chromium
Hazard to Others
Posts: 284
Registered: 27-6-2005
Member Is Offline
Mood: reactive
|
|
AC power supply for experiments
Efficient electrolysis, electric arcs, glowing graphite rods, induction heating - quite a lot can be done if you have suitable power supply at hand.
On my continous strugle for getting most of PSU-s that are not designed for the task i want to do with them i now understand that i need some
experimenting with powerful multivibrators, transformers and DC to DC converters.
I do not want to make these using just someones worked out solution without understanding exactly how it works and how it can be changed to suit
different tasks. Instead i want to experiment with capacitors, coils, transformes and alternating currents to discover main principles myself.
For start i need something that converts DC (for example 12V 8A from computer PSU) to AC with selectable frequency (for beginning 2 or 3A of maximum
AC output current is acceptable). This device should work with any kind of inductive or capacitive load and should be reliable even in case of harsh
treatment. Overcurrent protection for output stage is probably not needed as i can limit maximum DC current at input.
Generating logic pulses with selectable frequency (and selectable duty cycle) is absolutely no problem for me. I can use NE555 timer, microcontroller
or LPT port of my PC but i am not sure how to convert these pulses to high power AC in case of unknown inductive or capacitive loads (i have done this
in case of resistive loads as temperature controller for 40V heating element).
As i understand i have to use dual H-bridge as output stage. Now there are problems. Can i just switch H-bridge as if there were only resistive load
(this is easy for me) or maybe induced currents will do anything they want with voltages in my circuitry and something special is need to get reliable
switching?
What voltage and current ratings should i choose for my H-bridge? I suppose i need some reserve, at least regarding voltages.
I can use bipolar transistors and common FET-s but no IGBT-s. I also have some triacs, thyristors and lot of monolithic H-bridges from old printers (
L298, LMD18201 and L6204 ).
|
|
|
IrC
International Hazard
Posts: 2710
Registered: 7-3-2005
Location: Eureka
Member Is Offline
Mood: Discovering
|
|
Inverters do what you ask but the waveform sucks and the efficiency is low. Nothing beats a variac and transformer combo for a lab AC supply. You can
also drive an alternator with a DC motor if you must.
|
|
|
12AX7
Post Harlot
Posts: 4803
Registered: 8-3-2005
Location: oscillating
Member Is Offline
Mood: informative
|
|
So, what do you want? A 12VDC to 120VAC inverter? Get an off-the-shelf model, they honestly are much cheaper, even understanding included. (You can
cover understanding on your own DC-DC converter experiments, however. If you wish to include a 12V-160V converter with 60Hz "magic waveform" chopper,
that's your choice......)
If you want to deal with 12V at any more than a few amps, use push-pull. Any bridge is just utterly silly. It's just another four turns on the
transformer, anyway.
Inductive loads are easy -- the flyback current is clamped by the intrinsic diodes in your MOSFETs (don't even give BJT's a second thought these days;
at high voltage (>300V), IGBTs are worth a look, too). Fast switching is easy because current responds slowly; filtering can remove the harmonics,
allowing highly efficient class D (switching) operation, pushing efficiency in the 90-99% range easily. But capacitive loads will easily smoke
things. You're a sick, twisted person (not in the good way) to load a class D amp with capacitors! I mean, if you're synthesizing a ~60Hz waveform,
sure you can get some capacity. But be very mindful of the feedback system, if you're using one. Too much L or C on the load can easily cause a
fatal phase shift, putting more than your happy sinewave on the output.
For HF outputs such as induction heating and welding, you can use the clock frequency. Welding may need PWM, and special processes may want a
specific waveform spec. Output may be CC or CV. Induction heating needs a sine wave, but the fundamental can be coupled through an inductor to allow
switching. Mind the 180 degree phase shift in such a topology.
