I'm posting this here in hopes people more knowledgible (sp.. heh what a word to get wrong eh?) in SS will see it... anywho...
The goal is DC somwhere between 100 and 500V, 10 to 200mA. A bit more specific goal would be say, 200V at 20mA to run a tube preamp, and 300 to 400V at 100mA or so, for a nice 6V6 or 6L6-ish power amp.
Here's what I've breadboarded so far. Description:
A basic multivibrator for 'clock'. A variable voltage source (not really, but gets the job done) controls the bias to one side of the oscillator; the imbalance changes the duty cycle (more like 'on' time, since frequency is changed, but that doesn't matter), varying the charging time the inductor recieves, thus varying its stored energy. This control circuit will later apply DC-coupled NFB to regulate the output voltage.
In the other PWM control circuit, which I tried on a whim and worked first time(!!), a pair of transistors resting on a CCS form a differential pair. One is biased fixed, the other adjustable, to vary the ratio of output voltages (rather than just one), and thus give a more accurate change in duty cycle.
The signal is taken off and fed to a driver transistor, which then powers the output transistor. In one case, an NPN is driven from the osc's emitter (with a small resistor to ensure cutoff), which then drives a large plate load, err sorry, collector load, with the attached output transistor. In the other case, a PNP is driven from the osc's collector instead. I tried it both with and without current limiting and 'speedup' resistors (to swamp Miller effect I presume) and noticed little difference.
For output, a high voltage transistor clamps an inductor for that 'on' period of time. When it releases, the pent up energy explodes into a flyback pulse, which would be bad mojo if not for the damper diode (hey, it's even being used as it was made to be used!) which funnels the pulse into a filter cap.
The questions: how can I improve 1) power output and 2) efficiency? I already know the answer to #1, lower inductance or longer 'on' time, and generally beefier parts, but I haven't been able to get either an efficient driver circuit of hefty enough capacity, or good output transistors (I'll be stripping some more computer PSU boards later on) for either. As for #2, by the nature of the output transistor, the driver's load will always have +V-1V or so across it, which burns something like a watt. All the other transistors contribute maybe 1/2 or 1W as well, and the rectifier doesn't help any. All in all I get maybe 70% efficiency for 5W output (current record, 215V across 9k). I'll admit that's not bad for someone who doesn't know squat about SS, but this digital kind of circuit can push a lot more than that.
Tim
The goal is DC somwhere between 100 and 500V, 10 to 200mA. A bit more specific goal would be say, 200V at 20mA to run a tube preamp, and 300 to 400V at 100mA or so, for a nice 6V6 or 6L6-ish power amp.
Here's what I've breadboarded so far. Description:
A basic multivibrator for 'clock'. A variable voltage source (not really, but gets the job done) controls the bias to one side of the oscillator; the imbalance changes the duty cycle (more like 'on' time, since frequency is changed, but that doesn't matter), varying the charging time the inductor recieves, thus varying its stored energy. This control circuit will later apply DC-coupled NFB to regulate the output voltage.
In the other PWM control circuit, which I tried on a whim and worked first time(!!), a pair of transistors resting on a CCS form a differential pair. One is biased fixed, the other adjustable, to vary the ratio of output voltages (rather than just one), and thus give a more accurate change in duty cycle.
The signal is taken off and fed to a driver transistor, which then powers the output transistor. In one case, an NPN is driven from the osc's emitter (with a small resistor to ensure cutoff), which then drives a large plate load, err sorry, collector load, with the attached output transistor. In the other case, a PNP is driven from the osc's collector instead. I tried it both with and without current limiting and 'speedup' resistors (to swamp Miller effect I presume) and noticed little difference.
For output, a high voltage transistor clamps an inductor for that 'on' period of time. When it releases, the pent up energy explodes into a flyback pulse, which would be bad mojo if not for the damper diode (hey, it's even being used as it was made to be used!) which funnels the pulse into a filter cap.
The questions: how can I improve 1) power output and 2) efficiency? I already know the answer to #1, lower inductance or longer 'on' time, and generally beefier parts, but I haven't been able to get either an efficient driver circuit of hefty enough capacity, or good output transistors (I'll be stripping some more computer PSU boards later on) for either. As for #2, by the nature of the output transistor, the driver's load will always have +V-1V or so across it, which burns something like a watt. All the other transistors contribute maybe 1/2 or 1W as well, and the rectifier doesn't help any. All in all I get maybe 70% efficiency for 5W output (current record, 215V across 9k). I'll admit that's not bad for someone who doesn't know squat about SS, but this digital kind of circuit can push a lot more than that.
