48VDC to 160VDC (for inverter)

[Siwastaja]

This is a funny thread because everyone seems to have a completely different idea how to do this "easily", and the ideas of others are seemingly unsuitable.

Of course you can MacGyver this up with almost anything you have lying around, but that's completely up to what OP happens to have.

I stand by my idea that boost converter is the simplest with easiest-to-understand schematic and working principle, with guaranteed zero need for building custom transformer(s) or inductor(s). Yes, you need to minimize the switching loop area, but OTOH, you don't need to think about transformer leakage inductance, minimizing it with proper winding patterns, obtaining litz wire, and adding snubbers nevertheless.

In other words: if all the hardships of boost described in this thread prevent OP from succeeding, I'm 100% sure any transformer-based topology will definitely fail, they have inherently more things to go wrong (also more potential to learn).

Transformer based circuit which needs the MOSFET Vds(max) to be theoretically rated "just" 2x input (not accounting for any parasitics) does not sound, IMHO, very appealing at all compared to simple boost which needs to be rated 3x input.

Same with flyback for example, the output voltage is reflected back to the input by the turns ratio making high conversion rates appealing, but if the conversion rate is just 1:3, the difference is small, and the new added parasitics may make it actually worse unless you really know how to build (and snub) it properly.

By all means go for transformer-based topology if you want to learn about designing and building them, but if the aim is to make a simple, easy to design, easy to build, low-cost and efficient converter, 1:3 boost is hard to beat. That's exactly why I changed a transformer-based 1:7 topology to a boost-based 1:3.5 topology when I had the chance to halve the output voltage requirement, and oh boy did it get so much simpler to design and build.

And I'm sure many will disagree.

And yes, 1:3.5 boost is getting close to the edge of the "makes sense" boost sweet spot but 1:3 is clearly within the sane area.

[penfold]

In other words: if all the hardships of boost described in this thread prevent OP from succeeding, I'm 100% sure any transformer-based topology will definitely fail, they have inherently more things to go wrong (also more potential to learn).

Can you clarify that? I only disagree because the push-pull and forward are usually my go-to since I've never had an issue with one, and that includes the occasional lash-up when I've not had the ideal core to hand. In comparison, something I noticed from being a teaching assistant the vast majority of students' projects' problems involved boost and buck converters whereas a typical student producing a transformer based design tended to have fewer problems. Most problems were due to bad PCB layout or control loop - both of which are coincidentally(?) something the push-pull is a little bit more forgiving over.

[Siwastaja]

Did the students wind the transformers, were skin/proximity effect losses estimated, how about pri/sec coupling, was the criteria just "works somehow", how were the switching parts selected?

Or was it a design-done-by-the-teacher, parts supplies by the teacher, including pre made transformer?

Boost converter control loop compensation surely sucks due to the right half plane zero, but there are a lot of appnotes available calculating the correct loop compensation components. I think no topology is immune to selecting loop compensation components "randomly".

Minimizing one simple loop (switch - diode - output capacitor) is IMHO easy. Something to keep in mind, though.

I don't know, maybe push-pull transformer-isolated topology isn't that difficult for a beginner, either. I'm remembering my own experiences and my first buck and boost ever obviously were a breeze. I also spent time winding transformers, enjoyed it and were successful but I remember taking a lot more time - and measurements - doing it.

[penfold]

Did the students wind the transformers, were skin/proximity effect losses estimated, how about pri/sec coupling, was the criteria just "works somehow", how were the switching parts selected?

Or was it a design-done-by-the-teacher, parts supplies by the teacher, including pre made transformer?

I'm mentally combining two cases of independent design and a more teaching based exercise,
In cases where the student chose to design their own PSU instead of buying one, part of a larger project but not the sole design focus, "works when testing in application" would be fine, so metrics such as "doesn't catch fire", "doesn't melt insulation", "keep the output in regulation", "still works under load" would be enough. If the PSU was the sole design project then it would be a specific topology, probably something more "interesting" than push-pull, boost or buck and that's a whole new world of problems! 

For the more teacher lead exercise, the student was still expected to design their own transformer, and given a typical design procedure to follow which did include skin and proximity.

