Fully Discrete Switched PS

Greetings everyone,

In http://www.diyaudio.com/forums/showthread.php?s=&threadid=57281, i was looking for suggestions on an amplifier to be built. I ended up going for a modified A/B Class Citation 12 schematic (http://passdiy.com/pdf/citation.pdf).

Well, the original supply voltage is +-42V, but I have decided to go for 30-35V, saving me some money on the transfo :dodgy:. The "problem" is: i really don't want to use the originally proposed power supply (which is not regulated, using only filter capacitors).

When simulating, I realized that an ordinary linear regulator wouldn't handle it, as the circuit demands ~4A :bigeyes: on each rail when the output transistors are conducting. Power dissipation on the series pass transistor would be huge, and the ripple voltage is a little too high for me. BTW, I simulated using a 36Vrms transformer, using 6800uF filter capacitors.

So I wonder: what if I use a fully discrete (no specialized ICs) switched regulator in this case? Why discrete? Mostly for fun :D ... I am aware on the restrictions (/www.diyaudio.com/forums/showthread.php?s=&threadid=28028) on the discussion of mains operated SMPS :att'n:, but i would like to know: how feasible is the use of a switched regulator to supply 4-5A @ 35V out of a transfo+rectifier+filter combo...

As I can't afford another transfo, i am considering only transformer-less topologies, specially boost/buck... AFAIK, i can build the pwm controller using opamps + capacitors + transistors, but I haven't done it yet (in PSPICE; I may give it a try on this weekend)...

In case I haven't been clear, I am planning on using a sec center-tapped transformer (36Vrms), plus a bridge rectifier, and two filter capacitors before the switched regulator... The input dc voltage would be between 40 and 50 V (an estimate), for a 35V regulated output voltage. By using the boost circuit, I (if I'm not wrong) could go for a cheaper transformer (around 24-25Vrms)...

What do you guys think?
 

Dem

Member
2005-05-07 12:32 pm
Israel
When I was young, I built a lot of idiotic things, one of them - relay-type buck regilator 30V to 24V 5A, The main idea - a comparator with small hysteresis, when output voltage drops a little - it opens the switch, when grows to much - closes it. It even worked :) ... (see picture) . You can adjust Vout vith R5, and output ripple with R3, of course You can't us LM393 when Vin more than 30V, just build another "full-discrete" comparator...

About boost - I don't think that it's a good idea, and transformer will not cheaper - it will lower voltage, but higher current, so it will have the same power, and tr-rs with same power cost the same...
 

Attachments

  • relay_stab.gif
    relay_stab.gif
    5.6 KB · Views: 978
Kenshin said:
transformer-less: you can't! it's too dangerous.

Is it dangerous due to the non-isolation from the mains line? Or something else? So, in this case, I have to go for a topology using a transformer... Any recomendations? Flyback, Half-bridge?

BTW, Has anyone successfully used transfos from "old" PC power supplies (i'm thinking about ~200W AT ones, which should be easy to find in old PCs, for a good price) ???
 
Mr Evil said:
If power dissipation is your main concern, then you might consider a capacitance multiplier instead of a regulator. It would allow you to achieve low ripple and low power dissipation, at the expense of regulation.

The thing that keeps on bothering me is: should I rely on the feedback to prevent hum due to supply voltage ripple??? If so, then I guess I could go for a capacitante multiplier...
 
In the experiments I have done with capacitance multipliers, it's not hard to get ripple down to 4mV with 4A output current. Even if the amp has zero PSRR it's getting on for 80dB down, which isn't bad, plus it has a relatively benign, almost sinusoidal shape, as opposed to the higher frequencies from linear of switching regulators. Even lower ripple is quite achievable.
 
You can certainly make a discrete SMPS, but why bother when chips are so much better? LOL

You can try anything from a variable-bias multivibrator to a full-blown expanded SG3524 or TL494 internal diagram, using darlington long tailed current-mirrored differential amps for the comparators, a current mirror to charge the timing cap and an open collector to discharge it, to form the triangle wave.