Tim
Attachments
Hello Tim,
Check out the website for Linear Technology or any of the other big semiconductor makers that sell PWM chips. Can get a full regulated voltage PWM controller on a chip. For higher power units use a mosfet driver chip as well between the controller chip and the mosfet. There are some chip designs for low noise using slowed down transition times too, important for an audio power supply. The flyback converter, as you have outlined, is probably the easiest for HV since it only needs an inductor (rather than a full transformer), but may be noisy. Boost convertors are another possibility. The chip designs are fairly straighforward (and well explained in chip application notes) if powered from a DC source, but get complex when isolation from an AC line source is required. Lots of cheap surplus low voltage switchers available for the DC source.
Another route is to get a low (say 5V) surplus switcher that has just a single turn strap thru its ferrite transformer for the low voltage secondary. Get a unit with some heft to it so there is plenty of room on the xfmr. Remove the strap and wind your own secondary, 5V per turn (use small TFE insulated wire to keep distributed capacitance down). By using a voltage doubler circuit, can cut the # of turns in half, the PWM regulator will keep the output voltage stable. Have to redo the subsequent rectifiers/filter-inductor/capacitors and voltage divider for the voltage regulation feedback signal. Use low ESR caps and fast rectifiers.
Don
Check out the website for Linear Technology or any of the other big semiconductor makers that sell PWM chips. Can get a full regulated voltage PWM controller on a chip. For higher power units use a mosfet driver chip as well between the controller chip and the mosfet. There are some chip designs for low noise using slowed down transition times too, important for an audio power supply. The flyback converter, as you have outlined, is probably the easiest for HV since it only needs an inductor (rather than a full transformer), but may be noisy. Boost convertors are another possibility. The chip designs are fairly straighforward (and well explained in chip application notes) if powered from a DC source, but get complex when isolation from an AC line source is required. Lots of cheap surplus low voltage switchers available for the DC source.
Another route is to get a low (say 5V) surplus switcher that has just a single turn strap thru its ferrite transformer for the low voltage secondary. Get a unit with some heft to it so there is plenty of room on the xfmr. Remove the strap and wind your own secondary, 5V per turn (use small TFE insulated wire to keep distributed capacitance down). By using a voltage doubler circuit, can cut the # of turns in half, the PWM regulator will keep the output voltage stable. Have to redo the subsequent rectifiers/filter-inductor/capacitors and voltage divider for the voltage regulation feedback signal. Use low ESR caps and fast rectifiers.
Don
Another way would be to use say a surplus 48V switcher and just replace its rectifiers/caps with a high ratio voltage multiplier circuit (existing inductor and a bunch of caps and rectifiers, can find a schematic in ARRL handbook or in lots of circuit textbooks) Re-do the resistive voltage divider circuit for the PWM controller feedback so it regulates the resultant HV correctly (as in method 2). No need to rewind the transformer this way. I would use a unit rated for well above the required power again though since high ratio V multipliers have poor inherent regulation, don't want to give the PWM controller fits.
problem with V multipliers
On second thought, using voltage mutiplier circuits may cause instability for the PWM controller due to the delayed reaction time between input and output of the V multiplier circuit. Would be worth trying as an experiment to see if it can be done or not. If it causes instability, a somewhat flimsy backup idea would be to take the PWM voltage feedback signal off the first capacitor in the multiplier chain. Regulation would be poor this way. Otherwise, back to method #2 with the usual full wave rectifier setup.
Also, just as a note, there is a two phase version of the usual high ratio voltage multiplier circuit that eliminates most of the ripple voltage in the output from input charging. Basically,it's two multiplier chains side by side using a shared stack of capacitors in the center, grounded leg. Requires a center tapped output transformer to drive the two phases. (looks like three vertical series chains of caps, the center one grounded on bottom, and rectifier ladder-like steps, diagonally up one and down one step, across between capacitors, output comes off the top center cap)
On second thought, using voltage mutiplier circuits may cause instability for the PWM controller due to the delayed reaction time between input and output of the V multiplier circuit. Would be worth trying as an experiment to see if it can be done or not. If it causes instability, a somewhat flimsy backup idea would be to take the PWM voltage feedback signal off the first capacitor in the multiplier chain. Regulation would be poor this way. Otherwise, back to method #2 with the usual full wave rectifier setup.