Its that last point which steers me more toward transformer based designs (I do still agree that the boost is a technically suitable option when done right so I do apologise if I sound argumentative). In complete ignorance of the OP's PCB layout skills I would find it easier to recommend a good transformer design guide (Ned Mohan's for instance) and a guide for selecting snubber components than it would be to start describing ways of keeping current loops low and speculating the possible sources of certain oscillations through the medium of the inter-web, especially when some of these concepts may very well be brand new.

[David Hess]

For 200W? That's going to be a fairly big heavy transformer.

200 watts is not very large for a 60 Hz transformer and operating at a higher frequency increases the power capability within the limits of the winding resistance.  If the operating frequency is raised, then the transformer has to be sized for the winding current and not the output power.  Hysteresis losses are higher but they started out small and remain small at lower frequencies.

I like the inverter method for simple designs because it avoids the complexities of feedback control.  The big disadvantage is lack of short circuit protection although that could be added.

For myself, I would implement a high frequency design using a current mode switching regulator but it is not something I would recommend for the casual designer.  And I might still use an inverter as outlined above with primary side regulation to avoid making a high voltage switching transformer.

[Zero999]

I don't see the problem with either approach. A boost converter can be made with a transformer.
https://www.eevblog.com/forum/projects/12v-to-400v-boost-converter/
https://onlinelibrary.wiley.com/doi/10.1002/cta.1994

Here's the principle applied to a self-oscillating topology. Q1 and Q2 need to be rated to double the supply voltage. The output voltage is equal to double the supply voltage, plus the normal expected voltage on T1's secondary winding. For example if T1 has a 1:1 ratio, the output voltage, will be triple the input.

[ogden]

I would discourage self-oscillating topologies for anything above few watts, well... unless it is homework for "PSU (in)efficiency" class. Better use push-pull SMPS controller IC or MCU with low side MOSFET drivers - if SMPS IC for some reason is not welcome.

[Zero999]

I would discourage self-oscillating topologies for anything above few watts

Why? If it doesn't need to be regulated, what's the point in using an SMPS IC? Self-oscillating transformer drivers are used in many commercial products. They're simple, cheap, reliable and high efficiency.

[ogden]

I would discourage self-oscillating topologies for anything above few watts

Why? If it doesn't need to be regulated, what's the point in using an SMPS IC?

Efficiency.

Self-oscillating transformer drivers are used in many commercial products. They're simple, cheap, reliable and high efficiency.

Really? I would love to see commercial product with [simple, cheap] self-oscillating supply providing 200W.

[edit] You most likely were joking when said "high efficiency"?

[Zero999]

Really? I would love to see commercial product with self-oscillating supply providing 200W.

[edit] You most likely were joking when said "high efficiency"?

No, I'm serious. Self-oscillating transformer drivers can be very efficient: >90% is pretty common and easy to obtain. Classic examples are electronic halogen lamp transformers and CLF ballasts. The main reason why they aren't used for switched mode power supplies is because they aren't regulated. Here are some electronic transformers with >90% efficiency.
https://www.powerled.uk.com/wp-content/uploads/sites/48/2020/05/AC-PLED-Series-Data-Sheet.pdf
https://www.osram.com/ecat/HALOTRONIC-PROFESSIONAL%20HTL-Electronic%20transformers%20for%20halogen%20lamps-Electronic%20Control%20Gears%20for%20Lamps-Digital%20Systems/com/en/GPS01_1028160/

They all use the same simple, self-oscillating topology:
https://www.st.com/resource/en/application_note/cd00003902-electronic-transformer-for-a-12v-halogen-lamp-stmicroelectronics.pdf

[ogden]

No, I'm serious. Self-oscillating transformer drivers can be very efficient: >90% is pretty common and easy to obtain.

I would not consider 90% as efficient, especially for hi power (200W) supply. Note that electronic transformer in ST appnote has AC output, thus no rectifier losses.

They all use the same simple, self-oscillating topology:

Where's proof that Osram HALOTRONIC-PROFESSIONAL HTL you mention uses such topology?

[edit] Not exact model, but anyway - no IC. Your point proven. Thanx.
http://ixbt.photo/?id=photo:610174
http://ixbt.photo/?id=photo:610175

[Zero999]

No, I'm serious. Self-oscillating transformer drivers can be very efficient: >90% is pretty common and easy to obtain.