The former would look something like this:
http://webpages.charter.net/dawill/tmoranwms/Circuits/More_SMPS_Junk.gif
While the latter would look like this, 50 years ago:
http://webpages.charter.net/dawill/Images/Tube PWM Generator.gif

Tim
 

Kenshin

Member
2005-01-09 9:51 am
China
Not idiotic things, but fairly good design.

I have also done something like this, then develop the thing into some class D audio amplifier.

A year ago I started a project of building a 24V to 5V 7.5A supply for a industrial PC on a robot. SSEEKER, my companion, suggest to use 3x LM2576 in parallel, and I think it's not reliable. So we decided to build one use discrete power stage.

Soon, I sliped into designing some advanced thing operating at 200KHz. Also, I considered too much problems such as speed of comparator or overshooting of loop responce. So I didn't really made anything until end of the robocon2004.

Then we, the whole team lost the game in Beijing -- since we spend too much time in PSU instead of sensor.

This year, for robocon2005 (my schoolmates of lower grade do it, :D ) I built one with IR2111 + IRF540NS power stage, it works at 100KHz. IR2111 has too much dead time and the MOSFETs gets overheated due to body diode reverse recovery (IMHO.) when the load is heavy.

I got angry of it. Instead of operating it at lower frequency, I give up the whole project and threw it to SSEEKER (companion of last year :D )

He build it with three LM2576 in a week. Now, it runs perfect on our young school brother 's robot.

Class D audio ampilfier technogy may be the biggest good thing left in this MEGA failure. I have bulit some class D amplifiers of 10w power level, it sounds fairly well.

Dem said:
When I was young, I built a lot of idiotic things, one of them - relay-type buck regilator 30V to 24V 5A, The main idea - a comparator with small hysteresis, when output voltage drops a little - it opens the switch, when grows to much - closes it. It even worked :) ... (see picture) . You can adjust Vout vith R5, and output ripple with R3, of course You can't us LM393 when Vin more than 30V, just build another "full-discrete" comparator...
 

Kenshin

Member
2005-01-09 9:51 am
China
Of course, extremely dangerous.
How can you use an audio equipment without mains isolation? If you touch the audio input/output terminals, you will die.

For a isolated one -- the topic "Self-oscillating SMPS with saturable drive transformer" is discussing a simple self-oscillating one.
http://www.diyaudio.com/forums/showthread.php?s=&postid=652815#post652815

tguzella^TuX64 said:

Is it dangerous due to the non-isolation from the mains line? Or something else? So, in this case, I have to go for a topology using a transformer... Any recomendations? Flyback, Half-bridge?

BTW, Has anyone successfully used transfos from "old" PC power supplies (i'm thinking about ~200W AT ones, which should be easy to find in old PCs, for a good price) ???
 
Solid State DC-DC conversion

Well, if you plan on using a transformer for mains isolation, I think the safety subject is already solved.

But for control issues, I believe it is best to work on a buck topology rather than a boost converter. My experience is that if you apply conductance control, buck is best to regulate. You will have to overcome some design limitations such as (dis)continuous mode operation.

My biggest experience so far was a fully discrete boost-topology regulator for an interface with Solar Arrays. A 400W 100V output was realised with a single opamp, two comparators and a few transistors.

So buck/boost self-oscillating discrete DC-DC PSU? very feasible...

Bouke
 
How can you use an audio equipment without mains isolation? If you touch the audio input/output terminals, you will die.

That seems like somewhat of an exaggeration! :D Wouldn't this be ok if it was operated with a GFCI cable or outlet? I'm considering my first power amp project to be a high power (~1.5kw) Class D subwoofer amp that could shake my whole home theater like the woofers do at the cinema. The cheapest/lightest way to do it would be to rectify the mains directly to 170 Vdc then use an H-bridge mosfet switching configuration like you'd see with a hip4080. This is way beyond that chip's voltage rating so either discrete components would have to be used or a pair of half-bridge drivers with something driving them. I really like this idea because it should be very efficient due to the relatively low output current required to make this power at such a high voltage.

