SMPS Tracking pre-regulator for bench PSU

Discussion in 'The Projects Forum' started by afternath, May 7, 2012.

  1. afternath

    Thread Starter New Member

    May 7, 2012
    17
    0
    Hi All,

    My on going project is to build a bench power supply with these specs :
    - Two CV Outputs
    - Adjustable Output voltage from 0V to 24V DC (fine+coarse)
    - Rated Output current of 1 A

    The circuit is currently composed of :
    -22V/100VA double secondary transformer (2.27A each secondary)
    For one channel :
    -one LM338K in TO-3 (improved LM317) for current limiting
    in serie with
    -one LM338K in TO-3 (with heat sink) for voltage adjust

    The problem that i have is, of course, heat dissipating...
    For example : output 12V @ 1A, the heat dissipated by the last LM338K is
    P=(30V-12V)x1A=18W
    30V corresponds to 22V*sqr(2) - 1V (diodes bridge drop)

    Junction temp is given by
    Tj=(rthjb+rthra)*p+ta=(1+10)*18+25=223°C > 125°C -> the IC will thermal shutdown
    where rthjb is the thermal resistance of the LM338K
    and rthra thermal resistance of the heat sink (no air flow)

    So what i want to do is keeping constant dropout voltage across the linear regulator in order to remove the thermal problem.

    To do this, the best solution is to use a tracking pre-regulator which keep constant voltage between its output (connect to linear regulator input) and the final output.

    I bought recently a SMPS adjustable step-down module based on LM2596S, see link below
    2pcs LM2596 DC-DC Step Down Adjustable Power Supply Module | eBay

    This IC (and others) has a FEEDBACK pin which is used to program output voltage with a resistor divider between output and ground.

    Now the final question is : what is the best schematic you can show me to track final output voltage of the linear regulator with this feedback pin?
    Maybe OPAMP circuit...
    (you can show me another kind of step-down IC)

    (see attached my current design + link to LM2596 datasheet)
    http://www.ti.com/lit/ds/symlink/lm2596.pdf

    Thanks in advance ;)
     
  2. afternath

    Thread Starter New Member

    May 7, 2012
    17
    0
    I found 2 possibles solution, that i need to test:

    1/ Complex one: create a feedback loop from the final output (linear reg output), this loop contains a non-inverting summing amplifier which add the desired drop out voltage (5V for example) and return it to the feedback pin of the LM2596 smps.

    2/ Simple one: modify the existing resistor feedback loop of the LM2596 by connecting R1 to the output of the smps instead of GND and connecting R2 to the output of the linear regulator
     
  3. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    I built one of these a long time ago. I recall the best way was to build a linear regulator designed for about 3V and then reference it to the regulated output. The variable switcher feeds te voltage to the linear, so the switcher voltage output is always 3V higher than the regulated output. You adjust the switchers voltage and the linear (which feeds the regulated output) follows along since it is always 3V below.

    This is the best way to get a decent bench PS with lower internal power dissipation.
     
  4. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    This design is easily doable in linears: you only have about 50W of total power with both sides shorted and current at 1A. Easily manageable with a decent heatsink. My bench PS is a dual 16V/3A unit so it cranks a bit over 100W with both outputs shorted at 3A. It's doable with a good heatsink. The one I have is about 0.7C/W.

    Building in a switcher is a cool idea, but it's a hassle and you will have switcher noise coupling out the leads. IMHO, I would use a linear for a 50W bench supply because the size and weight is not that bad. Above 100W, probably need a switcher.
     
  5. afternath

    Thread Starter New Member

    May 7, 2012
    17
    0
    Thanks for your advices
    The HF noise will be mostly reduced by the linear regulator if care is taken for ground path.
    What is the spec of your 0.7C/W heatsink (size, weight) and where did you buy it?
    Making a PSU only linear will waste more and more power with increasing output voltage range vs rated current.
    The heatsink helps do dissipate heat in a certain amount, but it is very huge and hard to implement in a case.
     
