12v dc to +\-35v dc for amplifier

Discussion in 'The Projects Forum' started by MattStrike, Jun 12, 2010.

  1. MattStrike

    Thread Starter New Member

    Jun 12, 2010
    I have built an amplifier based on the LM3886 reference design http://www.national.com/ds/LM/LM3886.pdf

    The amplifier requires a +v and a -v supply between 20v and 84v (35 seems good). I want to convert this amplifier to run from 12v dc car power. The only thing I've found so far is the LM2587-ADJ reference designs. I don't see a 68w-135w amplifier being able to operate at about 1 amp, which if I understand correctly is the theoretical limit that the LM2587 can output given my input voltage and output voltage.

    So, I need to build a power supply that can supply enough wattage to operate the amp. I understand I can use a switch-mode power supply for this? Are there any other ways? I would prefer a cheaper, compact or simpler solution.

    So far I've found this write-up about pretty much exactly what I need:

    If no other options exist than the smp, I need to learn as much about them as possible before I try to build one. My understanding is that the 12v input is put into a square wave generator to convert the 12v dc to ac? The AC is then stepped up through a center-tap transformer to the required voltages, rectified and filtered. What I don't understand is the purpose of the transistors and the gates and how that whole mess functions. Also, the schematic I found is a bit odd to me referring to the multiple +12v tabs...usually that is due to my lack of understanding of proper schematics...

    I am wanting to put the amplifier right next to the speaker in the door of the car because I don't want an amp in the trunk and run long wires through the car to pickup noise. Also, a single channel amp is just as expensive as a dual or some quad channel car amps, and they are bulky and have way more power than I need to drive the speakers (~50watt RMS, 180w max) and the head unit can't drive them at a sufficient volume, and I build circuits as a stress-reducing hobby (I'm an ME). Worst case is a dual channel amp under the dash, but that is a last resort.

    I also want to take the lesson learned to built a 12v to 19.5v 2.4a laptop car supply in the future.

    Any help is appreciated, suggestions welcome.
    BEJOYJG likes this.
  2. SgtWookie


    Jul 17, 2007
    Hello Matt,
    Welcome to AAC.
    SMPS's are tough to get a good handle on.

    It sounds like you want to start with the project on Rod Elliott's site, but that's kind of like climbing a mountain while you're learning how to walk - it might not end well.

    I suggest starting with some easier projects, working up to your laptop car supply, and then finally tackling the Elliott Sound SMPS. By starting off with the easier projects, you will have a far higher probability of success, and each project will build on the knowledge you've gained.

    Here's a great site for switching supply n00bs: http://www.dos4ever.com/flyback/flyback.html

    Ronald Dekker's done a nice job keeping things pretty easy to understand for lay persons.
    To keep the first supply output to a lower (and more safe) voltage, change R4 to 47k Ohms. I like to keep projects for electronic newbies below 50v; above that you start running into shock hazards. Keeping it below 50v means that the neon bulb at the output won't light; they take about 65v to ionize initially.

    Build Ronald's "Inductor Test Bench" and experiment with toroids and transformers.

    I hope that you have a decent oscilloscope. I picked up a used 60MHz bandwidth dual trace O-scope for next to nothing on an auction site. It doesn't have to be fancy or cost a lot of money to be very useful.

    In the flyback converter schematic near the bottom of Ronald's page, use a 1:3 ratio instead of a 1:10 ratio for the transformer, and again use a 47k or lower for R4.

    After you've completed Ronald's projects, you'll have some introductory knowledge and some practical experience that will be very helpful towards larger projects.

    Here is a very useful site with loads of information:

    SMPS's are the most efficient and cost-effective way to produce the power output at the voltage levels you need in an automotive environment. They have a pretty steep learning curve, particularly if you are a newcomer to electronics.

    That's the general idea. On the input side, the MOSFETs are switched at high frequency to sink current out of either end of the transformer primary. The transformer can be small due to the high frequency. If you were to use a transformer with 60Hz on the input, it would have to be much larger and heavier to provide the same power out. The IC that Rod & co. used switches in the 150kHz region, so that the transformer can be considerably smaller.

