Hey, Svart! Big switcher needed

[PRR]

Oh great switcher-hacker:

Got a crazy product idea. It won't sell in Atlanta, but there may be a global market.

I have heard that there are US-market guitar amplifers in Australia, and surely all over the world.

They use step-down transformers to turn 240V into 120V so they don't explode.

But they still have 50Hz, which strains iron wound for US-type 60Hz power, and the transformers overheat after a while.

Cranking the "wall" voltage to 96V gets the flux down, but also under-volts everything in the amp enough to sound lame.

It is "simple" to make arbitrary "wall power". Get a big audio amp, a step-up transformer, and an oscillator of the desired frequency. There are amps made for this work. They are big and heavy and expensive.

You'd think that a motor-generator would work, but it isn't as simple as that. (Aussie motors turn 3,000RPM, while we get/need 3,600RPM for 60Hz machines.)

The 60Hz isn't really critical for an amplifer: 70Hz would be fine too. I have a hack that probably won't work.

So what about rectifying the 240VAC 50Hz to DC, then chopping it very fine (>5KHz) with a PWM, and reassembling the slices in a low-pass filter?

I think even I could do that.

BUT. Noise rejection and safety laws and old baked insulation really require that one side of the "120V 60Hz wall outlet" be common with the local wall-power and nearby dirt. A simple bridge would give an output 160VDC above local earth. A voltage-doubler could give +/-320VDC and a hot leg symmetrical around earth, but that's heroic voltage. Of course a 60Hz transformer would let us float the output down to any earth we like, but that's a big lump.

Any thoughts? Could it be a PC power supply size box for $99, or does it end up looking like a big audio amp and transformer?

 

[moamps]

[quote author="PRR"]
....or does it end up looking like a big audio amp and transformer?....[/quote]

Like PS Audio P1200, IMO.
Why not using good transformers (40-80Hz) in guitar amps in the first place?

Regards,
Milan

 

[Svart]

PRR, thank you for this honor! However, I am not a guru of any sort. A hack.. yes. A master.. uh no.

I'll try nevertheless:

Oh, this is totally do-able for 99$ and probably much less. the only signifigant product cost that I can forsee is the trafos, which would have to be sourced and *maybe* even wound for us. Unfortunately most *affordable* versions of these are going to be chinese and of marginal quality unless you do the design and QA yourself (much like studio projects and their 747 made mics..). Either way you are looking at buying a boatload, literally, to get a good deal.

most switchers can actually be used in a universal situation like this as they truly only modify their dutycycle to adjust their outputs. ~120v and ~220 are usually stepped up to around 340v-440v then chopped. we could likely size the trafo up and go for a higher initial V then chop, or we could run the switcher at a much higher dutycycle but this could heat things up a bit and infringe on our >70% that we would prefer to keep the dutycycle under.

What would you say your max output power needs to be in both I and V? Some of the chinese powersupplies I have seen coming out these days are as small as a postcard but can supply huge amounts of power density.

The topology is also important here, we could do the simple brute force type of supply with one MOSFET/IGBT running up to 99% and simple current/voltage feedback with a simple single channel PWM IC, or we could go with dual MOSFETS/IGBTs running alternate cycles and more in depth current/voltage sensing across multiple outputs run by programs on MCUs (which I am not very educated in). We could go as far as mounting daughterboards with components on them to save space but adding cost and complexity.

So while designing our own would allow us to get the features we absolutely want and none that we don't need, unless we sell tons of them we won't make too much money in the open market. I suggest that we look at what is available and what might be able to handle a component change to adapt the device to our needs, likely a feedback resisitor or zener to set the voltage output. This would be our cheapest route for small numbers of PSUs, at least for testing and verification of manufacturers of the trafos at least..

:grin:

oh and I'm not sure I answered one of your questions directly:

So what about rectifying the 240VAC 50Hz to DC, then chopping it very fine (>5KHz) with a PWM, and reassembling the slices in a low-pass filter?

that's pretty much the process of a switcher in a nutshell, so you are on the right track! the hard part seems to be getting the trafo specs right.

