Analyzing Internals of Analog Semiconductors - Mystery Component

Discussion in 'General Electronics Chat' started by MachineHum, Mar 28, 2015.

  1. MachineHum

    Thread Starter Member

    Nov 3, 2014
    70
    4
    Hey, does anyone know the component I've marked out on this schematic? It's TI's LM386, my best guess is current source? I'm trying to analyze the internals of how this amp works, I'm unsure as to what's going on during the positive half of the waveform input...

    http://www.ti.com/lit/ds/symlink/lm386.pdf
    upload_2015-3-28_13-3-56.png
     
  2. bertus

    Administrator

    Apr 5, 2008
    15,646
    2,345
    Hello,

    That is the symbol for a constant current source.

    Bertus
     
  3. AnalogKid

    Distinguished Member

    Aug 1, 2013
    4,523
    1,247
    That is indeed a current source pulling up the quasi-complementary symmetry output stage. Very common in opamps, and in most discrete audio amps. Remember, what you are looking at is not the full schematic of the part, just enough of a representation to give a designer an idea of how it behaves internally.

    ak
     
  4. OBW0549

    Well-Known Member

    Mar 2, 2015
    1,308
    884
    Yes, that's a constant current source, most likely a current mirror of some sort.

    If you're trying to analyze the innards of the LM386, I developed and posted a SPICE subcircuit model for it on the sci.electronics.design Usenet group 20 years ago or so:

    https://groups.google.com/forum/#!topic/sci.electronics/Ssv3p_PDaw8

    (last post in the thread)
     
    Last edited: Mar 28, 2015
    Roderick Young likes this.
  5. #12

    Expert

    Nov 30, 2010
    16,262
    6,769
  6. OBW0549

    Well-Known Member

    Mar 2, 2015
    1,308
    884
    Thanks.

    I had pretty much the same learning experience. A pretty significant chunk of my electronics knowledge and understanding came from studying (and scratching my head over) the IC internal diagrams in National Semiconductor's datasheets, and Linear Technology's also. There's a lot of good stuff in there.

    Regarding those "little circuits," IMO they're the backbone of being a productive design engineer. Anyone who does analog for a living for very long, I suspect, builds up a mental "library" of scores, or even hundreds, of those basic functional building blocks. And much of the design work I do, comes in selecting the appropriate combination of basic functional blocks and tailoring them for the specific task at hand.

    And after 50 years working in the field, I'm still learning new ones and filing them away in my cluttered brain...
     
    KLillie likes this.
  7. MachineHum

    Thread Starter Member

    Nov 3, 2014
    70
    4
    I've tried analyzing it using currents and my transistor knowledge, but it turns into a huge equation that's impossible to solve and doesn't really help the understanding. Do you have a resource for understanding these building blocks? Are most of the transistors operating in their saturated region? IE either fully "on" or fully "off" or in their "linear" region, running up and down the load line?

    I'm also getting into transconductance amps, so if the knowledge crosses over (which it usually does) bonus points
     
  8. OBW0549

    Well-Known Member

    Mar 2, 2015
    1,308
    884
    That's why SPICE was developed, because trying to understand circuit operation in the "equation domain" quickly becomes useless-- and insanity-inducing-- once you get past a couple of transistors. Linear Technology, Inc. has a free SPICE program, LTSpice IV, that a lot of people use.

    Get a copy of The Art of Electronics by Horowitz & Hill. Download and study every application note you can find on the Linear Technology, Texas Instruments, Analog Devices and Microchip websites. Lots and lots of good information there.

    Except when the amplifier is overdriven and its output is clipping (i.e., slammed up against the positive supply rail or down against the ground rail), all transistors in that circuit are operating in their linear region.
     