  6. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    If you want quality just go linear (as Bountyhunter said) and eliminate all the noise and possible instability. 24W is quite low in a bench supply and won;t be needed all the time anyway (only at 1A max). Even those little square heatsinks they use on Pentium chips with a tiny fan will easily move 50W.

    Also if you don't mind voltage ranges you can use a transformer with multiple output taps, I have a linear PSU with a rotary knob that selects different voltage taps and it runs quite efficiently.
     
  7. afternath

    Thread Starter New Member

    May 7, 2012
    17
    0
    Ok for linear i will probably keep my design, but i would like to know where to buy efficient TO-3 heatsinks? If you know a good website please let me know.

    One question about the transformer:
    I already bought a 22V/100VA double secondary transformer (2.27A each secondary)
    do you know if it will supply correctly (without heating or dropping the voltage) my psu rated at 1A DC in all the voltage range?

    Regards
     
  8. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    No, believe me, I spent may years designing them and EMI containment is very difficult. It radiates directly into any available conductor and magnetic flux goes through just about anything. It is not easy to design it right.

    It's an extrusion type from wakefield or AAVID. About 2" high maybe ten inches wide, fins about 3/4" length. I'll try to find a picture.

    Typical extrusions:

    http://www.aavid.com/cgi-bin/proces...irUnits=LFM&Shape1=ON&SortBy=V2&NumPerPage=v3



    That's true, it's a tradeoff for reduced noise, simpler design.

    Not really, if it's a bench supply. Mine is reasonable size and weight.
     
  9. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    TO-3 is a massive and practically obsolete transistor package meant to dissipate 150W to 200W. You are dissipating less than 24W at all times, and likely less than 10W in most circumstances.

    I would look at making your PSU using the LM317 regulator IC, it is easily good for your 24W and has overcurrent regulation and thermal cutout built in.

    For heatsinks pretty much anything larger than your hand and preferably with one of those little 40mm fans attached. All hobby elec shops have many different heatsink types, they are not that expensive.

    A 100W transformer that can deliver a total of 22v 4.5A is massive overkill for your PSU. Maybe you could look at the big brother of the LM317 which is good for 3A and make a 0-30v 3A supply?
     
  10. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    If you want 1A load current, that transformer is about perfect. Transformer secondary current is about 1.8X load current using full wave bridge, so you need about 2A ballpark (minimum) so 2.3A is perfect. Always good to derate it a bit because it will run cooler.

    The 22V rating cuts it close for a 24V output: you need enough DC voltage to allow a few Volts of ripple on the filter cap and a couple of Volts for the regulator headroom.

    I recommend the following design since it is a very good lab power supply design and only needs about 1V of drop across it to regulate.
     
    Last edited: Jul 4, 2012
  11. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    Bountyhunter, all the dual-secondary toroids I have used are label rated for full load output current, so 22v 2.3A will produce 22v DC at 2.3A load current (and then x2 as there are two output windings).

    At 1A output current the DC voltage after rectification will be around 35v, maybe >40v if he uses both windings in parallel (0.5A per winding).
     
  12. afternath

    Thread Starter New Member

    May 7, 2012
    17
    0
    Thanks for your feedbacks guys.
    I'm going to change my transformer with a dual 15V one, with same power, because i will design two outputs, voltage range can be reduced and if i need more than 15V i can use outputs in serie.
    I will have less power to dissipate for sure.
    I would like to use the LM317K (TO-3 package) to benefit of the 1.5A current limit, thanks to this feature, my first regulator configured as a current limiter will be removed. So i will have 2 regs in less => 2 heat sinks less, good point.

    The only problem is the LM317K has a higher RTHJC, (worst case 3C/W vs 1C/W for the LM338K).
    If i use a 1C/W heat sink, the junction temp calculation is :
    @ 1.5A, max drop out of 20V = 30W to dissipate
    30*(3+1)+25=145C... => IC will thermal shutdown
    I could use thermal grease to improve the heat transfert, the only way to be sure of the good operation of the PSU in all cases is to test it.
    The AAVID heat sinks you shown me are really cool but not appropriated for TO-3 because there is no space to fix it with screws...

