Home-brew 13.8v, 30A PSU regulation prob

Discussion in 'General Electronics Chat' started by Dave_UYZ, Jan 16, 2014.

  1. Dave_UYZ

    Thread Starter New Member

    Jan 16, 2014
    29
    0
    I'm trying to finish a 13.8v dc linear PSU to run to 30A.
    [This may sound strange but I'm now retired, and had most, if not all, the bits so got started.]

    General circuit:
    The Pass transistors are MJ11016 (30A each, 250w). There are four of them on two enormous fan-cooled Heat-sinks.
    The transformer is a 7 inch C-core cube delivering about 28v. I've satisfactorily tested this to 50A load.
    Mains is applied in two stages via a small relay and a series resistor.
    The rectifier is a 50A rated device, well bolted to the chassis.
    There's about 75000uF, 40-60v rating in smoothing.
    Wiring is in thick wire (generally around 2.5mm (but it's all AWG sizes)).

    The problem as experienced:
    The regulator is a standard 723 Circuit (by GI3HXV in the RadCom Handbook); it can also be found in Fig 17 of the data sheet.

    Using a car headlamp as a first-step test load (5.7A), I have noticed that the output voltage fell from 13.8 to 13.65v,
    and I do not know why. If I have read the data sheet right, it should hold to within 15mV.
    Careful checking of the circuit showed no apparent problems.

    Can someone please give me a few pointers while the three test loads arrive from China ?
    Or have I missed something important? :confused:
    Photos of the story so far can be provided at need.

    73
    Dave
     
  2. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    So, are we to assume you connected the output of the chip directly to the bases of 4 Darlington transistors and the current limiting resistors are the right values and you measured the voltage at the beginning of the feedback divider? The actual load on the chip is 3 ma because the Darlingtons have a gain of 10,000 and the load is distributed evenly because you have a matched set of 4 output transistors?

    If that's true, it works and you measured it wrong. If otherwise, we'll need YOUR schematic.
     
  3. MikeML

    AAC Fanatic!

    Oct 2, 2009
    5,450
    1,066
    Are you remote sensing at the output terminals? Do you have a separate higher voltage supply?

    I have an Astron supply for my ham station. Here is a link to its description and schematic. It uses the 723 chip for its regulator. You might learn from the schematic.

    I have fixed several of these over the years. Just a good basic supply.
     
    Last edited: Jan 16, 2014
  4. Dave_UYZ

    Thread Starter New Member

    Jan 16, 2014
    29
    0
    Thanks, guys.

    The 723 output goes directly to the bases via a 10R in each.
    Output current from the MJ11016 is via a R1.
    Sampling of output voltage comes for the back of the output terminals on the front panel.

    I've wondered about a separate supply for the 723, but never tried it.

    I'll see what I can do about a circuit in a little while.
     
  5. Dave_UYZ

    Thread Starter New Member

    Jan 16, 2014
    29
    0
    I usually do my notes in WORD, so I've put it into a PDF
    I hope this is acceptable.
     
  6. alfacliff

    Well-Known Member

    Dec 13, 2013
    2,449
    428
    how much voltage drop across the wires connecting the load? are you testing the voltage across the output terminals of the supply, or load? did you use emitter balasting resistors on each output transistor? and a seperate supply is usually used for the 723, even if it is a seperate winding on the power transformer.
     
  7. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    What value is OR1 and where is the I lim pot set?

    You're chasing about a 1% difference. You will have to do some "before and after" measurements. Measure the reference voltage of the chip, then turn on the load and see if the reference sags. Measure pin 10 to each pass transistor base to see if they are carrying equal shares of the load. etc.
     
  8. ScottWang

    Moderator

    Aug 23, 2012
    4,850
    767
    The properly way is using the power bjt <= 1/7 rating current when we used it in a power supply, and keeping the continuing temperature raising <= 60℃.

    If you want to using less quantity of power bjt to build the power supply, the heatsink and fan is needed, and keeping the continuing temperature raising <= 65℃, if the temperature over 65℃ may not burn out the bjt, but it will reducing the using life of parts.

    The current of each MJ11016 (30A each, 250w) is I=30A/4=7.5A, actually that is a challenge, how about the size of heatsink and the V/I of fan?
     
  9. Dave_UYZ

    Thread Starter New Member

    Jan 16, 2014
    29
    0
    Output testing from behind the front panel connect, although this may have to change.
    The wire to the present load (60w) is 1mm mains wire running about 3ft from the front panel. I'll measure the volts drop tomorrow.

    The Current Limit (I lim) pot is roughly half way. Calculations indicate that the load (a 60w car headlamp) should pull about 5A and so it is indicated on the front panel Ammeter.