    Q1 through Q4 are gate driver transistors for their respective MOSFETs. MOSFETs are generally used as electronic "switches". If the Vgs (voltage on the gate terminal, referenced from the source terminal) is equal to 0v, the drain terminal to source terminal is essentially an open circuit (infinite resistance, no connection, no current flow). If the Vgs = 10v, the Rds(on) (Resistance from the drain to the source terminals) is very low.

    The gate of a MOSFET is more or less like a small capacitor. In the datasheets for MOSFETs, the gate charge is specified in nC's, or nano Coulombs. http://en.wikipedia.org/wiki/Coulomb
    Once a MOSFET is turned ON or OFF by charging or discharging the gate, it takes practically no current on the gate to maintain it's state. However, in order to turn it on or off quickly (thus spending as little time in the partially-conducting state where power is dissipated as heat), one needs to provide for rapid charging and discharging of the gate. That's the point of Q1 thru Q4.

    However, charging/discharging the gate too quickly will result in "ringing" on the gate. The wiring/traces from the gate to the transistor driver has parasitic inductance, and the gate has capacitance, so the two make up a series LC circuit. The small values of R that you see between the transistors and the MOSFET gates serves as a damper, to minimize the tendency of the gates to oscillate after the state has been toggled from ON to OFF or vice versa.

    You'll find that node names (like +12, Vcc, Vdd, GND, etc.) are used very frequently on schematics. This helps to eliminate visual clutter. All points with the same node label should be considered as electrically connected together.
  3. MattStrike

    Thread Starter New Member

    Jun 12, 2010
    Thank you for the links and information!

    I had planned on buying a portable oscilloscope (MCP CQ5002 or DSO Nano) for troubleshooting a 1986 GM ecm/sensor problem (can't get a new car until I fix this one, personal vandetta, i have to win), but I get the feeling that the sample speeds and single channel aren't enough for this project.

    I would like to stay portable if possible so I can use it to monitor the car during driving conditions as well. I have a laptop that I use to monitor the ALDL data (160 baud is frustrating) from the car, so a USB oscilloscope could work.
  4. SgtWookie


    Jul 17, 2007
    As designed, Rod Elliott's DC-DC converter's oscillator frequency is roughly 65kHz.

    In order to be able to view a decent representation of the oscillator output, you'll need a bandwidth of at least 10x the fundamental frequency, or 650kHz. The higher the bandwidth of the scope, the better the representation you'll see.

    This is due to the harmonic content of ideal square waves. An ideal square wave is composed of the fundamental frequency, plus ALL of the odd harmonics of the fundamental.

    Have a look at this Wikipedia entry: http://en.wikipedia.org/wiki/Square_wave
    Scroll down to the 2nd image, and click on it, or just click here:
    You'll see that as the harmonic content increases, the shape of the square wave improves.

    This implies that transmission of an ideal square wave requires unlimited bandwidth. Since unlimited bandwidth is not practical, you go for as much bandwidth as you can reasonably afford, and understand the limitations of your device to display the waveforms.

    I have an "old school" affinity for the older analog scopes. They are bulky, but you can actually repair them. The new digital scopes are great, but if they break or need calibration, good luck.

    Meanwhile, here are some interesting finds for you.
    Electronic Goldmine sells lots of stuff (usually surplus) to hobbyists. They sometimes have some pretty good deals, if you know (or have a pretty good idea) of what you're getting.

    Here's a 5-pack of toroids for $1: http://www.goldmine-elec-products.com/prodinfo.asp?number=G6683
    If nothing else, they would be very good for use as baluns to stop electrical noise from entering/exiting your various projects.
    eta: see this Wiki entry on baluns: http://en.wikipedia.org/wiki/Balun - note the large image on the left halfway down the page.
    They could also serve in your experimentations on various inductors/transformers.

    Here is a random selection of toroids, cheap: http://www.goldmine-elec-products.com/prodinfo.asp?number=GP62
    You won't know exactly what you are getting, but you can figure it out using an O-scope and the inductor test bench on Ronald's page. You will learn a lot from the experiment.

    Here are toroids that will work very well for your eventual project:
    I looked them up. Here are the specifications:
    I suggest picking up a couple of them.