    I have found a good heat sink on farnell:
    http://fr.farnell.com/abl-heatsinks/520ab1250mb-to-3x2/dissipateur-thermique/dp/253730

    What do you think guys about that, may i change my design for LM317K?
     
  13. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    The Theta JC for an LM317 in the TO-3 package should NEVER be 3C/W, I don't know why it would be that high.
     
  14. afternath

    Thread Starter New Member

    May 7, 2012
    17
    0
  15. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    But this data sheet says 2C/W.

    http://www.ti.com/lit/ds/symlink/lm117.pdf

    Believe me when I say: I spent 30 years working on power devices, the last 20 at National Semiconductor. Any semi company who is getting 3C/W for a TO-3 device must have screwed the package up completely. The typical values should be closer to 1.5C/W.


    Funny that this power transistor in a TO-3 has a thermal resistance under 1C/W. Larger die, but still shows the package is capable of very good thermals. Have no idea how they screwed up the 317 to make it that bad.

    http://www.onsemi.com/pub_link/Collateral/2N5883-D.PDF
     
    Last edited: Jul 5, 2012
  16. afternath

    Thread Starter New Member

    May 7, 2012
    17
    0
    Allright no problem, it will be more often close to the typical of 2C/W, i've just taken the worst case of 3C/W.

    The topic is moving to a thermal problem instead of SMPS, i will open a new one, with my last though about this PSU.
     
  17. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    The only reason you have a thermal problem is you are making a mistake in using the LM317. I posted a schematic for a very good linear regulator that uses an external power transistor (which would have the lower 0.9C/W resistance) and give superior performance. The LM317 is unsuitable for any kind of usable bench supply for a variety of reasons, one large one being you won't be able to set the current at any desired value. It also won't adjust to zero Volts. If you are going to spend the time building a supply, build a better one.
     
  18. richard.cs

    Member

    Mar 3, 2012
    162
    31
    I won't comment on the linear section but I can tell you that there used to be a commercial PSU that reduced dissipation on the linear reg by triac-control of the transformer primary. A little noisy but less so than a switcher. We've still got quite a few of these at work and they're perfectly good for normal lab use.

    If you go that route you might have to oversize the transformer as the high peak currents will increase the RMS dissipation, and probably add some series L to the rectifier output to reduce the peak ripple on the filter capacitors.

    I believe you should be able to filter a switcher adequately though, and you could potentially get away with stepped pre-regulation for simplicity (pre-reg to 10V for output <8V, pre-reg to 20V for output <18V, disable switcher above that).
     
  19. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    I'm not the OP, and I understand you took the time to post a schematic for a bench supply that is superior to a LM317 PSU in a number of ways, but I think you are being unfair with the above statement.

    An LM317 based PSU is an excellent starting point for a person building their first bench supply, due to the extreme simplicity of needing only one 3pin semiconductor and having all the inbuilt automatic protection etc. It has quite a bit of usefulness, and a very high likelyhood of success for a first time PSU builder.

    The issue with having only one current limit (due to the internal protection) can also be dealt with by using a second LM317 as a current regulator, before the LM317 voltage regulator.

    I have a few bench supplies but the one I seem to be reaching for the most these days is a dual LM317 setup that I whipped up in a hurry one day.
     
  20. bountyhunter

    Well-Known Member

    Sep 7, 2009
    2,498
    507
    I'm not being unfair, just pointing out what I learned the hard way:

    Building a bench supply is a TON of work, building the chassis, fitting the meters, rectifiers, heat sink assemblies, etc.

    And the first one I built DID USE the LM317 and it was a short time before I realized how useless it was and upgraded it.... which was a lot more work than just building it right the first time.

    I offer the design to anybody who wants to take advantage of it and step onto the learning curve about ten feet above the 317 level.

    Or, plow ahead and re invent the wheel is OK too.

    Attached is a simplified version of the schematic.
     
    Last edited: Jul 9, 2012
Loading...