Hey, Svart! Big switcher needed

[PRR]

OK, I see my hidden fallacy.

I was pretending that Austrailians (etc) know which side of their line is groundED (what we call "White" in US Code). Even if Australian (etc) electrical code enforces this, Murphy's Law can not be repealed. There WILL be a mis-wired outlet somewhere in this box's life.

Therefore our 50Hz/60Hz converter NEEDS to have complete isolation, plus groundING and groundED connections. (I'm assuming that at least the groundING wire, green in US custom, will be correctly connected; if not, the venue's problems are bigger than our box.)

OK, get cheap. I pick up a PC power supply. The secondary has a fully-floating winding making 12VDC (among others). I can (in principle) pull off that 12V (and 5V) winding, count the turns, put 14+14 that many turns on, fiddle the voltage control, and get a fully floating +/-166VDC output. I feed that to an "audio amp", which could be a low-frequency high-THD chopper, and get a sine that swings +162V to -162V relative to a common terminal. That common terminal returns to the Aussie Ground Pin, to the US White hole, and to the US U-hole (just in case there is a vintage amp with a ground pin).

As a minor bonus, the "115VAC" is regulated, and probably low-passed, so it is "better" than you find on US stages. Since the PC power supply is switchable 120V/240V, it actually could sell in Atlanta GA or Lodi NJ: it is a "power filter" to turn crap "120V" power, or any global power, into "vintage 115V 60Hz" power.

Another frill: amps always hum and we don't play in the key of 60hz. 60Hz isn't any relation to a tuning-fork. It isn't necessary to feed the old US stuff exactly 60Hz: they traditionally were designed to work to 57Hz (some bookdock utility companies had slow days) and the iron will not object to higher frequency, 90 or 100Hz. You could "tune the hum" so at least it was on the 12 tones based on 440Hz, maybe even into the key you are playing in, so it "sits with the notes" instead of being a dissonance.


As an order of magnitude for size:

A 2*6L6GC Fender amp will eat 24 watts in the heater, 75 watts in the plates. The heater is a dead resistor. The plates are a cap-input rectifier, so it draws all its current in narrow spikes at the peaks. It is not uncommon to double such a load when estimating the equivalent resistance; I dunno if a switcher needs that allowance.

But the cold heaters draw 3 times the current, so we have 72 watts for several seconds at switch-on.

The power cap also has to charge. If it is a tube rectifier (older Fenders), that won't happen until after the heater start surge is falling. If it is sillycon rectifer (Marshall and newer Fender), it will cap-clamp the line. If the first cap is 40uFd, it reflects to 120V line as about an equivalent 400uFd cap-input rectifier with maybe 20 ohms in series. Could be an 8 Amp spike at the first half-cycle, though you could choose to limit it a bit lower and let it come up over more cycles (that's actually good for the vintage rectifier).

So it could be 1,000VA for a few cycles, 200 Watts for a couple seconds, and 100VA cruising but with a large part of that power at peaks.

There are 4*6L6 amps, and mighty monsters like the SVT, with double or sextuple the power demand, but that might be a higher-price converter. Many of the beasts were produced only as "Export" models and may not have the problems that the old small cheap US-only stuff has in today's global village.

 

[Svart]

Idea:

Bcarso mentioned a class D amp which will be PWM output which we couldn't use directly but through a R/C network could be smoothed well enough to work.. It didn't sink in until i actually relaxed a bit last night and thought about it without 100 other things going on around me..

It could work well enough to fool those trafos.

I already have PWM circuits that could easily drive a couple IGBTs at whatever freq and duty we decide.


also,

what if we do a class AB power amp and feed it rails from a step up switcher (at whatever V we decide) and then just stick a 60hz oscillator on it's inputs? it's a little more crude than a well designed inverter but could do the trick eh? it will likely dissipate a LOT more heat than the class D version but likely save us money on the output filtering needed.

either version will need some kind of flyback converter for the main switching element, likely the same one could be used for both. then we can decide which would be more suitable?

