getting my feet wet with SMPS design

[imaradiostar]

I know, they have no place in a tube amp. Still, I want to play with them. Any suggestions for how to get started? Books, websites?

I was thinking I'd buy the chip mentioned in this thread then maybe use the cores from some computer power supplies and rewind them.

Any suggestions on where to buy ferrite cores for the 200 to 600 watt range and how to wind them?

I still think a lightweight SVT would be awesome! I also like the appeal of winding my own transformers and having control over voltages.

jamie

[exclamationmark]

There's a whole bunch of topologies, but only a certain handful are suitable for the 300v+ and high wattage range. Specifically the isolated variety (half bridge, full bridge, push-pull and some resonant types). I've built a couple prototypes, but nothing serious. I started with some simple non-isolated designs like your general buck/boost then gradually moved up to some isolated types. There are really no websites with express instructions on how to build a 600 Watt smps from the ground up, but SMPS.us will get you started if you browse around. Be prepared to unleash a lot of magic electronic smoke though! Maybe there are even kits around. If at all possible I'd start with one of those, because layout can be VERY finicky with these things. The math behind these, while simple at times, can also unravel into an ugly mess.

All smps's generally follow the same block diagram, which is rectification, filtering to rough DC, then actual DC/DC conversion (which almost always involves a FET, capacitor, inductor and diode). You'll also have a feedback loop happening somewhere, which controls the duty cycle of the unit to compensate for various things. For higher wattages, PFC is another thing to worry about. As well as short circuit protection. Many driver chips have these things integrated however. I wouldn't know anything bout winding my own cores, however...

[madkatb]

I've been playing around with some sub-miniature tubes in pedal applications. The power supply I've been using is based on a Nixie power supply. It's good for up to 250V but wouldn't have the current capability for a higher (>2 or 3w) wattage amp. There's some good info at diystombox.com (follow the links at the "effects" section of this forum) or google "nixie power supply". It's basically the circuitry as described above by exclamationmark.

[Steve Conner]

ETD core assemblies are easily available on EBay, in Europe at least. The size I've used is the ETD39, good for about 400W. I think they go all the way up to ETD54.

Assuming you're comfortable with general electronic construction and troubleshooting, the main areas of difficulty remaining are:

How many primary turns do I put on the transformer? You can calculate a lot of math (it's that equation with 4.44 in it, but you use 4.0 for a square wave) or you can guesstimate it. I believe that 50 turns is about right for an ETD39 driven by a full bridge converter off a 320V DC bus at 50kHz.

How do I stabilise the feedback loop? I hope you weren't napping in control theory class. If you were, well you can sometimes get away with no feedback at all! Which is just as well as the IR2153 doesn't have anywhere to connect it.

But only in forward-derived converters, the boost and flyback variety need feedback to keep the unloaded output voltage under control, or they will explode messily. (You'll be seeing a lot of this "explode messily".)

I heard PCB layout matters in these high frequency, high power circuits? Yup. Everything needs to be rammed together for low stray inductance, but it also needs to be spread out for cooling! Most designs place the DC bus caps right next to the MOSFETs and their hot heatsink, and we wonder why the caps don't last so long.

If considering a DIY high power SMPS, you may find some help over at 4hv.org.

[imaradiostar]

Thanks guys. I bought some tiny nixie switchers on ebay for guitar effect use but haven't had a chance to use them yet. I intend to stuff an SIB Varidrive into a boss enclosure as a joke for a friend. I wouldn't use that kind of design for a tube amp though.

I like the little IR chip because I intend to use a free running isolated design- basically the same circuit as a standard tube amp but running at 50 to 100k so I don't have to have a giant transformer. I suppose I could have a pro help me with the design.

I'm an electronics and mechanical engineering tech in my day job so I'm familiar with a lot of different designs. I like the MarkBass design because it is rather simple.

Thanks for the tip, Steve. Ebay is probably a good place for cores because you can get things from other countries pretty cheaply.

jamie

[MarkusBass]

I wouldn't start from high power SMPS.
But if you have necessary knowledge you can go to: diysmps - The Front Page
or PowerEsim - Free SMPS Switching Power Supply / Transformer Design Software
or diy smps for audio amplifier
BTW, several valve amps use SMPS. They even work better than classic power supply - they have less hum if designed properly (unless they fail). I think that SMPS in AMPEG SVT is possible but what would be still left is the output transformer, which is still heavy. And this cannot be replaced unless you go for Class-D power amp. And such Ampeg amp already exists.

Mark

[Steve Conner]

Well, for me the flaw in the "lightweight SVT" plan is the speaker cabinets. A big bass rig won't sound big without some bigass heavy speakers, and if you're already lugging those, a tube amp head isn't really that much more bother.

Of course you still want to build it anyway so let's continue. I think the Markbass SMPS actually has a rudimentary form of feedback, the transistor slung across the IR2153's timing resistor. Maybe the transformer is a resonant design and the feedback shifts the frequency until the voltage is right.

In any case, it would probably run with no feedback at all, but a minimum load would be advised! If the power supply gets completely unloaded, the output voltage can almost double because of ringing.