Also, just as a note, there is a two phase version of the usual high ratio voltage multiplier circuit that eliminates most of the ripple voltage in the output from input charging. Basically,it's two multiplier chains side by side using a shared stack of capacitors in the center, grounded leg. Requires a center tapped output transformer to drive the two phases. (looks like three vertical series chains of caps, the center one grounded on bottom, and rectifier ladder-like steps, diagonally up one and down one step, across between capacitors, output comes off the top center cap)
My goal with this was also to use discrete parts... I hate ICs... transistors are bad enough that you can't see inside them... but what the hell do these other 11 pins do... 
I've seen one appnote that showed a device which rectifies the line in such a way as to draw current during most of the cycle, increasing power factor. With appropriate wiring, it would easily do 500V at what, 3A?
Besides, saves you the hassle of bothering someone for chips, something that's part of the fun in tubes is doing it yourself. You SS guys got it too easy, amp in a chip my ***!
Although... I have always wondered just what I can do with a SG3524(?).
Tim
I've seen one appnote that showed a device which rectifies the line in such a way as to draw current during most of the cycle, increasing power factor. With appropriate wiring, it would easily do 500V at what, 3A?
Besides, saves you the hassle of bothering someone for chips, something that's part of the fun in tubes is doing it yourself. You SS guys got it too easy, amp in a chip my ***!
Although... I have always wondered just what I can do with a SG3524(?).
Tim
Tim,
If you really want discrete Switch Mode, why not use valves?
Efficiency isn't much of an issue for valve people, but reasonable efficiency could be got from a class C resonant system
Edit: Maybe even a non-resonant class C system would be good. There are loads of nice RF double tetrodes out there Eg QQV03-20, QQV06-40.
If you really want discrete Switch Mode, why not use valves?
Efficiency isn't much of an issue for valve people, but reasonable efficiency could be got from a class C resonant system
Edit: Maybe even a non-resonant class C system would be good. There are loads of nice RF double tetrodes out there Eg QQV03-20, QQV06-40.
Tim,
OK, a discrete design, first thing to improve efficiency is to use a mosfet for the final switching transistor. Mosfet speed makes for much more efficient operation in switching operation. Also, mosfets will require less drive. (There's also some nice, ONLY 8 PIN!, mosfet driver ICs around too that will switch the mosfet input gate capacitance efficiently and rapidly, but not absolutely necessary) Your inductor needs to be ferrite cored for efficiency and wound with a minimum distributed capacitance coil layout. Also, needs to be properly air gapped for average DC current handling without saturating the magnetic material.
You will need a feedback loop to control pulse width if you want a regulated output voltage.
The power factor controller ICs are tempting, but they rarely provide isolation from the power line. In fact, most are of a "boost" type design that doesn't even provide short circuit protection. I know, I've blown a few up!
Modifying surplus switchers is still an attractive approach for low cost, and minimal effort, if you plan to make a bunch of them. The voltage multiplier designs would require some modification of the controller feedback loop to drop loop gain (versus frequency) below unity before instability sets in. (Similar problem to that in amplifiers with global feedback except at a much lower loop frequency here.)
SG3524 and related models are like the 6L6 or uA741 of PWM controller design. Good chip, a little dated.
Using tubes for a switching design will unfortunately require another HV supply to get them started. Of course, could try bootstrapping from the output voltage, but will need a TFT backlight power supply to charge up the HV caps first. (or try a time machine to reorder causality!)
Don
OK, a discrete design, first thing to improve efficiency is to use a mosfet for the final switching transistor. Mosfet speed makes for much more efficient operation in switching operation. Also, mosfets will require less drive. (There's also some nice, ONLY 8 PIN!, mosfet driver ICs around too that will switch the mosfet input gate capacitance efficiently and rapidly, but not absolutely necessary) Your inductor needs to be ferrite cored for efficiency and wound with a minimum distributed capacitance coil layout. Also, needs to be properly air gapped for average DC current handling without saturating the magnetic material.
You will need a feedback loop to control pulse width if you want a regulated output voltage.
The power factor controller ICs are tempting, but they rarely provide isolation from the power line. In fact, most are of a "boost" type design that doesn't even provide short circuit protection. I know, I've blown a few up!
Modifying surplus switchers is still an attractive approach for low cost, and minimal effort, if you plan to make a bunch of them. The voltage multiplier designs would require some modification of the controller feedback loop to drop loop gain (versus frequency) below unity before instability sets in. (Similar problem to that in amplifiers with global feedback except at a much lower loop frequency here.)