    Each of the four Pass stage transistors has a 0.1Ω (0R1) in the emitter output. After that the connections are commoned up.
    Since the Rear Panel of the unit is detachable, a heavy-gauge terminal block is used (as can be seen in the photo on page 9).
    All connections are 'greased' with 'Copperslip' to maintain conductivity in an adverse climate (my unheated garage).

    Future plans include using the Solid-state FET switch in the crowbar circuit (a 3423 chip, which I have mis-wired so far). The idea is to switch the output OFF in a hurry (a relay can take 300mS, by which time the radio (the load) has dissolved).

    My sincere thanks for the help so far. :)
     
    Last edited: Jan 17, 2014
  10. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    I calculate 1.14 ma through the I Lim pot for a 5.7 amp load. If it is at 500 ohms, you are delivering .57 volts to the current limiting resistor in the 723 chip. That's enough to cause a 1% error. Turn your I Lim to zero ohms for accuracy tests.
     
  11. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    The LM723 is a very old, low performance reg IC.

    The most likely cause of load regulation drift is voltage drop along ground or power leads. You have to have the regulator's reference circuitry "Kelvin connected" out to the load point when dealing with such high currents.
     
  12. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    Which is actually pretty good for the old dog LM723.:D
     
  13. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    I've pulled them in at .1%, zero to 4 amps.
    But then, I did these for a living back in the 1970's.
    I know tricks that aren't on the datasheet.
     
    Dave_UYZ and tvtech like this.
  14. ScottWang

    Moderator

    Aug 23, 2012
    4,850
    767
    What' s the secret of that tech?
     
  15. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    I think at close to 10X that current (30A) that 1% is decent.

    I could do a lot better with discrete components and references.
     
  16. Dave_UYZ

    Thread Starter New Member

    Jan 16, 2014
    29
    0
    Yes, I agree, the 723 has been around a few years, but what's to replace it that's any better? And - is it commonly available ?
    TI reckon 0.03% Vout in a data-sheet revised April 2013.

    I'll check 'duplication' of wiring, but if you mean that a separate pair come off the output terminals to the ic, they do.

    All the -ve bits come straight off the main capacitors to the -ve output terminal.

    But it's early, and I'm un-breakfasted.
    I'll dive into the garage and see if I can see more a little, later.
    Many thanks, Chaps. :)
     
  17. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    My bench supply which has load/line reg specs probably in the .01% ballpark is built using LM358 op amps and basic references similar to the LM385, and discrete devices like 2N3904 and standard NPN power transistors which a 723 design would need anyway. And my design has precise current regulation which a 723 can not do at all.

    What the LM723 does is easily replicated with discrete devices for under $1 with superior performance because the reference is not getting noise/heat from any other drivers. 40 years ago the 723 was something because it made a discrete design smaller and gave people a way to make cheap linear regs with adding a pass transistor. For cheap production designs, it makes sense.

    Unless space is absolutely critical, I would just build a superior design with the discrete parts... which is what I did. Considering the fact your heatsink is probably the size of a bowling ball, allocating room for a 2" x 2" board to hold the control circuitry isn't so terrible.
     
    Last edited: Jan 18, 2014
  18. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    As for trouble shooting the cause of your load regulation problem, here are some tips:

    1) make sure the output is not oscillating as that will give a erratic reading.

    2) Use a precision DVM to measure V drops along lines with respect to a "single point ground". See if the 723 gnd node is moving up or down, even a few millivolts gets gained up at the output.

    You should be able to find the place where the V drop is happening.

    If not, it's possible the design just doesn't have enough DC gain. Load regulation is basically a function of open loop gain for the control loop.
     
    Dave_UYZ likes this.
  19. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    Bounty: The 723 lacks a sharp cutoff for current limiting, but it can be added with a 358, just as you did in your discrete design. The foldback feature of the 723 is very convenient in my opinion. In addition, the reference voltage stability can be enhanced by placing the chip in a zener regulated supply of about 12V to 15V, with or without a separate secondary winding. The op-amp inputs can be impedance matched to minimize thermal drift.

    Including Scott: The first feedback resistor can be bypassed with a capacitor to minimize power supply ripple at the output. You can add a resistor in the range of 3.3M to 15M from the base of the output transistor back to the positive input of the difference amplifier. The increased Vbe under load will compensate for load droop. This method requires that you divide the reference voltage with resistors so there is an impedance to work with instead of trying to skew the reference voltage. Results of .01%, zero to full load can sometimes be achieved.

    Been there, done that, almost broke my arm patting myself on the back.
     
    Dave_UYZ and tvtech like this.
  20. #12

    Expert

    Nov 30, 2010
    16,248
    6,745
    I don't think loop gain is the problem because the OP design requires 3 ma drive current from the 723. (It is allegedly capable of 150 ma.) The usual method is to use a driver for a bunch of parallel output transistors, but this design uses Darlingtons which should accommodate the load nicely.

    Still, we'll see when we get some measurements.
     
Loading...