 

[BradAvenson]

I think the Art of Electronics has a drawing of an AC regenerator that used a 12 VAC to 115 VAC stepup transformer on the output of the power amp. I don't know if a step up transformer relaxes the amp design any, but its something to consider.

 

[Svart]

Thanks Brad. a problem still lies with driving the trafo. would could drive a 12v trafo with a 12v/60hz sine which would make some things easier, but then we must supply HUGE current, much more so than driving the PSU trafo of the amp directly with high potential rails. I will keep this in mind though.

 

[bcarso]

[quote author="Svart"]Idea:

Bcarso mentioned a class D amp which will be PWM output which we couldn't use directly but through a R/C network could be smoothed well enough to work..
[/quote]

L/C network.

But I'm still intrigued (but way too busy to investigate) the idea of switched rectification of the high frequency carrier on the other side of the HF isolation trafo, rather than a.c.-to-d.c.-to-HF-across-trafo-to-d.c.-to a.c.

We know the frequency/phase of the hf carrier; we know what we want as an averaged output; we can contrive a bipolar switch with a FET in the middle of a bridge of SiC or other fast/soft recovery diodes and not get horrendous losses; we can sync a sine wave osc to our HF carrier or vice versa; our error amp compares the smoothed L/C filtered output to the desired sine wave ref.

We wind up with an offline rectifier and bulk cap, a PFC prereg with one FET, a HF trafo for isolation, and one FET in the middle of a bridge with trafo-coupled drive or some such, an output filter, and some control electronics. I guess it really doesn't save that much, but avoids deadtime correction issues I think and another bulk cap on the load side. Might be fun. As such things often go, if it's reasonable it's been done, and if it's unreasonable it's probably been done and discarded for some good reasons. Looking to poke holes in it one thing that comes to mind is the flyback energy of the L/C filter needs to be snubbed, ideally in a way that recaptures the energy.

 

[Svart]

:shock:

i think you might be on to something. I think we have at least 3 distinct tests to conduct here with mixing between the topologies very likely.

Bcarso, you only lightly touched on a subject that I suppose needs to be addressed in this stage of design, exactly what are we looking for in regards to efficiency? we could do this easy, dirty and quick and have world class waste of power or we could have something so complex that you'd think it was designed in Germany and of course run clean and efficiently.

:guinness:

 

[bcarso]

The prereg/PFC stage should run >90%. The trafo and (if the bridge + FET is used) diode drops will be about a volt and a half drop on peaks, more if SiC (but less loss to reverse recovery), so with average currents of a couple amps that's only a few watts out of a few hundred.

Overall without too heroic of measures I'd guess around 85% could be achieved at moderate to full power out.

 

[Larrchild]

Whats the loss (and weight) of the 50 hz smoothing network? It would be working less hard with a sinusoid going in.=)

 

[Svart]

well the point is to create sine output at 60hz.

essentially we are taking 120-240vac 50hz, rectifying, chopping, reconstructing 120-240vac 60hz.

there are a number of things we could do, one of which I am interested in, another that Bcarso is interested in and has me intrigued by, and yet another that I think could work but would work outside of the bounds of rational power waste.

Bcarso, I have seen something similar to the bridge/FET idea somewhere, In high current multiphase drives but I can't remember where now.. I'll try to find it. I also saw something similar featuring SCRs as both the recifiers AND the switching elements. I know it's not the same as what you were envisioning but i thought that since that switch arrangement need not be as fast as the HF FET/IGBT switch we might be able to adapt it? Just a thought.


On another note, I have been using a new little device that I think is worth mentioning, an optoMOS relay. these are nifty little optoisolators that use MOS as the switching side. much much faster than the photodiode or transistor based isolators.

 

[Larrchild]

Erk, Yes I meant 60 hz. I have'nt gotten to Austrailia and I already have this upside down.
So upon synthesizing a sine wave, the catalytic converter requirements are minimal.=) cool.

 

[Svart]

I was thinking.. even though we are planning to drive a reactive load, is (active)PFC really necessary? It would be nice to gain that 5% more effeciency, but is it worth the complexity? just a thought due to a number of articles i have read that tend to say that other than keeping the mains clean, PFC isn't going to give us much else in return.