Tim
|
|
|
chromium
Hazard to Others
Posts: 284
Registered: 27-6-2005
Member Is Offline
Mood: reactive
|
|
| Quote: | Originally posted by 12AX7
So, what do you want? A 12VDC to 120VAC inverter? Get an off-the-shelf model, they honestly are much cheaper, even understanding included.
|
I probably should have been stated this but i have not interested in generrating 60 or 50Hz as there are much easyer ways to get this. I want much
higher freqs like those that are used in DC to DC converters or switchmode PSU -s. In this case on can use smaller transformers for given power. If
everything is understood in case of low voltages then i can use same principles using recitified mains voltage as DC supply. Now got the idea?
|
|
|
not_important
International Hazard
Posts: 3873
Registered: 21-7-2006
Member Is Offline
Mood: No Mood
|
|
DC to DC converters useually run fixed frequency or over a narrow range of frequences, before the output filtering they make rather trashy pulse-like
'AC'. Some designs depend on a resonance in the reactive parts of the cuircut, really fixing the frequency. They are rather different than your stated
goal of variable frequency, and don't make general purpose AC.
As 12AX7 said, it depends on what you want it for; the waveform needs to match the application and load.
The Motorola/Freescale 56F8xx DSP family has built in PWM control for motors and the like, reading its documentation might be the quickest way to see
what goes into controlling power PWM, and faking sine waves.
|
|
|
chromium
Hazard to Others
Posts: 284
Registered: 27-6-2005
Member Is Offline
Mood: reactive
|
|
Ok, thanks so far but i still do not get answers to my questions. I try to explain once more what exactly i want to know.
Suppose i have a coil. Suppose i have DC power supply. I want to connect coil with supply and want to change direction of current, say, 1000 times per
second. (I do not know how this kind of current is named, i thought AC but it probably is not)
I need to do various measurements with different coils, cores and secondary windings on various switching frequencies. For example i want to measure
voltages on secundary windings with certain load, or changes in core temperature depending on frequency and input power.
Can i do this using H-bridge? Are there better ways than H-bridge for doing exactly this (switching direction of DC through coil with externally
controlled frequency)? As i understand from what 12AX7 said inductive loads do not blow output transistors (and do not affect switching) but if there
is some capacitance, problems my arise. Is this right?
[Edited on 3-8-2006 by chromium]
|
|
|
not_important
International Hazard
Posts: 3873
Registered: 21-7-2006
Member Is Offline
Mood: No Mood
|
|
It sounds as if you are just switching the current direction, with no other waveform shaping. Guess it could be called bipolar class-D.
That's not quite right, inductive loads slow the current rise rates and give 'inductive kickback' when you turn the current off. Sometimes the
intrinsic diodes in MOSFETs are enough protect, especially if you go through the design and calculate what it will be - making sure the kickback and
MOSFET get along. If you are just tossing windings on you might be able to zap the MOSFETs, external protection is smart.
Capacitive loads draw maximum current as the voltage is applied, and drop down as the capacitance is charged. You can limit the maximum current
through adding small value series resistance, designing the switches to current limit (see that DSP), and so on.
You're planning to direct drive your coils, best seeing as you want a range of frequencies. Either H-bridge or totem pole can be used, push-pull means
putting a pair of windings on where one transistor of a out of phase pair drives one of the windings (sometimes called center tapped). For a desired
applied voltage of V, the total pole requires +V and -V, H-bridge and push-pull just need V.
You'll likely want to test each new coil modification at reduced power, interesting things happen as an LC load goes through resonance.
|
|
|
12AX7
Post Harlot
Posts: 4803
Registered: 8-3-2005
Location: oscillating
Member Is Offline
Mood: informative
|
|
Oh, well then. Read my inductor page.
A single transistor, clocked appropriately, with a current sense resistor (or current transformer, for AC) and a snubber to absorb the flyback pulse,
works quite nicely. Without the load resistor wasting current, this would otherwise be a simple boost/flyback supply. (Upside-down, it becomes a
boost/buck negative voltage generator, and with a seperate winding it becomes an isolated supply.)