SG3524 and related models are like the 6L6 or uA741 of PWM controller design. Good chip, a little dated.
Using tubes for a switching design will unfortunately require another HV supply to get them started. Of course, could try bootstrapping from the output voltage, but will need a TFT backlight power supply to charge up the HV caps first. (or try a time machine to reorder causality!)
Don
Tubes would be nice, a lot easier to see if they're past their ratings... but the idea is to run tubes off either a low voltage (as in a car), or a high voltage (340V or so) as in most SMPSs. Obviously, it'll be a little hard to get any more than 5% efficiency with tubes at 12V. 
Ah yes, MOSFETs - I had one before, was idling around 4W output, and with no heatsink on the thing
it wasn't even getting luke warm. Guess I need to mooch some more IRF640s from something...
Besides that, very nice rise time was had using the basic circuit's driver, only with 680 ohms driver load.
I actually have such a PWM IC, pulled from a monitor - should check it out some time.
Oh - and my inductor is already toroidial, about 50T 26AWG on a 1" o.d. core (was a medium-sized SMPS filter choke). Saturation is a good point, one I hadn't thought of; come to think of it, that might be why I can't get more power out of some of these things! Thanks for the reminder!
Now where's the cutting disks for the Dremel...
Tim
but the idea is to run tubes off either a low voltage (as in a car), or a high voltage (340V or so) as in most SMPSs. Obviously, it'll be a little hard to get any more than 5% efficiency with tubes at 12V. Ah yes, MOSFETs - I had one before, was idling around 4W output, and with no heatsink on the thing
it wasn't even getting luke warm. Guess I need to mooch some more IRF640s from something...Besides that, very nice rise time was had using the basic circuit's driver, only with 680 ohms driver load.
I actually have such a PWM IC, pulled from a monitor - should check it out some time.
Oh - and my inductor is already toroidial, about 50T 26AWG on a 1" o.d. core (was a medium-sized SMPS filter choke). Saturation is a good point, one I hadn't thought of; come to think of it, that might be why I can't get more power out of some of these things! Thanks for the reminder!
Now where's the cutting disks for the Dremel...
Tim
No pressure...
But Tim could derive all his voltage rails from one SMPS transformer. And, HV rectification is more efficient.
I'm not going to suggest that he uses mains directly. That is beyond the scope of DIY to do safely. Don't
Edit: this was composed before I saw Tim's last post. But it's all entertainment...
Indeed
But Tim could derive all his voltage rails from one SMPS transformer. And, HV rectification is more efficient.I'm not going to suggest that he uses mains directly. That is beyond the scope of DIY to do safely. Don't
Edit: this was composed before I saw Tim's last post. But it's all entertainment...
toroid surgery!
Tim,
Some precaution is in order before performing surgery on the toroid. SMPS usually have a couple of types of inductors in them. The 60 Hz AC input circuitry usually has a common-mode choke(s) with two windings on it. This is likely a high u ferrite core since the two windings cancel out incoming powercurrents and don't saturate the core. Only low level common-mode noise is impressed across the core normally. The other core type is in the DC output circuitry as part of the output filter. It handles the DC output current and so cannot be a high u ferrite core unless already air gapped. Usually these are made from pressed powdered iron which has an effective distributed air gap to lower u so high DC current can be handled without saturation. These unfortunately have high losses in a flyback design where large AC swings are occuring rather than just ripple currents. From the 50-100uH figure you quoted, I would guess the latter type. But can only be sure by doing some electrical measurements or some mechanical core measurements and calculations using the # of turns. Maybe can tell by how hot the core runs in flyback mode too. A hot core means wasted power.
In any case, gapping the core will lower the inductance dramatically in the case of a ferrite core, modestly in the case of a powdered iron core. So more turns will be required to get the same inductance. Ferrite E cores with a few sheets of paper or plastic sheet to gap them is the usual way to get a gapped ferrite core. Some POT cores also have provision for a gapped center post, or can just grind some off the center.
The usual xfmers used in SMPS are ferrite (unless from a really old unit) but often are epoxied together hopelessly.