I would not consider 90% as efficient, especially for hi power (200W) supply. Note that electronic transformer in ST appnote has AC output, thus no rectifier losses.

They all use the same simple, self-oscillating topology:

Where's proof that Osram HALOTRONIC-PROFESSIONAL HTL you mention uses such topology?

90% is the minimum and is pretty decent for such as simple circuit. The rectifier losses will be lower, at higher voltages. An couple of additional windings and MOSFETs on the secondary could be added to give synchronous rectification, but it isn't worth it for 160V.

And, no, I don't have any proof that the Osram electronic transformers use such a topology, but given the data sheet says it has a 50kHz output and it's what's inside every other electronic halogen transformer I've seen, I'd be surprised if it's something different.

It's possible it uses an oscillator & MOSFET driver IC, such as the IRS2153D, FAN7387, L6571 etc. but I don't see why they would go to the additional expense.

If you don't want to go with self-oscillating, then use an oscillator & MOSFET driver IC, such as the IRS2153D, FAN7387, L6571 etc. The main advantage is no feedback winding on the transformer to worry about, but I doubt it will gain much efficiency-wise.
https://www.infineon.com/dgdl/Infineon-IRS2153D-DataSheet-v01_00-EN.pdf
https://www.mouser.com/datasheet/2/149/FAN7387-1006880.pdf
https://www.st.com/en/power-management/l6571.html

The nice thing about the self-oscillating topology is it tends to set the frequency to the optimal point. Here's a data sheet for the LTC1697, a CCL lamp driver IC, which still uses external transistors, in the classic, self-oscillating, Royer converter topology. If it was that bad, the IC would control the driver transistors directly, rather than PWMing the Royer oscillator.
https://www.analog.com/media/en/technical-documentation/data-sheets/1697f.pdf

The circuit I posted previously, should be more efficient, than an isolated converter. It's configured like an autotransformer, so the circuit is only doubling the voltage, which is connected in series with the primary side, giving triple the voltage, thus cutting the losses by around a third.

Two of the diodes can be eliminated, if one winding is used for the primary and secondary, making it into a true autotransformer.

[Siwastaja]

Should and should. Such topology-based assumptions about efficiency and voltage peaks are of little value. For any topology, I suggest you always run a spice simulation with explicit series inductances all over the place and realistic coupling coeffs like 0.95 especially if you are not a professional well-taught SMPS transformer winding specialist.

I actually compared the efficiency of two different converters made for the same task, one with autotransformer-like tapped boost, one with simple traditional boost, and the latter was more efficient (and much simpler to build), because the snubber losses needed to keep the MOSFETs working with the theoretical "you only need this and this Vds rating" were way above I thought they were in a transformer-based design. Also transformer losses were higher than inductor losses in the simple boost.

Stray inductance is a bitch as are transformer coupling and winding AC losses.

Classic non-isolated buck and boost are pretty appealing when the conversion rates are not too big so that duty cycles don't wander too far from 50%. (Multiphases are utterly trivial to design and build and ease this condition to some extent.) The reason is in their simplicity and lack of parts that cause cost and losses. PC motherboard manufacturers use multi-phase synchronous buck converters for surprisingly high (over 1:10) conversion rates.

[ogden]

My argument is not against push-pull topology. I say that self-oscillating BJT's in push-pull converter cannot reach same switching performance, thus efficiency, as properly driven MOSFET's in converter of same (push-pull) topology.

[T3sl4co1l]

Not to mention startup and fault conditions, which are likely to destroy a basic oscillator.

You're pretty much hand winding the oscillator transformer, unless you run at such a low frequency that a mains transformer is suitable.  The boost can use off the shelf inductors (at least up to modest power levels; for >kW you're probably using many in series/parallel, or winding your own).

Oscillator has no regulation, and is made ever worse by poor transformers (e.g. split-bobbin vs. shell vs. toroid mains-frequency type).  Can be controlled, say with a buck reg in front of it, but, y'know..?