 

[bcarso]

I guess it is not really required. Maybe it reflects my gradual acceptance of good powerline citizenship :green:

However, you have to get across the galvanic barrier with something, be that a flyback or forward converter---so I'm thinking it might as well synthesize a resistive load too.

 

[Svart]

true true. I guess the next stage is to gather some elements of the design and start to do sims. When/If get some time this week I'll start to look more at the HF trafo. I think that will be the part that is either going to make or break the project.

 

[PRR]

> what if we do a class AB power amp

Such as, say, the Bogen 300W tube amps I rescued from cine duty? Where they were making 115V 59.9Hz at a couple hundred watts all day long?

Well, of course the modern version would be a Crown or similar audio power amp. The bigger ones do use switcher power supplies; at the powers we need, dumb-iron lingers I guess because it is simpler.

> but then we must supply HUGE current

Transistors love current (until they die); hate voltage.

The advantage of a DC-300 or similar is that they are mass-market and cheap anywhere there is a Banjo Center. (They call it something else in Germany, a family business fills the same niche.) The disadvantage is that they are scaled for 4-8 ohm loads which comes out to be 30V-40V for our power levels. Yes, a 32V:120V 60Hz transformer makes a DC-300 or equiv. into a 100+ watt "US outlet". Some of the hugest ones can be bridged to 115V output, but they are not cheap since that computes to 1,650 Watts in 8 ohms (look for a 2*800W/4ohm rating).

And actual efficiency is closer to 50% than the theoretical 78% of AB. That does not matter on the electric bill, but does mean a big lump to carry.

I'm envisioning a "PC power supply" but with a US-type wall-outlet instead of PC-mobo cables. I really think it could be that size (though at that size, it may have to be fan-cooled, a drag in the recording studio).

To get in that range, it needs to be a switcher "audio" amp with a low-pass output. 10% third harmonic is not a serious problem if it is in the direction of peak-flattening (spike peaks would be bad). Higher harmonics will radiate interference all over the stage or studio, and there are Rules about that which are sometimes enforced.

> even though we are planning to drive a reactive load, is (active) PFC really necessary?

The "reactive" (not the right concept) guitar amp has nothing to do with using a PFC input.

The classic DC supply, including PC power supplies, pulls all its current at the peak of the voltage wave. Fluorescent lamps do a similar thing. That makes the RMS higher than the average, mildly overheating supply wires. But in a 3-phase building, these peaks do not cancel in the neutral, they add. Older 3-phase assumed resistive loads, neutral cancellation, and used an undersize neutral. As fluorescent lamps and cap-input supplies came to dominate office buildings, some of these 3-phase neutral wires burned off from the huge uncancelled peak-current. And in general, distribution losses are higher with cap-input systems. In the USA, we increased the neutral on new 3-phase installations and forgot it. In the EU, they mandated PFC-input supplies. These emulate a choke-input rectifier, nice smooth current, easy line loading. They couldn't reduce the cost of 50Hz chokes, so they chop into a 50KHz choke for similar effect. As a designer, you buy a chip and implement the reference design, done.

Is it necessary? One non-PFC box is not going to burn the place down, is no worse than plugging the vintage cap-input amp directly (through a step-down transformer). But import inspectors may be watching out for non-PFC gear, and they have no sense of proportion.

 

[PRR]

I hadn't realized how simple PFC is.

Normal input has a rectifier feeding a bulk cap.

For PFC, you cut the wire, insert a 1mH choke and diode, and run a big FET across it, and a brain-chip controls it. The resulting waveform on the line is pretty ugly, but current is drawn on 80% of the half-cycle instead of 10%, and the guys who sell the brain-chips assert that you can meet the IEC specs.

It's a frill, but it can add slow-start and other frills, so what the heck.

The result is the same: a 100uFd-470uFd 400V cap charged to 1.4 times the AC input.