Don
Tim,
Some precaution is in order before performing surgery on the toroid. SMPS usually have a couple of types of inductors in them. The 60 Hz AC input circuitry usually has a common-mode choke(s) with two windings on it. This is likely a high u ferrite core since the two windings cancel out incoming powercurrents and don't saturate the core. Only low level common-mode noise is impressed across the core normally. The other core type is in the DC output circuitry as part of the output filter. It handles the DC output current and so cannot be a high u ferrite core unless already air gapped. Usually these are made from pressed powdered iron which has an effective distributed air gap to lower u so high DC current can be handled without saturation. These unfortunately have high losses in a flyback design where large AC swings are occuring rather than just ripple currents. From the 50-100uH figure you quoted, I would guess the latter type. But can only be sure by doing some electrical measurements or some mechanical core measurements and calculations using the # of turns. Maybe can tell by how hot the core runs in flyback mode too. A hot core means wasted power.
In any case, gapping the core will lower the inductance dramatically in the case of a ferrite core, modestly in the case of a powdered iron core. So more turns will be required to get the same inductance. Ferrite E cores with a few sheets of paper or plastic sheet to gap them is the usual way to get a gapped ferrite core. Some POT cores also have provision for a gapped center post, or can just grind some off the center.
The usual xfmers used in SMPS are ferrite (unless from a really old unit) but often are epoxied together hopelessly.
Don
Could always burn off the crap from a core (ferrite is ceramic, right?) and glue up a paper bobbin.
I figured these cores would be lossy... seem to remember something like that. But at least it's made to handle DC..
I'll check out some ferrite cores around here and see if I can get at any without breaking them.
Oh, I have a few line filter chokes which look kinda loose, I'll try one of those (with a bit of air gap of course).
The existing cores run a bit warm, but 26AWG is good for about 400mA, and with up near an ampere through it, it's no suprise it's getting warm. But I'll consider core loss...
Tim
I figured these cores would be lossy... seem to remember something like that. But at least it's made to handle DC..
I'll check out some ferrite cores around here and see if I can get at any without breaking them.
Oh, I have a few line filter chokes which look kinda loose, I'll try one of those (with a bit of air gap of course).
The existing cores run a bit warm, but 26AWG is good for about 400mA, and with up near an ampere through it, it's no suprise it's getting warm. But I'll consider core loss...
Tim
Here's another idea for getting HV from a surplus SMPS. Get two identical supplies and yank the xfmr out of one of them. (well maybe more gently). Remove the rectifier/inductor/cap DC output filter connection to the output xfmr in the other supply. Now connect the loose xfmr in reverse to the xfmr in the power supply unit. (ie. secondary to secondary) Its primary will then produce something like the usual 370V peak voltages originally sourced from the rectified AC power. Unfortunately, this is only a peak voltage output, for 50% PWM duty cycle this can only be used to generate steady 370/2 Volts. But if the primary was of the center tapped type, can get back up to full 370 V with a full wave bridge rectifier. The output DC filtering components will as usual have to be changed to accomodate the new higher voltage, as well as the controller voltage feedback resistive divider. This approach may cause some feedback stablility problem since the distributed xfmr capacitance and leakage reactance are doubled. So some fixing of the feedback loop gain rolloff may be necessary like with the HV voltage multiplier idea. But will require a far less drastic cut in loop bandwidth than with the voltage multiplier approach, which slows response time to over several clock cycles.
smps recommendations
The discrete route has its limitations. Trying to vector board this circuit could introduce problems with noise emission and pickup, as well as point to point wiring inductance. For switching, MOSFETs should be used for lower loss. If the discrete approach doesn't give you the desired results, do consider an integrated switcher IC, such as the TOP Switch family from Power Integrations (www.powerint.com). The TOP247, 248, or 249 should do the job. Included are application notes, and an Excel spreadsheet program to help design the flyback transformer. At 400 volts, 100 milliamps, or 40 watts, the flyback converter is a good choice. You can use one transformer with dual secondaries to get both the plus and minus output voltages. I hope this helps.
The discrete route has its limitations. Trying to vector board this circuit could introduce problems with noise emission and pickup, as well as point to point wiring inductance. For switching, MOSFETs should be used for lower loss. If the discrete approach doesn't give you the desired results, do consider an integrated switcher IC, such as the TOP Switch family from Power Integrations (www.powerint.com). The TOP247, 248, or 249 should do the job. Included are application notes, and an Excel spreadsheet program to help design the flyback transformer. At 400 volts, 100 milliamps, or 40 watts, the flyback converter is a good choice. You can use one transformer with dual secondaries to get both the plus and minus output voltages. I hope this helps.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.