    Note that Electronic Goldmine has a minimum purchase of $10. Also, your shipping costs will likely not be less than $9, so you might as well get your shipping dollars' worth.

    You might pick up a few of these MOSFETs: http://www.goldmine-elec-products.com/prodinfo.asp?number=A20388
    Not appropriate for your final project, but plenty good enough for a variety of experiments. Selection of MOSFETs for the application will require a bit of thought.

    Here is a hard-to-find RF transistor that will come in handy:
    Use it for this project:
    A simple crystal tester.
    Last edited: Jun 14, 2010
  5. SgtWookie


    Jul 17, 2007
  6. SgtWookie


    Jul 17, 2007
    My order from Electronic Goldmine arrived a bit ago.
    The 77443-A7 toroids appear to be in unblemished condition. I have not had time to perform any testing on them at this moment.

    I have a couple of SG3524's but no 3535's as of yet.

    Have you made progress in reading those links I posted?
  7. MattStrike

    Thread Starter New Member

    Jun 12, 2010
    I have read and bookmarked the links and I understand what is going on, but have to buy new tires so can't get the oscilloscope this paycheck. I will be able to order some of the stuff I need for the projects. Where can I find some cheap coated wire to wrap the toroids in? I have some dead pc motherboards and psu's and and old xbox 360 board that I'm going to scavenge for some toriods and caps as well. I haven't had time to play with the toriod calculator program yet. Can it calculate the toriod value if I can measure the toriod it with a calipers?

    Also, how can I calculate how much current a given wire can handle? I can measure the wire guage with my calipers.
  8. SgtWookie


    Jul 17, 2007
    You have to use magnet wire. Magnet wire has a thin insulation coating on it, sometimes it's lacquer, sometimes it's a synthetic coating. You can't use wire that has opaque plastic insulation.

    Sometimes you can salvage it from solenoids, old-fashioned doorbells, motors, transformers, switching power supplies, etc. if you're careful and don't nick the insulation or kink the wire. You're actually better off to use multiple smaller gauge wires than one large gauge wire due to the skin effect.

    Radio Shack carries a 3-pack of magnet wire for a few bucks.
    I've bought magnet wire from auction sites.
    I have a 10-lb spool of AWG36 magnet wire that must be 40 or more years old.
    There is a store near me called Skycraft Parts and Surplus that carries all kinds of magnet wire for decent prices, but they don't sell that via their website.

    Since I have absolutely no clue where in the world you are, it's kind of hard for me to suggest where else you might find it.

    Here is yet MORE reading on magnetic core materials: http://focus.ti.com/lit/ml/slup124/slup124.pdf

    Unless you know what the material of the toroid is, you can't guess what it's characteristics are until you start experimenting with it by winding on a thin layer of tape (to delay the onset of saturation and reduce eddy currents) and then wind on some wire and start doing some tests, like on Ronald Dekker's site.

    Speaking of which, I just ran some tests on that 5-pack of toroids I received from Electronic Goldmine: G6683
    The Al value is around 10,000, which is quite high. 10 turns gave me 936uH; 20 turns were 4mH. They would be good for low-frequency stuff, like power line chokes.

    More reading for you attached.
    Last edited: Jun 24, 2010
  9. SgtWookie


    Jul 17, 2007
    I picked up some UC3825BN's at a local supplier. They can run a good bit faster than the SG3525's, have more gate drive (1.5A vs 500mA) and the rest of the features are similar.

    Product folder:http://focus.ti.com/docs/prod/folders/print/uc3825.html
    Datasheet is available for download here: http://www.ti.com/lit/gpn/uc3825

    Scroll down to the Application Notes. The first one, entitled "Current Doubler Rectifier Offers Ripple Current Cancellation" is quite interesting, because the MOSFETs in the output are used as synchronous switches, or the elusive "perfect" diodes (refer to schematic on page 2). This makes the supply more efficient (reduced power loss in the output stage) which means reduced power dissipation. You'd save roughly 12W of power loss, figuring a 1v drop across rectifier diodes with 6A current; one diode active at a time in both the + and - supplies. Your noise and ripple current would also be reduced.