    Last edited: Mar 31, 2015
  9. #12

    Expert

    Nov 30, 2010
    16,262
    6,769
    My favorite current generator is called, "jfet" or, Junction Field Effect Transistor.
    You connect the gate to the source pin and you're done. One part, one connection.
    Jfets are born in the, "on" condition. You can buy them idling from 50 ua to 150 ma.
    I expect they are a favorite of the IC designers.
     
  10. bertus

    Administrator

    Apr 5, 2008
    15,646
    2,345
  11. MachineHum

    Thread Starter Member

    Nov 3, 2014
    70
    4
    ^Awesome link
     
    JWHassler likes this.
  12. WBahn

    Moderator

    Mar 31, 2012
    17,720
    4,788
    Not if they aren't available in the process you are using. If you are designing CMOS circuits then you are very limited in the devices you can draw upon and JFETs are almost never on the list. Mostly you use enhancement mode PFETs and NFETs. You also have some ability to make resistors and capacitors. The linear capacitors tend to be very low valued and the larger ones, which use the gate capacitance of FETs, are very nonlinear. You can usually made lateral NPN transistors but they aren't very good. They are useful for things such as bandgap references, pad protection structures, and thermometers. You often have the ability to use different implant masks to get different threshold voltages. Other than that it becomes very process specific.
     
  13. studiot

    AAC Fanatic!

    Nov 9, 2007
    5,005
    513
    Good link ,Bertus.

    I doubt the current mirror in the output OBW0549.
    What would be the point?
     
  14. WBahn

    Moderator

    Mar 31, 2012
    17,720
    4,788
    It would be an active load for that stage.
     
  15. OBW0549

    Well-Known Member

    Mar 2, 2015
    1,308
    884
    Are you talking about I1, Q8 and Q9 in the model I linked to? That's how I modeled the current source for driving the upper NPN output transistor, Q13. I could have omitted Q8 and Q9 and connected I1 directly to the base of Q13 (in other words, exactly as shown in the diagram in the top post of this thread), but if I'd done so it would have allowed the base of Q13 to be driven above the +Vs rail and therefore Q13 to be driven into saturation, which wouldn't have been realistic.
     
  16. studiot

    AAC Fanatic!

    Nov 9, 2007
    5,005
    513
    No I was talking about the TI diagram presented by the OP and the current source between the base of the upper LIN output transistor and the + rail.

    I further doubt that all transistors are operating in their linear region.
    This is definitely contradicted by the TI curve of output v supply voltage in the OP linked datasheet.

    At least the output ones are active, yes, but, like all large signal amplifiers, non linear.
     
  17. #12

    Expert

    Nov 30, 2010
    16,262
    6,769
    It seemed like a good answer while I was looking at post #1
     
  18. WBahn

    Moderator

    Mar 31, 2012
    17,720
    4,788
    My guess (and I don't know this for sure) would be that the availability of JFETs on a bipolar process would be even harder to come by. Although I know that National developed the Bi-FET process that allowed them to fab reasonable JFET devices on a bipolar process. I believe that nearly all bipolar processes that support JFETs are derivatives of their work (in the 70's). But I don't know how widespread such processes are. Even if the standard process physically permits JFETs to be realized, there would be additional masking steps, which are very expensive, that most people are not going to want to pay if they don't have to.
     
  19. #12

    Expert

    Nov 30, 2010
    16,262
    6,769
    I don't know Jack about fabrication. I only know that I love j-fet input op-amps. They enable common mode inputs outside the power supply range, have low noise, high input impedance, and enough speed for audio work. They are a Ferrari compared to an LM741.

    Don't bother arguing. I'm emotional about this one. :D
     
  20. WBahn

    Moderator

    Mar 31, 2012
    17,720
    4,788
    Oh, I agree. JFET input opamps are definitely sweet. For a variety of reasons JFETs have never really captured a portion of analog design that they probably should have from a pure performance standpoint. Mostly it's because their uses, in high volume and/or integrated designs, is enough of a niche realm that other, more widely available and low cost, options are at least "good enough".
     
Loading...