Besides instant death, is there a really good reason not to rectify the mains directly? :scratch:
 

mzzj

Member
2005-07-02 8:17 pm
65N 25E
rth said:


That seems like somewhat of an exaggeration! :D Wouldn't this be ok if it was operated with a GFCI cable or outlet? I'm considering my first power amp project to be a high power (~1.5kw) Class D subwoofer amp that could shake my whole home theater like the woofers do at the cinema. The cheapest/lightest way to do it would be to rectify the mains directly to 170 Vdc then use an H-bridge mosfet switching configuration like you'd see with a hip4080. This is way beyond that chip's voltage rating so either discrete components would have to be used or a pair of half-bridge drivers with something driving them. I really like this idea because it should be very efficient due to the relatively low output current required to make this power at such a high voltage.

Besides instant death, is there a really good reason not to rectify the mains directly? :scratch:
I am around same ballpark, exept planning to run it from rectified 230V mains. Lundahl seemed to have 4kV ac rms isolation specified input transformers wich is requirement in here, but maybe i wound my own prototype since in subwoofer application its far more easier than full-range input transformers. I have fluke insulation resistance tester availlable at work(5kv dc), but usually i use variac and oil burner ignition transformer at home, up to 10kV ac rms and measure leakage current with ordinary multimeter standing on top of plastic box. Speaker terminals are big NO-NO at these voltages, and actually normal loudspeakers are also big NO-NO. not any speaker manufacturer can quarantee that their cone is not going to flash trough at 4kV, so i had to build that amplifier directly inside sub box, ground element chassis and fix fine-mesh grounded metal net directly in front of entire speaker element. And i am probably still breaking at least 10 different safety regulations. GFCI is definetely good idea, but even that is useless if useless user is putting his fingers directly to both of output terminals at max volume.

Running it with isolation transformer makes things somewhat easier and thats my option 2, since i have few pieces of 3000VA 230V/230V toroids hanging around.
 
What exactly is using 4kV? :bigeyes:

60hz transformers are big, heavy, waste power, and cost money ;) I'll probably sample the power devices and I have all the small parts so I think the only thing I'll have to actually pay for is the heatsinks :cool:

Even if I decide that isolation is required, I'd prefer to do a switchmode config that chops up the rectified mains then runs through a small high freq transformer... way smaller, very light, and they are cheap to wind :nod:
 

mzzj

Member
2005-07-02 8:17 pm
65N 25E
rth said:
What exactly is using 4kV? :bigeyes:

60hz transformers are big, heavy, waste power, and cost money ;) I'll probably sample the power devices and I have all the small parts so I think the only thing I'll have to actually pay for is the heatsinks :cool:

Even if I decide that isolation is required, I'd prefer to do a switchmode config that chops up the rectified mains then runs through a small high freq transformer... way smaller, very light, and they are cheap to wind :nod:

In here safety regulations call for 4kV dielectric strength between ground and L or N conductor. Means that it has to be able to withstand 4kV rms AC for 1 min without arcing or starting to leak. Bit less in some cases if i remember correctly but i usually built everything to withstand 4kV so i dont need remember every exeption of regulations. And 6mm creepage distance is required for uninsulated parts. This can be a real bugger in SMPS transformers when half of core widht is lost to creepage distances.

In US somewhat less than 4kV is sufficient but its not my business what UL or whatever says there.
 
You can easily build a high quality linear regulator which will do more than 4 amps -- use the LM317HV and wrap an NPN power transistor around it. National Semi will tell you how in the application notes for the LM317HV -- although they show use of their super-matched pair transistors. Using the LM317HV will definately be quieter and easier to build than an SMPS.

You can also scale the Pooge 5.5.1 for higher current, higher voltage...