Any kind of push-pull system will work for testing an inductor. PP, from a single supply, with a CT'd winding, works just the same, since the
voltages are both the same direction, but the windings and thus flux are in the opposite direction. Half bridge works intuitively, though needs a
split supply. You can just as well split a single supply using coupling capacitor(s), with the advantage that a mismatch in ON-time or supply voltage
doesn't cause a DC current. An H-bridge (two half bridges in series, thus double the voltage drop!) is harder to functionalize because there is no
ground point, both ends of the load move.
I would be more interested in seeing how you're going to handle drive. Simple voltage coupling through a transformer works (see my induction heater
page 3-4ish), but sucks for frequency response and somewhat for drive (especially for loadey gates). For beef, a floating power supply works well
(sometimes supplied from the low side supply using a capacitor and diode, but this depends on the bridge constantly switching, and the low side
transistor(s) saturating good and low). Signal can be supplied by optoisolator, but slowly (1-20us response time!); transformers are still limited at
LF, but transmitting current (at a low voltage difference) instead of a voltage waveform will net you more bandwidth. Of course, P-channel FETs will
work too, with a suitable driver, but are hard to get in big power.
I'm off to the head, more later..
Tim
|
|
|
Twospoons
International Hazard
Posts: 1326
Registered: 26-7-2004
Location: Middle Earth
Member Is Offline
Mood: A trace of hope...
|
|
Here's the short version :
Yes, use an H-bridge. There are some very nice fully integrated H-bridges around that will handle 25V or more, 10A or more, complete with overcurrent
and back emf protection. They have logic level inputs, making them dead easy to drive. This is by far the simplest solution.
Look for automotive ones from companies like National Semiconductor. ST have one that will handle 30V, 30A loads. They will be listed for driving DC
motors and/or solenoids.
The simplest way to handle a capacitive load (we are talking 10uf or more - not pF of parasitic C) is to put a resistor in series to limit the peak
current.
If you need more current or voltage you can use an H-bridge driver chip with external MOSFETs or IGBTs - this is the way to directly handle 100's of
volts and /or 100's of amps. IRF make some nice ones - up to 1200V!
Helicopter: "helico" -> spiral, "pter" -> with wings
|
|
|
bio2
Hazard to Others
Posts: 447
Registered: 15-1-2005
Member Is Offline
Mood: No Mood
|
|
....use NE555 timer,..... PC but i am not sure how to convert these pulses to high power AC .......
This is easily implemented by driving an opto isolator/triac driver directly from the
timer.
As far as unknown loads go triacs are made up to 40A/800V and a rugged snubber should prevent problems.
Teccor claims their "quadracs" (4 quadrant triacs.) do not need a snubbing network.
I've had good luck with this 40A triac
and using fast fuses makes it pretty hard to fry them as long as the case temperature is kept below70deg.
|
|
|
chromium
Hazard to Others
Posts: 284
Registered: 27-6-2005
Member Is Offline
Mood: reactive
|
|
Thanks, your replies are very useful.
I did some experiments with LMD18201 H-bridge. Bridge was driven by logic pulses from microcontroller. I could change direction of current in bridge
output with any desirable frequency. 12V DC from computer PSU was used as main supply. H-bridge was capable of max 1.5A of current. This, aproximately
18W, input power was converted to biderectional pulses and driven into coil (unknown type with 1cm ferrite rod as core). Frequency was set to 7 KHz.
All this setup consumed less than 100mA current. For some fun i placed copper ring (1mm diameter wire, both ends soldered together) around the coil.
Copper ring turned burning hot very soon while current consumption increased to something over 1A.
I also tried kind of coil as commonly used in induction heaters (12 windings of 1mm diameter copper wire, no core). I changed frequency to something
like 100 KHz to get any results. This setup was able to warm screwdriver to aproximately 45C.
To get any glue how efficiently this kind of current can be converted by transformers i used one with two identical windings on toroidal ferrite core.
Primary and secundary, both 40 (or maybe 50) turns. I connected full wave rectifier and 12V 20W halogen bulb to secondary winding (primary was
connected to H-bridge). By measuring currents and voltages at input and output i got that 67% of initial DC power from computer PSU reached light bulb
through H-bridge transformer and recitifer.
