms would decrease the power consumption, but like I said there are a number of variables that I simply don't know about; you are going to have to validate the circuit design using actual components; this is just a starting place.

    Hardly any. You could probably even increase the R1/R2 values; as it is R2 will pass around 1.2mA, R1 will pass ~1mA. That's ~33mW for both.
     
    Last edited: Nov 4, 2011
  10. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    Thanks for your help SgtWookie, I think I'm gona have to go back and open my textbooks on transistors. I practically never use them and its been 20 years out of school. Think I need to touch up on my transistor theory.

    I should have this prototyped buy Sunday and I'll post my Layout and parts count if anyone would be interested.
     
  11. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    I just added more text to my previous reply; please re-read it to make sure you got the new material.

    You can use 2N3904, 2N4403 transistors, or many other small-signal types.

    [eta]
    You can get the above transistors in small SMD packages. You can also get them in 2/package and 4/package, but the price goes up astronomically. For example, Mouser has MMBT3906 single transistors for $0.02/ea. The dual transistor packages start out at $0.12/ea for the MMBT3906SL; 6x as much. They only carry one quad 3904 package; MMPQ3906 for $1.40/ea, or $0.35 per transistor. That's over 17 times more expensive than the individual transistors.
     
    Last edited: Nov 4, 2011
  12. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    As always, That's why I have stayed away from transistors for so long. With op-amps the math is straight forward but with Transistors you need to know the gain and currents and voltages and such. I'm almost tempted to just use inverting op amps to drive the j111's but that should be super overkill.

    Mouser has TL084's for5 23 Cents, that's 5.75 Cents per switch and 2 resistors. I would have to add a -18V rail to get the -15Volts though due to it being an opamp, but that's just one more regulator at the power supply, not to big a deal.

    [​IMG]
     
  13. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    AvnetExpress has TL084C's for ~$0.18/ea
    CD - 14-pin SOIC
    CN - 14-pin PDIP
     
  14. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    I haven't tried Surface mount yet, but that video Gadget Gangster did on using a hot plate looks like a cool idea, I may give it a try also this week.

    So would you agree this would be the easiest way to do it then? I feel like it should be overkill but in all reality I haven't found a simpler way.
     
  15. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    It's 9 parts for 4 channels, PLUS you will need an 0.1uF cap across the supply pins; total 10 parts per 4 channels; 150 parts for all 60 channels, 2.5 parts/channel.

    You can use 0805 size SMT resistors if you're handy with tweezers.
    You could use resistor arrays, too. Here's four 10k resistors in an 8-terminal 0804 SMD package; $0.23 for 10:
    http://search.digikey.com/us/en/products/CRA06S08310K0JTA/CRA6S810KCT-ND/1285853

    There are similar available for the 47k resistors. 4 parts for 4 channels; 60 parts total? Is that more like what you're looking for?

    [eta]
    There are also people using toaster ovens to solder these things together.

    You may want to stick with individual resistors for at least one of the values though, as using arrays for both could make routing traces on the board just plain nasty.

    Thru-hole resistors would be a snap to route, as you could even use them as jumpers.
     
    Last edited: Nov 4, 2011
    RickH likes this.
  16. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    Dual sided it comes out too 2.5x2.75" board for 8 switches, thats not too bad hu?

    [​IMG][​IMG]
     
  17. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    What are you using for schematic & board layout?
     
  18. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    Eagle PCB, its free from cadsoft, I like it much more than Design Spark.
    Design Spark pro's
    1- Its free and any board size is possible
    2- The component Library is extremely easy to use, Packages/Symbols/Devices are all in their own Libraries so you don't need to keep drawing an 8 pin Dip every time you start a new library
    Cons
    1- Its clunky as all get out, its not smooth on panning or scrolling, placement of components is clunky.
    2- copper pours are a total pain in the but, it will make islands of copper if you don't set the properties correctly.
    Eagle PCB pros
    1- It feels like AutoCAD to me, its smooth once you figure it out.
    2 - Design Rules are a breeze to set
    3 - Copper pours are super simple once you understand them.
    4- Its printer setup works great for Laser jet printing
    Cons
    1- Max board size is small for free version 3.2"x4" ( you can draw outside of the max area but no pads from components can be placed out side max area)
    2- The library is a pain to manage (But it comes with a massive library)
     
  19. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    I thought it looked like an Eagle schematic, but your symbols look a bit different from the default libraries - and you're using REALLY large pads for everything. That's OK though, as you don't have to do any routing between IC pins.

    I try to avoid having more than one trace per pad or via. For example, on layer 1 (top) of your board, on the very bottom, R2 has just one trace to the pad on the left, but two to the pad on the right. Having two traces on the pad means you'll have to apply more heat to get the solder to melt/flow, which increases your risk of lifting a pad, ruining the board. Sometimes, it's just not feasible, like pin 11 of IC1 & IC2 because the pads are so large.

    By the way, you don't have a bypass cap on IC2.
     
  20. RickH

    Thread Starter New Member

    Dec 10, 2010
    21
    0
    Yes I made all my own Library's mostly, I don't trust a lot of the eagle ones, I have run into problems.
    All my pads are .07" and I could use long for IC's but I don't genraly run traces between pins, its just something I have gotten used too. I like the large pads becuss I am getting old and soldering is much easyer and more forgiveing on large pads.

    I typicaly don't run multiple traces too a pad either mostly becuss it hids the pad too much for me but I just did this layout today and haven't optimized it yet. I will typically do a layout until I'm satisfied and then come back too it the next day to check it and optimize it. Same concept of having some one else bleed on your drawings, its always easier for someone else to tell you whats wrong with your work lol.

    Also I don't add bypass caps to every op amp in most any circuit. their is a bypass at the farthest op amp and then their is the 1pF or so bypass from the ground and power plain(I added this island under the Op amps after I posted it. I think for this circuit it should be enough. I do a lot of audio stuff and I almost never add more than 1 cap(Per Rail) per 2 IC's and one set of Electrolytic per board.
     
    Last edited: Nov 4, 2011
Loading...