 

[Svart]

Moamps, I think he is trying to save space, money and weight. a good trafo of the correct size will be much more expensive, much heavier and potentially larger than a switcher of comparable power output.

 

[northsiderap]

Perhaps the transformer heating is from the extra distortion on the ground leg of the transformer secondary inside of the amp due to insufficient capacitance on the amp's PSU filter to 'smooth out' 50Hz, causing the transformer to see increased assymetrical phase loading?

It makes me wonder if there's a way to identify that distortion and create a negative feedback into the primary??

Should amp-makers just put more capacitance in the PSU filter section when shipping to Europe?

I guess I'm still stuck on trying to understand the exact effects of a lower mains frequency.

Do I need more coffee?

 

[Svart]

mmm more coffee...

 

[northsiderap]

You got it

 

[Svart]

It makes me wonder if there's a way to identify that distortion and create a negative feedback into the primary??

possibly IF the problem was introduced by the primary itself and we could guarantee that the signal we inject is absolutely the same or we are just going to create different problems. Seems like a lot of time and effort that could be spent just purchasing the *right* parts be it a switcher or linear. This is just a speculation though.

EDIT:

:shock: that looks tasty as can be.. too bad i'm at work..

 

[Svart]

Interestingly enough, Amveco specs their (toroid) trafos for both 50/60 as the same part numbers. It makes me wonder if they don't just overbuild the part and just let it heat up a bit more?

 

[moamps]

[quote author="Svart"]Moamps, I think he is trying to save space, money and weight. [/quote]

Sorry to have barged in uninvited.

 

[Svart]

oh no problem Moamps, all comments are invited, i just think he had a specific purpose for asking his question, which is to get around the obvious problems of trafos with switching technology in an effort to make a truly universal PSU for amps at high power densities.

and for some background, read some of TI's abstracts/appnotes on the uc3842/ucc38c42. The uc3842 and it's variants are, in my estimate, seen in more than 50% of switchers today. Great little ICs for what you can do with them.

 

[northsiderap]

[quote author="Svart"]possibly IF the problem was introduced by the primary itself and we could guarantee that the signal we inject is absolutely the same or we are just going to create different problems.
[/quote]

I'm talking mostly about the heating POSSIBLY caused by assymetrical phase loading on the secondary due to distortion loaded on the secondary ground leg of the transformer. I am referring to this in order to try to understand the exact cause of the heating - and in turn have a more clarified vision of the simplest solution!

Therefore, my questions now are focusing on the secondary of the amp, and the consequential assymetrical loading of the primar.

I understand that the objective here is to design a standalone unit, but I am just trying to understand the cause of the heat.

 

[moamps]

[quote author="Svart"] and for some background, read some of TI's abstracts/appnotes on the uc3842/ucc38c42..... [/quote]

I did it a long time ago.
"Industry Standard Analog ICs" Databook, SGS Thomson, 1st edition, pages 1105 to 1113.

The uc3842 and it's variants are, in my estimate, seen in more than 50% of switchers today.

IMO, UC384x is used in 95% of CRT VGA monitors.
Btw, TL494 is used in 95% of PC power suplies.

 

[Svart]

check out the new ucc38c42.. using MOS outputs.. >1mhz switch speed!

 

[bcarso]

If this is envisioned as an AC-AC converter I think the optimal strategy would be to incorporate PFC (power factor correction) as part of the initial off-line rectification of the 220VAC, since eventually everyone is going to insist on it anyway, and you get isolation along with some decent preregulation out of it as well, with opto or other galvanic isolation feedback. Then on the other side of the high freq trafo and rectifier you need a switchamp.

Although it is a daunting voltage for DIY the offline stuff has spawned a whole bunch of power FETs that are happy in properly designed circuits at high voltages.