The oscillator probably has lower EMI, owing to the low loop gain.  (A typical switching reg has many stages of digital gates cascaded, effectively having extremely high loop gain, not that it's a useful metric at that point.  It can even be difficult to tame that speed, with such heavily optimized power transistors as we have these days.)

Tim

[uer166]

I don't get the point of self-oscillating topologies. They might look simple, but the number of edge cases, states, startup and fault conditions, potential for lockup, makes it probably the most difficult ones to analyze and prove that they work. They might work "on average" but you certainly can't claim any reliability figures.

It's much easier to be truly in control of what's happening and it'll work fine and be easy to validate. I'd go with a simple multiphase boost for this any day, the ratio isn't difficult at all (use SiC diodes for reverse recovery losses, straight-forward). Don't need to care about leakage inductances, winding transformers, coupling coefficients, relying on thermal noise to start-up a royer oscillator, bs, bs, and more bs.

[Zero999]

Don't want BJTs? Then it can be built with MOSFETs, which will be more efficient at low voltages. Don't do BJTs down too much though: they're more cost effective than MOSFETs, at higher voltages.

I think one important point has been forgotten here: it's driving an inverter, so the lack of short circuit protection and regulation are less important. The inverter can implement short circuit protection and vary the duty cycle to compensate for changes in the voltage.

I'm not aware of any reliability issues. Self-oscillating transformer circuits are used extensively in many products, without any issues. I've seen cold cathode lamp drives and halogen lamp transformers last for over ten years. More complex switched mode power supplies just as regularly go wrong.

I'm not saying thou shall use self-oscillating transformer, just that it's not as bad as many people think. Like any topology, it has it's advantages and disadvantages. I'd certainly give it ago.

[ogden]

Don't want BJTs? Then it can be built with MOSFETs, which will be more efficient at low voltages.

Changing BJT to MOSFET won't improve efficiency much. Simple 2-transistor self-oscillating topology can't achieve fast turn-on/turn-off times by design. Slow switching will result in conduction losses disregarding transistor type. Fast MOSFET switching needs gate driver. As many SMPS controllers includes gate drivers, it is logical choice when one needs better than said 90% efficiency. Hope this explanation helps.

Don't do BJTs down too much though: they're more cost effective than MOSFETs, at higher voltages.

48VDC shall not be considered as high voltage.

[Zero999]

Don't want BJTs? Then it can be built with MOSFETs, which will be more efficient at low voltages.

Changing BJT to MOSFET won't improve efficiency much. Simple 2-transistor self-oscillating topology can't achieve fast turn-on/turn-off times by design. Slow switching will result in conduction losses disregarding transistor type. Fast MOSFET switching needs gate driver. As many SMPS controllers includes gate drivers, it is logical choice when one needs better than said 90% efficiency. Hope this explanation helps.

That's not true at all. The feedback winding of the transformer has a very low impedance, which is ideal for driving either a BJT, or MOSFET. A BJT will take a little longer to turn off, due to the storage time, but once it starts to turn off, it doesn't spend very long in the linear region, so the switching losses aren't too high. The good things about this topology when it's used to drive BJTs are: the voltage is stepped down by the feedback winding, so there aren't the huge losses in the base resistors and the base voltage also goes negative, which vastly improves the off time. It's also great for MOSFETs, as it can rapidly charge/discharge the gate capacitance, and it can drive a high-side N-MOSFET in half bridge configuration, as the feedback to the positive side can be floating.

Don't do BJTs down too much though: they're more cost effective than MOSFETs, at higher voltages.

48VDC shall not be considered as high voltage.

I agree. MOSFETs would be a good choice for this application, probably better than BJTs, even if they have to be rated to double the supply, if the push-pull centre tapped primary is used, rather than half-bridge.

[Siwastaja]

I think one important point has been forgotten here: it's driving an inverter, so the lack of short circuit protection and regulation are less important.

It's the opposite! Inverter is a short circuit! Depending on model, they have a shitload, or two shitloads of DC link capacitors.

Don't count on it of them implementing precharge. They are supposed to take many tens of A of inrush.

Obviously you can design the self-oscillating push-pull with a soft-start, or even better, current limit.

Obviously you can wind a custom transformer with feedback windings.