    The secondary would obviously need either two windings, or a center-tapped winding in order to produce your + and - outputs.
  10. MattStrike

    Thread Starter New Member

    Jun 12, 2010
    This may seem like a silly question, but if I'm wrapping my own toriod how long should I limit each wire to? Theoretically I should be able to obtain any ratio of turns that I needed by wrapping each wire more times, but what are the limits when dealing with the high-frequencies? I found some very fine magnet wire locally that's smaller than 30 guage by a lot, as well as some 30 guage wire. I figured I would experiment by wrapping an audio step-up transformer and a 1:1 isolation transformer.

    Also, I have salvaged a few transformers from some pc power supplies. They have the "E" shaped core and figured the core was perfect for a switch mode power supply because that's what they were originally used in. These transformers have a series of numbers printed on the top, in a "xxxx.xxxxxx.xxxx" format. Is there a way to determine the core material from this number?

    Edit: I found a breadboard workstation with a +/- 1.5v-16v dc supply and a square and sine wave generator for small projects :)
    Last edited: Jun 26, 2010
  11. SgtWookie


    Jul 17, 2007
    That question is not silly at all, nor is it easy to answer!

    That "Mini Core Ring Calculator", in conjunction with measurements or specifications for your toroid, will help a great deal to determine resonant frequencies and the impedance of the windings at particular frequencies - but you need to know the characteristics of the toroid to begin with. Inductance is just one part of the equation.

    It's more complicated than that. You're going to be driving the primary using a square wave, somewhere in the vicinity of 20kHz-50kHz. Remember earlier when I linked to the definition of square wave in Wikipedia? An ideal square wave is the fundamental frequency (ideal sine wave) plus ALL of the odd harmonics of the fundamental frequency. This implies that a very broadband transformer is required.

    If the impedance is high at the fundamental frequency, you won't get enough current through the primary winding to transfer the energy to the secondary. This implies that the primary must have a low impedance in order to transfer the energy required at the frequencies you will be operating the switching regulator at; which means few primary windings. If you re-visit Rod Elliott's page, you'll see that they are only using 4 or 5 turns on each side of the center tap of the primary.

    If you wish to use MORE turns, then you would need to use a toroidal, pot, or E-core with even lower Al than what he was using. I haven't managed to find the specs for the EL39 core they're using; it would have been very helpful if they had included a link to a datasheet or manufacturer.

    This is a broadband power transformer. If you try wrapping a gazillion turns around a toroid, it'll have a high impedance, implying a low fundamental frequency requirement.

    Note that a transformer designed for 50/60Hz is like a big chunk of steel made up of a series of E-cores (it's a special iron alloy) with LOTS of windings on the primary side.

    OK, the "E" shaped cores, if they are laminated iron/steel cores, are designed to be operated at low frequencies. If you tried to use the same cores with high frequencies, you would have very high losses in the cores.

    Now, there ARE E-cores and pot cores that are made from ferrite or powdered iron, which ARE designed for higher frequencies; they are generally molded in two parts. However, the part numbers on them may be proprietary. You would have to use something like Ronald Dekker's test bench to evaluate them - but that would be just the beginning.

    If you want to build a high-power DC-DC converter, you better stick with purchasing a toroid or E-core with known properties. Otherwise, you will likely wind up with a power supply that is woefully inadequate and/or woefully inefficient.

    Nice. :)

    Beware though; breadboards really are not good for much more than audio frequencies - there are too many parasitic properties for higher frequency and/or higher power projects. The jumper wires have inductance at a rate of about 15nH per 10mm, or about 450nH per foot. The parasitics will wreak havoc on broadband and/or high frequency circuits.
    Last edited: Jun 27, 2010
  12. SgtWookie


    Jul 17, 2007

    Have a look at the attached; it's a 12v to 19.5v DC-DC boost converter for operating your laptop in your vehicle.

    I'm also attaching the LTSpice .asc file. If you don't have the free LTSpice simulation tool, download and install it. Google "LTSpice download", it's on Linear Technology's website. You don't have to register if you don't want to.

    There is an LTSpice support group on Yahoo! groups. I suggest that you join it; lots of help and lots of models that drop right in to LTSpice.