If we use two big transistors to drop a 50KHz square-wave across a primary, we get a square wave on the secondary. Wind it for 340V center-tapped. Put a diode bridge on that: we get +/-170V. Because it is almost a square-wave, it is almost DC. The fact that it has deep 50KHz nicks in it won't matter in the end. No real need for a bulk-cap here. (And if we do need it, it only has to bulk 50KHz.)

To get a 60Hz square-wave output, we connect two FETs, one to + one to -, to the output. Turn one on for 16mS, then turn the other one on for 16mS, repeat.

That gives a wave that is ~160V peak, same as a 115VRMS sine line, so cap-input rectifiers (such as most guitar amp plate supplies) will be happy. However it is also 160V RMS, which means tube heaters will run twice as hot as they should, no good!

There is a 3-level waveform that lingers at zero long enough so the RMS is reduced to 115V while the peaks are 160V. That's what comes out of many UPSes. It sucks.

Obviously we could PWM the FETs and get a near-sine. We might PWM at the same rate as the main converter, say 50KHz. We could go a step further and use the FETs as synchronous rectifiers, as bcarso has been muttering, though at 170V the advantage is less clear than when doing 3V power.

If the switching is 50KHz, we could avoid linear PWM and just drop pulses. At the peak of the wave, the FET is on all the time. Just before zero-crossing, the FET drops 99 out of 100 pulses to deliver about 1.7V. This might make a ton of sense with SCRs, if they are fast enough to drop-out at 50KHz. It does seem to make a spread-spectrum of output garbage, extending to 500Hz? That puts more demand on the low-pass filter, but I think it has to be pretty stiff anyway, to keep audio-band racket low.

What is this low-pass looking like? Say 230 watts at 115V, and 300Hz. We have equivalent 60 ohm load. Something like 10uFd (150V bipolar!) and 30mH at 2A. The cap is a small motor-cap, about $4 and 1.5"x1.5"X1". The choke is a very big speaker crossover part, but we would use iron-core because we won't hear it.

 

[bcarso]

The modulated rectifier approach probably gets shot down by the difficulty of controlling the reactive energies from load and output filter, the more I contemplate it. Good "old" class D is probably the way to go, as it keeps a reasonably low impedance going, even though deadtime tuning is crucial.

 

[Svart]

hmm I slept on it. Just like you say Bcarso i think the mod. rectifier adds a complexity that gets us away from our intentions. I agree that class D is the way to go here. we rectify 120-240v, chop, feed a class D amp with both the chopped HF and 60hz wave and get PWM output, then through the filter for 120-240v 60hz.

I think this is going to be the easiest way to get it into a small box. The only thing that concerns me is the HF chop and the PWM freq not playing well with each other. there will come a time during the cycle that the PWM will be high and the chop will be low unless we create a global feedback that ensures that both are high. The way I'm envisioning this is that the PWM cycle will not change it's charachteristics on it's own if the 60hz sine is firm but the chop will change due to demand.

I wonder if we could use this as a part of PFC.


PRR, I didn't follow the "The "reactive" (not the right concept) guitar amp" statement, as the guitar amp does have an input PSU transformer which would indeed be reactive. Since our power amp is driving the primary of this trafo I assumed it to be reactive in nature. As I thought anything that required power to setup a mag field before operation was reactive besides being inductive in nature when actually doing work.

As far as PFC was concerned, I was envisioning a form of active PFC for the output of our box since it would be feeding another line level device. PFC on the input of our device would be nice to have and relatively easy to implement these days. heck, I have seen a few stand-alone ICs just for that. But now that I think about it, I don't think that a PWM output would give a wet fart about what it's actually supplying to the next device. Trying to change that would change our PWM's nature.

the PWM created sine would inherently supply it's peak current at.. the peak.. since that is the area with the greatest pulse width. current here is DC so it goes hand in hand with the average voltage. A couple Rs Ls and Cs on the output to average/smooth the voltage/current to a sine and we can adjust the Cs to create a simple PFC for an inductive/reactive load.

 

[Svart]

http://www.irf.com/product-info/audio/classdtutorial.pdf

check out the schematic. semi-discrete quantization. easy enough to feed it a 60hz sine with rails sufficient enough to output 120vac.