The ideal topology for the switchamp would be the Crown OCA (opposed current amplifier) which is basically two buck converters of opposite polarity with their output inductors tied together, with only two FETs required along with a proper modulator. But since that patent is fiercely enforced, a decent standard class D high voltage amp would suffice. The nice thing is the output frequency is low, so the switch frequency will be chosen to work best with given magnetics in the output filter, and a little bit of deadtime is an issue only in terms of inductive kickback at the FETs. I'd avoid anything as low as 5kHz for reasons of acoustical emission. But 100kHz should be fine, or maybe higher if it made sense for reducing the size of the output filter.

It is tempting to synchronously rectify the high frequency output of the secondary to make the 60Hz sinusoid more or less directly. But of course DMOS always has the pesky body diode. They are lossy as can be with tons of reverse recovery charge, and to make a biploar withstand switch requires a diode in series or another DMOS part, so this is not very attractive IMO.

Or you can put a single FET in the middle of a diode bridge and switch it to the output filter appropriately. Actually, with SiC diodes available now that sounds like not too bad an option. I used a scheme like that once for a bidirectional motor drive using in this case IIRC a high voltage bipolar with optoisolated drive---the load was a d.c. (brush) motor, used for a two-axis positioning system to move photographic plates around for manual inspection under a magnifier. In this case I wanted variable d.c., and the motor was quite happy if it was pulsating.

 

[bcarso]

If this is envisioned as an AC-AC converter I think the optimal strategy would be to incorporate PFC (power factor correction) as part of the initial off-line rectification of the 220VAC, since eventually everyone is going to insist on it anyway, and you get isolation along with some decent preregulation out of it as well, with opto or other galvanic isolation feedback. Then on the other side of the high freq trafo and rectifier you need a switchamp.

Although it is a daunting voltage for DIY the offline stuff has spawned a whole bunch of power FETs that are happy in properly designed circuits at high voltages.

The ideal topology for the switchamp would be the Crown OCA (opposed current amplifier) which is basically two buck converters of opposite polarity with their output inductors tied together, with only two FETs required along with a proper modulator. But since that patent is fiercely enforced, a decent standard class D high voltage amp would suffice. The nice thing is the output frequency is low, so the switch frequency will be chosen to work best with given magnetics in the output filter, and a little bit of deadtime is an issue only in terms of inductive kickback at the FETs. I'd avoid anything as low as 5kHz for reasons of acoustical emission. But 100kHz should be fine, or maybe higher if it made sense for reducing the size of the output filter.

It is tempting to synchronously rectify the high frequency output of the secondary to make the 60Hz sinusoid more or less directly. But of course DMOS always has the pesky body diode. They are lossy as can be with tons of reverse recovery charge, and to make a bipolar withstand switch requires a diode in series or another DMOS part, so this is not very attractive IMO.

Or you can put a single FET in the middle of a diode bridge and switch it to the output filter appropriately. Actually, with SiC diodes available now that sounds like not too bad an option. I used a scheme like that once for a bidirectional motor drive using in this case IIRC a high voltage bipolar with optoisolated drive---the load was a d.c. (brush) motor, used for a two-axis positioning system to move photographic plates around for manual inspection under a magnifier. In this case I wanted variable d.c., and the motor was quite happy if it was pulsating.

 

[Svart]

No need to say it twice! JK!

and the motor was quite happy if it was pulsating.

stories from the strange but true category:

I've run into a few that weren't. some smaller 2 brush motors tended to pulse a bit in resonation if you weren't careful. I ended up just using 4 brush motors for this device but later found that it was only one type of motor of one brand which turns out didn't use brushes but had spring "fingers". turns out the commutator had gaps between the contacts that were not even and the oscillation was actually the speed of the motor at a lower resonant freq of the switching speed.

:shock:


Maybe I read what PRR wanted wrong but I just thought he wanted to use a switching PSU instead of a brute linear? Inverting is a crazy science and actually pretty tough to do *properly* without MCUs and hellish programs IMHO. The bean counters here decided to use a brand of inverter for their mobile units. these units were "modified sine" wave inverters. A couple of our units used large toroidal trafos for isolation etc. Upon plugging them in, these trafos would "sing". Sing of course being audibly vibrating at an obvious 60hz/120hz. A quick scope test confirmed a PURE square wave of alternating polarity with a little overshoot thrown in for good measure. the hash just from the harmonics was tearing the video signal to bits besides rendering the digital control completely inop.

all to save a few hundred bucks. We quickly returned to "pure sine" inverters with stepping so small it can hardly be seen on a scope. Of course with the MCU and 3 banks of 3 IGBTs it's not easy to do in any respect.