Or you can just use off-the-shelf boost controller IC (with softstart and current limit!), off-the-shelf MOSFET, off-the-shelf inductor and off-the-shelf diode and that's it!

Custom magnetics and more-complex-than-non-isolated-buck-or-boost topologies are great, even fun, but a simple buck or boost is a perfect fit when conversion rates are below 1:3 to 1:4 and power levels are below about 500W. This projects falls within this range.

Now 48VDC to 320VDC at 5kW would be something else!

[Gibson486]

That actually seems like a straight forward thing to design with a boost converter. 48 to 160 is actually not difficult, but the added current constraint is a little worrisome (your current on the input needs to be ATLEAST Vo/Vi times the output). Also, I am not sure how many IC chips will handle that 48V input. I have only done boosts with max of 24V. For a motor, you can actually try to make an unregulated boost converter. It is basically the boost converter without feedback (so no IC to do it for you...just use a 555 and go to town). Again, you will plenty of input current to support this. It is a good learning experience, but I would be inclined to do this more with AC wall output power as the input, than DC.

[Zero999]

I think one important point has been forgotten here: it's driving an inverter, so the lack of short circuit protection and regulation are less important.

It's the opposite! Inverter is a short circuit! Depending on model, they have a shitload, or two shitloads of DC link capacitors.

Don't count on it of them implementing precharge. They are supposed to take many tens of A of inrush.

Obviously you can design the self-oscillating push-pull with a soft-start, or even better, current limit.

Obviously you can wind a custom transformer with feedback windings.

Or you can just use off-the-shelf boost controller IC (with softstart and current limit!), off-the-shelf MOSFET, off-the-shelf inductor and off-the-shelf diode and that's it!

Custom magnetics and more-complex-than-non-isolated-buck-or-boost topologies are great, even fun, but a simple buck or boost is a perfect fit when conversion rates are below 1:3 to 1:4 and power levels are below about 500W. This projects falls within this range.

Now 48VDC to 320VDC at 5kW would be something else!

Good point about the DC link. If the DC link capacitance is big, that pretty much rules out a straightforward boost converter, or an autotransformer, as they will blow up, when the output is short circuited.

Anyway, I thought the original poster was designing the inverter, so is in control of the DC link capacitance? As it isn't being powered off the mains, the DC link capacitance can be smaller, assuming it's not going to be doing a crazy amount of braking.

If you want to do it the easy way, just buy a cheap 48V to 120V inverter and connect the motor inverter to its output, but where's the fun in that?

[uer166]

Just don't forget to precharge the DC link to the battery voltage in the first place before connecting battery, or make sure that the boost diode can handle the inrush. The boost can't output less than the 48V input, but once they're connected, can handle any and all inrush on power-up if it's a current mode controller as it should be.

[Siwastaja]

Also, I am not sure how many IC chips will handle that 48V input.

Many controller ICs available rated way above 48V, but you don't need to limit yourself to such parts if you feel more familiar with something rated lower; just regulate the 48V down for the controller IC Vcc pin with a simple linear regulator.

In the non-synchronous boost topology, the lower MOSFET (and low-side current sense resistor) sit right on the ground plane and that's all the controller IC needs to know.

[Siwastaja]

I thought the original poster was designing the inverter

I missed this completely.

If they are designing a VFD for ACIM from scratch, it is at least an order of magnitude more difficult and complex than any of the possible voltage boosting stages, so I don't think it makes sense to concentrate so much on finding a boost circuit which can be built in absolute minimum time and skills. I think OP should be able to do this on their own, if not, I wonder how they are going to do that VFD?

I think the whole thing could be integrated and designed in one big blob. As the design will have six-MOSFET bridge driven by the MCU anyway, I see no reason not to use the same MCU to drive the boost stage as well.

The simple boost may need some precharge anyway because the boost diode conducts the input voltage to the output even if the boost is turned off. Often this isn't a huge problem because at lower voltage (here 48V) the inrush current limited by the total ESR is much lower than at full voltage, and boost current limit starts to work right when the boost is enabled, but for a really robust system in cases with very large very low-ESR capacitor banks or actual short circuits, even the simple boost requires some additional protection (like an extra FET switch in the input, driven by a controller) to protect the diode.