With a MCU we could likely get rid of most of the first half of the design.

I haven't had much luck finding a good standalone Analog->PWM converter. IRF, TI and Zetex have products but they are much too integrated with other features.

EDIT: I don't think i was clear on that last statement, I am assuming that a normal PWM IC with a free error amp/comparator input will modulate the signal easily. A HV high/low FET driver IC should do the duties for a pair of high power MOSFETS for the output. I am astonished to see a lack of easy analog input/pwm output interfaces..

 

[Larrchild]

lack of easy analog input/pwm output interfaces.

NE 555 ? Too low-rent? Or you want to drive the mosfets directly..

 

[PRR]

> I was envisioning a form of active PFC for the output of our box since it would be feeding another line level device.

It is feeding the wall-plug of a 1961 Fender DeVille, or something.

We CAN'T "PFC" that: the old Fenders and other tube amps depend on their un-Green high-PF waveforms.

The antique amplifer has to see what it would have seen on-stage in Los Angeles: 115V, 60Hz, about a tenth of an ohm impedance. No PFC fake-chokes!

> I thought anything that required power to setup a mag field before operation was reactive besides being inductive in nature when actually doing work.

Inductance is a reactance. Or do you mean it will kick-back?

In an audio power transformer (delivering real audio power to a load), the input impedance is resistive for all audio frequencies. That's a goal, not a given: we know most trannies start to look like coils below 50Hz, and like coils or caps above 15KHz. For the narrow frequency range of wall-power transformers, the goal is again to have the load (heaters and rectifier) dominate all transformer impedances, and here it is easier. Inductive shunt reactance is higher than useful-load impedance, series inductance and resistance is small. These assumptions get dubious for low powers; at guitar-amp power levels the losses ARE fairly small, and resistance is the biggest issue (unless we reach saturation with over-volt or under-Hz). Transformer resistance is "part of the sound", not something we want to "correct".

This also hits my "Power Factor" hot-button. Power factor should be voltage and current out of phase, such as an unloaded transformer or an unloaded motor. These pull sines of current, just not in phase with the sines of voltage. But recently "Power Factor" has been extended, ambiguously, to things that don't draw sines at all. Rectifiers driving caps pull narrow spikes of current. In one sense, the current spike is (almost) centered on the voltage peak so the PF is (almost) zero. Yet the total energy is greater than the real energy absorbed, at least in the distribution losses, which is where low PF traditionally hurts.

> the HF chop and the PWM freq not playing well with each other

I intuit that we will need to have a Bulk Cap, possibly two. One at 50Hz scale, 375V(400V) and around 470uFd. Another (or a pair) at 115VAC 200 watt scale, though probably at HF so many times smaller. There may be a clever trick to avoid one or the other, but clever tricks sometimes turn out not so clever.

> check out the schematic.

Yeah, that's the second half. A Class-D Audio Amplifier is exactly what we want. That particular one is scaled for +/-50V rails, almost 150 Watts in 8 ohms. We want +/-180V rails to deliver 115VAC, 200 Watts, 1.74 AmpsAC, 66 ohms, with some to spare for start-up.

Then the first half IS any standard PC power supply, re-wound for +/-180V instead of +12V(etc). Turns are many-more but resistance can be much higher, so it should fit as a first approximation. Winding insulation thickness may force a larger window and core. Change the output rectifier and caps from 16V to two 200V. Since impedance is almost 1,000 times higher, our 470uFd cap on the 12V rail becomes something like a pair of 1uFd. These caps are actually choosen for delivered ripple, which upsets CPUs but will hardly bother our guitar amp behind a Class-D audio amp, so they might be even smaller (50KHz ripple rating will probably set a minimum physical size).

Then the second half IS one of the new Class-D Audio Amp chips, a couple of MOSFETs, the low-pass filter, and the USA Wall Outlet. Some difficulty working on ~360V instead of 100V as most speaker-amps do, but that's a detail.

> I am astonished to see a lack of easy analog input/pwm output interfaces..

Two-chip solution- (+/-22V supply; 28VRMS bridged output; not quite what we want).