The heater and idle HT draw of a tube amp would be plenty.

[imaradiostar]

I guess "lightweight SVT" is a relative thing. I plan to use big bottles at around 750 volts but only two of them and a single 4x10 cabinet. It might not be the best plan but it's just something I want to build for fun. I also like the idea of the experimentation. That's much of the fun for me.

jamie

[imaradiostar]

Oh, and you're totally right. I hadn't noticed the other winding feeding the transistor into the IC. For my purposes I still think I could get away with a free running design though. As you said, some load is needed. The small car inverter switcher I've been playing with has no feedback that I can find and it sags very little with changes in load- the sag is in fact pretty much directly related to the DC sag on the primary side.

jamie

[g1]

http://music-electronics-forum.com/t23974/

[R.G.]

Be aware that if high voltage AC power line and transformer based power supplies are dangerous, high voltage/high power switching power supplies are multiply so. Steve's "explode messily" was a daily occurrence back when I was working on these in the mid 70s, before we really had semiconductors which could do the job well. A modest failure to turn a power switch off fast enough (and that's nanoseconds at today's switching speeds) lets it go nova, dissipating the entire energy in the primary filter caps in itself.

One of my power supply lab techs refused to ever build and test anything for me again after a mistake I made had shards of diode-epoxy splattering all over the lab. Electrocution and fires are a distinct possibility until you get some experience in the bench work.

I have this vague idea that comes to me every now and then about the perfection required with higher power technologies. I can't state it succinctly yet, but someday I will. Humans have a fixed capacity to absorb energy. The amount is different depending on the energy type: mechanical stress, acceleration, heat, shock, point stress (that's the energy density needed to puncture human tissue), UV, ionizing radiation, etc. For each of these, humans can tolerate up to some level X before Bad Things happen to the human.

Humans do well at some level. For the mix of stresses and released energy in transportation, for instance, humans are well equipped to deal with the stresses and excesses involved in horse travel. Most horse mishaps are non-fatal to most riders. Cars involve higher speeds, and any "leakage" or imperfection resulting in energy transferred to humans in cars is more serious. Airplanes are even more so, to the point that simple exposure to the outside environment of the air during travel can be fatal. Crashes are another example. You have a pretty good chance of surviving a horse crash, less so of an airplane crash.

I put this down to the "perfection" needed in the process. As the power density and energies involved in any process get higher, the process must be more perfect in terms of not letting more than the tolerable amount of leakage happen. A 100W power supply leaking out 10% (10W) of heat will probably not even burn you. A 100MW power station leaking out 10MW of heat is positively dangerous. The reason is that humans have fixed thresholds.

So as the total power/energy involved in a process goes up and especially if the power density goes up, the perfection of design and reliability gets more critical to keep leakages and failures from exceeding fixed human tolerances. Antimatter and collapsed-matter processes are likely to never be safe to use.

[Steve Conner]

I agree. I know a guy who decided to get out of power electronics because of the danger. He was doing the industrial stuff, and had some nasty experiences at the 100 kilowatt level.

The components available now are much better, and building your own SMPS is probably quite safe as long as you use a dim bulb tester, an isolation transformer, safety goggles, a flameproof suit, and a 10 foot scope probe.

Smiley faces aside, I am serious about the goggles. I've been hit in the head by those little pieces of epoxy shrapnel, and you wouldn't want one in the eye.

Sometimes the components are too good. The other day I heard about some cheap TO220 MOSFETs that were far too fast and easy to drive. Hooked up to a powerful gate drive IC, they were probably switching in under 10ns. The fast switching was producing something like 300V of spikes and ringing on the DC bus at 90MHz. This was enough to confuse the gate driver chip and cause yet another "explode messily".

I have a corollary to RG's theory of energy density. The more efficient a power electronic circuit is, the easier it is to wreck. High speed electronic protection can help with that.

[MarkusBass]

I guess "lightweight SVT" is a relative thing. I plan to use big bottles at around 750 volts but only two of them and a single 4x10 cabinet. It might not be the best plan but it's just something I want to build for fun. I also like the idea of the experimentation. That's much of the fun for me.

jamie

I think that on diyaudio.com there was a project of a bass amp with some cheap Russian tubes running at 1000V or even more. The preamp was based most probably on one of AMPEG amps.
If you build it, I'd like to see the schematic of the power supply (SMPS of course). I had exactly the same idea.

Mark

[imaradiostar]

Thanks for the links!

Is there a middle ground that's safer? I see a lot of cores on ebay so the financial investment wouldn't be very big. I have a rapid prototyping PCB router at work that isn't used very often but still would allow me to make a PCB pretty quickly.

I regularly work with switchers at work, some of which are very fast and high power but they were built for us by another company. We've seen plenty of them blow on tests and thankfully have never seen epoxy fly. I've managed to blow up a few devices at home but I always wear safety glasses!

jamie

[MarkusBass]

If you have a PCB router, you may find this site interesting: Design Examples | Power Integrations - there are even Gerber files for their example designs.
You may also be interested in their SMPS design software: PI Expert Design Software | Power Integrations

Mark