For some fun i connected two more such transformers between H-bridge and recitifier so that power was transformed three times. In this nonsense setup
i got efficiency of 55%. Working frequency was in both cases 2400Hz.
Conclusions? Much more power is needed to get anything useful. H-bridge seems good for me as i can change working frequency easily - more chances to
get any results if one (just like me) does not know the theory behind coils and transformers.
|
|
|
Twospoons
International Hazard
Posts: 1326
Registered: 26-7-2004
Location: Middle Earth
Member Is Offline
Mood: A trace of hope...
|
|
Your efficiency measurement sounds way out to me. How exactly did you do your measurements? You mention a rectifier bridge - did you put a
substantial filter capacitor after it (100uf or more) ? If you are trying to measure unsmoothed DC ( i.e. no capacitor) most digital meters will give
you crap results.
Helicopter: "helico" -> spiral, "pter" -> with wings
|
|
|
chromium
Hazard to Others
Posts: 284
Registered: 27-6-2005
Member Is Offline
Mood: reactive
|
|
I measured current and voltage in both ends. Computed respective power by P = I * U and efficiency by dividing these. In one efficency measurement
output power was 1.08 A, in another 1.29A (could be changed by changing work frequency, i set frequency (by steps) to get brightest light from lamp).
My multimeter measures quite good aproximation of mean value of input signal during 0.2 ... 0.4 sec. so i did not use capacitor but as i have to do
similar measurements again i may try with large cap but i do not think this changes anything much. Do you think my reuslts are surprising?
|
|
|
12AX7
Post Harlot
Posts: 4803
Registered: 8-3-2005
Location: oscillating
Member Is Offline
Mood: informative
|
|
Ya, but are you accounting for peak and reactive currents?
Output power is NOT in amperes. Power is watts (V*I (real) = P).
As Don Lancaster says, lots of far-out results stem from BAD PRACTICE. You could easily run 10A reactive with only 1A really being consumed.
It especially seems to me that, at 2400Hz, you're probably saturating your transformer, causing a mess of peak current through the inverter and
primary, and attenuating the output waveform.
Likewise, your induction test at 100kHz was probably too high to get much current through the coil (although voltage induced ala shorted turn will
work out the same). Not like you can get any appreciable VAs without a capacitor to cancel the reactive current. 18W isn't much anyway, although it
might get a coat hanger red hot on the tip.
Tim
|
|
|
chromium
Hazard to Others
Posts: 284
Registered: 27-6-2005
Member Is Offline
Mood: reactive
|
|
| Quote: | Originally posted by 12AX7
Output power is NOT in amperes. Power is watts (V*I (real) = P).
|
I know this. I meant output current. It's 2:49 AM here now, so i am quite tired. Do you really believe its realistic to get iron wire to red hot by
inductive heating with only 18W power???
One source of efficiency loss in my circuit is rectifier as it will probably cut off something like 1.3V from output.
It seems you do not understand what i do and why. Those transformers and coils surely are not the best for this purpose. I never claimed (or thought)
they are. I just used what i have at hand without spending any money or designing my own. I probably will do both once i feel convenient but now i
want just see what happens if... This is not pyrotechnics where unplanned experiments may be suicide.
|
|
|
Twospoons
International Hazard
Posts: 1326
Registered: 26-7-2004
Location: Middle Earth
Member Is Offline
Mood: A trace of hope...
|
|
Ever seen the heater element on a soldering iron? They can be glowing with only 20W input. Put heat into a small enough space and you don't need
much to get something red-hot. Like in a 1/4 W flashlight bulb !
Your 1.3V drop on your rectifier is only losing you ~10% in a 12V system.
I will reiterate my earlier comment: only the very best "true RMS" meters are capable of measuring the the kind of pulsed DC you would get from your
rectifier. This actually is one area where the old analog meters have it all over digital ones! As 12AX7 points out - bad practice gives screwy
results. I suspect this is what causes all those "over-unity" results on those free energy websites. I really suggest you try your experiments again
with filter caps after the rectifier - its really easy to do, and eliminates a potential error source.