    I neglected to select a model for C3, so keep that in mind. Use a cap rated for >60v.
    Last edited: Jun 30, 2010
  13. MattStrike

    Thread Starter New Member

    Jun 12, 2010
    I've seen that program before in a dynamic systems class!

    Your attached .asc was only 28bytes and was blank in the LTspice IV. I made the circuit in a new file though have a question: What is the load? The load options are described as a "Current Source", which makes me think of it as supplying the current rather than pulling current from the circuit. Also, how is the output voltage adjusted? I changed a few components and it didn't seem to work how I expected.

    The LT1170 seems to make my laptop power supply very easy now.

  14. SgtWookie


    Jul 17, 2007
    Very strange about the .asc being only 28 bytes long. I frankly can't explain why that happened. I just tried re-uploading it, and it's still 28 bytes.

    I started off using the test fixture circuit for the regulator, then saved it under a different name and modified it. It comes up just fine on my system.
    OK, I see what happened now; I'd saved it in two different directories, and uploaded the wrong one! dOH! (in my best Homer Simpson voice...) The correct .asc file has been uploaded. [/eta]

    I1 is the load. It's a current source used as a current sink. The model is named "current". Once you place the current model in the schematic, right-click to get to the settings, and then click the "Advanced" button. On the next dialog page that pops up, you'll see a box under "Parasitic properties" next to "This is an active load" - check the box. That causes the current source to only sink up to the specified current (2.7A) if there is a voltage of the correct polarity across the model - otherwise, the simulation would wind up with a large negative voltage on the output until the regulator could catch up.

    The regulator tries to maintain the voltage on the FB (feedback) terminal at 1.24v. If it's below that, the regulator will attempt to supply more current to the output.

    I used 2.25k for R1, and 33k for R2.
    1.24v/2.25k = 0.551mA
    19.5v/0.551mA = 35.383k Ohms total for R1 and R2.
    35.383k Ohms - 2.25k = 33.133k Ohms
    I rounded it down to 33k.

    Changing the value of R2 will change the output voltage; decreasing the resistance will decrease the output voltage.

    If you want your simulation to have the same efficiency as the simulation I made, use the same capacitors and inductor, both in the simulation AND in the project itself. If you use the default "ideal" capacitors, you will get an artificially inflated efficiency. If you use "just any old caps and coils", you will very likely degrade the efficiency and performance of the circuit. I specifically selected components with low parasitic properties to get decent efficiency from it. See the Bill of Materials at the end.

    You'll need to make a circuit board for it; the smaller the better. The switching frequency of the LT1170 in this circuit is 100kHz, but as it's a square wave output, you need plenty of bandwidth - which implies very short and very wide traces on the board.

    Note that the LT1170 will dissipate 2W of power. Use a heat sink to keep it cool. Even though it's pretty efficient, don't forget that it still needs to be kept cool. If you put it in a project box, make sure that one panel is metal and the 1170's tab is thermally connected to it.

    Code ( (Unknown Language)):
    2. --- Bill of Materials ---
    3. Ref.    Mfg.        Part No.    Description
    4. C1  KEMET       C1206C105K3PAC  capacitor, 1µF, 25V
    5. C2  Panasonic   ECA1VFQ102L capacitor, 1mF, 35V
    6. C3  KEMET       T510E107K025AS  capacitor, 100µF, 25V
    7. D1  --      B530C       diode
    8. L1  Coiltronics CTX100-5-52 inductor, 100µH, 5A pk
    9. R1  --      --      resistor, 2.25K
    10. R2  --      --      resistor, 33K
    11. R3  --      --      resistor, 1K
    12. U1  Linear Technology LT1170    integrated circuit
    Last edited: Jun 30, 2010
  15. SgtWookie


    Jul 17, 2007
    I mentioned short/wide traces on the board, but didn't mention any tools.

    Download PCBTemp, freeware, here:

    The data is somewhat old, but Ultracad has discontinued support for this freeware tool, and the replacement tool costs $200. It's far better than nothing.
  16. MattStrike

    Thread Starter New Member

    Jun 12, 2010
    My order from electronics goldmine came in so I can start some of these projects when I get back from vacation. I've also been trying to find ~2mm cylindrical dc motor brushes to fix the power door locks on my car but haven't had any luck yet locally, hoping to find something online.