:thumb:

 

[bcarso]

There's a simple multilevel (that is, three: +, -, zero), used in recent mobile inverters, that cancels third. If you play with the concept you can use more levels for canceling higher harmonics. But at some point it just makes sense to synthesize the damn sine wave!

My experience with the three-level inverters has been mixed. They ran some things o.k., but when I tried to play cassettes on a boom box with it plugged into one (in the car, when my tape player was busted) the noise level was excruciating in the output tape audio (CDs were o.k.).

I was not advocating filterless output but just looking for a way to minimize the filter requirement. It does seem feasible to start with a relatively high frequency pulse output at the secondary of a switcher's trafo and rectify it with a bipolar switch, go to an output L-C filter, use some feedback for the whole mess, and get decent 60Hz 120VAC.

 

[PRR]

> Maybe I read what PRR wanted wrong

{sigh} Yes, you did, and it turned into a spitting match.

> use a switching PSU instead of a brute linear?

Sorry for not being clear.

Say you have a 1963 Fender DeLuxe that was once blessed by Eric Clapton. You do NOT want to modify it in ANY way.

BUT when you plug it into 50Hz power, even stepped-down to 115V, it overheats.

Why? Power transformers are designed for a specific flux density. If the AC wave is not reversing as fast as the designer intended, flux goes too high. At some point, the iron turns to jelly and huge primary current flows. For 60Hz iron on 50Hz power, just 10% of the tip of the wave is soaked with current, but that is enough to overheat the transformer in an hour (more or less).

"Export" transformers, rated for 50/60Hz, are available for many classic amps. But amp owners may resist ANY change to the amp, and certainly a major change like a power transformer transplant.

So what does it take to put a true 115VAC 60Hz "wall outlet" on a stage or studio in Sidney or Paris or Capetown?

Motor-generator? Industrial power amp? Ugh.

Certainly we can build a Class D "subwoofer amplifier" and make hundreds of watts of 60Hz with high efficiency in small size.

But, I thought, a minimalist switcher would make the AC riding on a DC level. I'm probably wrong there.

Still, the relation of the "fake 115V 60Hz wall outlet" to the real local power ground is critical.

If the vintage amplifer had a truly floating 115VAC primary, it would not matter. But the old iron had barely adequate insulation and poor isolation. Many of these amps still have the 0.047uFd "death-caps" that connect the chassis to either side of the 2-prong plug. Some vintage amp users would rather die than modernize a beloved amp.

Can we chop 50Hz wall-power into 60Hz wall power, and For-Sure stay at local ground reference, without 60Hz iron?

And preferably without special parts, but I'm still at the concept stage.

I'm doing a sim, but either it is wrong or my understanding of line-power rectification is wrong.

 

[Svart]

oh man i feel dumb. I read your post wrong, I thought you were talking about a drop in replacement for some reason, likely because I was trying too read quickly at work.

The class D amp is of course just spitting out alternating squares.

It's easy enough to create a much simpler inverter with alternating squares@ 60hz, but those old trafos are going to *sing* like crazy with +-DC on those old trafos 60 times a second... yeesh they aren't going to be happy for very long before giving up the ghost. we could play with high power R/C circuits to attempt to round the squares off but we are talking big dissipation and expensive caps.

Here is something I would try first. Find a 220VAC/50hz trafo that steps down to 12VDC at some high I output. now get one of these.

which i don't endorse at all, I just did a search for travel inverters and found this link. I'm not sure that you can reach the power levels up to what you need or IF they are true sine outputs at all.. but it IS an easy way to do what you need quickly.

BUT, I will start to look into what we can do to power these old devices. What current draw are we looking at? how much *would* you like to be able to source?