Helicopter: "helico" -> spiral, "pter" -> with wings
|
|
|
12AX7
Post Harlot
Posts: 4803
Registered: 8-3-2005
Location: oscillating
Member Is Offline
Mood: informative
|
|
I may or may not understand your goal (learning seems like a straightforward goal to me), but I can assure you I understand just about everything
electronic. One of the tools that's given me so much information is my oscilloscope, to me utterly indispensible when messing with switching
circuits. A current transformer is also indispensible, now that I've been using one in my experiments.
Now I said what I did, because you gave no mention of monitoring waveforms, be they voltage or current. Understanding the waveforms tells you what
the circuit is doing.
Tim
|
|
|
not_important
International Hazard
Posts: 3873
Registered: 21-7-2006
Member Is Offline
Mood: No Mood
|
|
I'm going to echo what Twospoons and 12AX7 said - when the waveform is very far from a sine wave most meters will not read correctly. Place I worked
at used an inverter to get 120V 60 Hz for testing product to be shipped to North America. They had several days of head scratching until it was shown
that the stepped 'sine' wave output of the inverter was causing bogus readings. The power supplies in the units being tested were happy with the odd
shaped applied power, but the measurement gear had all been designed with the assumption of standard clean sine waves power.
The switching currents in the inverter also sent all sorts of spikes out along the building mains, causing many problems - oscilloscopes
mistriggering, computers crashing, things like that. But that's an entirely different issue.
|
|
|
densest
Hazard to Others
Posts: 359
Registered: 1-10-2005
Location: in the lehr
Member Is Offline
Mood: slowly warming to strain point
|
|
If I might break in here - you really want to understand the magnetics in your transformers. The core materials are the "weak link" in most designs,
quickly followed by the slow turn-off of most semiconductors, followed by leakage inductance which stores energy which gets fed back into your
vulnerable switching components at exactly the wrong time. Look at your physics textbooks: amperes x turns x permeability (can be
measured if you don't know) x .0000012 = Teslas. Most ferrite cores can't handle above about .3 T (= 3000 gauss). Powdered metal cores can take up to
1-1.2 T (10000 gauss+). If the magnetizing force (amperes * turns / length) exceeds the saturation value for your core, it quits transferring energy,
it quits limiting current, and it gets very hot. The coils no longer limit the current to the switching semiconductors so the current increases very
very quickly to fry the semiconductors. The result is pretty spectacular but usually more expensive than most experimenters can deal with. As long as
you keep the magnetizing force (amps * turns / length of core) below critical levels you can use rules of thumb pretty easily.
Typical values for permeability for ferrite cores run in the 2000-6000 range. Toroids are good for transformers
if you wind the primary and secondary as close together as insulation and other nasty physical requirements allow. If you have access to an
inductance meter (less than $100 for a cheap LCR meter), measure the inductance of your transformer with the primary shorted as close and as well as
possible. Readings under 1% of the inductance with the secondary are usually necessary for painless operation.
If you can find out the characteristics of your toroids www.micrometals.com, the Phillips Ferroxcube site, and other core manufacturers' application notes have a lot of good information.
Snubber networks are your friends. They get hot so your switching transistors can live.
[Edited on 10-8-2006 by densest]
|
|
|
12AX7
Post Harlot
Posts: 4803
Registered: 8-3-2005
Location: oscillating
Member Is Offline
Mood: informative
|
|
Snubbers? Pah... I build bridges with transistors that automatically handle inductive kick. In fact leakage inductance would tend to protect these,
for reasons already mentioned. 
Tim
|
|
|
Twospoons
International Hazard
Posts: 1326
Registered: 26-7-2004
Location: Middle Earth
Member Is Offline
Mood: A trace of hope...
|
|
Better still use quasi-resonant switching - no inductive kick, greatly reduced switching losses, very little EMI. But I think this is getting out of
the realm of what chromium wants to do. 
Helicopter: "helico" -> spiral, "pter" -> with wings
|
|
|
![[*]](images/xpblue/default_icon.gif){kind=icon}
