DIY 50V 40A SMPS Buck Converter attempt and current status.

[MDM3D]

Hello Graybeards,

Let me start this off as to why I need such a large lab power supply. The goal is to be able to hard coat Type III anodize aluminum. In order to do this I need to be able to provide a constant current of at least 40A per surface foot of aluminum in the bath for periods of up to an hour per setup at up to 80V for the thickest coatings but at least 45V. I am not a EE, but I do Like Power electronics!

My 1st challenge was to find a way to get 50V DC that I could then regulate via a buck topology SMPS to gain the ability to control Volts and Amps like a basic lab power supply. I accomplished this by cutting the jumpers inside of 3 dell DPS-500CB A server supplies to make the outputs floating while retaining the safety ground on the case. The 4th was unmodified to reference the total supply. Then I wired all 4 in series and added switches to turn each supply on in the correct order. Thus I achieve 49.8V Power rail with all sorts of nice things like overcurrent protection, PFC, great voltage regulation and dirt cheap price point of $20 for the power rail at not a massive amount of space.

Then I started designing the control circuits I wound a 1mH inductor using a large transformer and measured it to confirm using a nice LCR meter at 20Khz which is what I am planning on setting my switching frequency at. My input filter cap is 40,000uF and my output Filter cap is 60,000uF they were salvaged from the scrap bin. Only using them because they were available and free. My controller is an Arduino Nano which I use pair of voltage dividers to measure the voltage across the output, a shunt resistor and an op amp to measure current and a couple of 10 turn pots to adjust voltage and current setpoints. The SMPS is setup as a synchronous buck converter with 2 N Channel mosfets with a dual gate driver chip that uses the bootstrap method to provide the high side drive voltage. 

This works fine as long as I keep the duty cycle under %50 and after I get above this I blow my mosfets when (what I think is happening) the low side switch appears to turn on when the high side fet is on and this shoot-through combined with my input filter and low rds on of the fets causes them to die a very sudden death. I believe this because increasing the current rating or voltage rating of the fets doesn’t mitigate the problem at all.  Thus I believe my problem is due to dV/dt. I could also have gate ringing problems but I don’t know why the fets only die when the duty cycle is above %50. I have added ferrite beads to the gates with minimal effects.  I don’t think it is heat that is killing the fets either as they are mounted to their own 40w heat sinks.

In short any advice to help get this circuit up to %95 duty cycle would be immensely appreciated. I also have a schematic but have not figure out how to make a decent pcb design so I have been using protoboards. I am making a list of all of the components I have used thus far. It will be complete soon. 

[T3sl4co1l]

Wires everywhere!  This is a disaster!

Even if you aren't driving the MOSFETs hard, it's still important that you control stray inductance and coupling.  Every inch of wire is another ~20nH you don't need in your circuit.  Indeed, if drive is slow, that's more time spent in the linear range, where all those inductances squish around, making any combination of oscillators or spike generators.  It may be a wonder it survives one switching cycle at all!

Is that inductor gapped?

Controlling output voltage, and then current, is exactly the wrong way to do it: control current first, then voltage.  See comments here: https://www.eevblog.com/forum/beginners/pfc-inductor-voltage-calculations/msg1238337/#msg1238337

Tim

[MDM3D]

When the controller is attached the gate wires are under 2 inches which was as short as I could go without a pcb.
The inductor is not gapped as far as I am aware. I just wrapped 10 awg wire around the EE core.
My control loop uses the pots as setpoint only which get read by the adc and which ever pot is higher is what controls what sensor is looked at so when I want CV I look at the voltage dividers and when I want CC I look at the shunt. Never both at the same time. I do understand a little of the post you provided and it sounds like for the inrush current was to high and that was blowing up the fets but if I did my math correct my inrush. current should be limited to 50 amps by the inductor. Unless you think it is saturated in which case that would be bad as I never thought I would be pushing the inductor that hard. I didn't think I would need a real control loop(pid or feedfoward) because the anodizing process is so slow.

Sent from my SAMSUNG-SM-G900A using Tapatalk

[MagicSmoker]

The core for any of the primary switchmode converter topologies - buck, boost and flyback - must always be gapped, either with an explicit air gap between core halves or else a distributed gap in the form of insulating binder as used in, e.g., powdered metal toroids.

Judging by that madcap lash-up of parts I suggest you start with a much less ambitious project - 48V/50A requires paying attention to a lot of parasitics that you simply don't even know about, as evidenced by the layout picture.

[mikeselectricstuff]

You can pick up surplus 48V telecom PSUs very cheap - that would be a good starting point.

"telecom rectifier" is a good starting point for searching

[MDM3D]

You can pick up surplus 48V telecom PSUs very cheap - that would be a good starting point.

"telecom rectifier" is a good starting point for searching

I looked at

http://www.ebay.com/itm/ELTEK-FLATPACK2-48V-RECTIFIER-241115-001-/232373636535

but I admit to being confused I already have a constant 48V source I need the CC control that is the key to hard coating.

The air gap increases the magnetization current in the core but isn't that only important is the core is saturating? I have load tested the supply using nichrome wire placed in a bucket of water and the no mater what load resistance(1,2,5,10 Ohm) I use once the duty cycle gets over %50 then the problems start.   

[Siwastaja]

When you want a current limit, you need the current limit . Even if you only wanted voltage control and didn't want a current limit, you'd still need the current limit to prevent the inductor from saturating, and FETs dying in overcurrent, while running into the empty output caps (or a shorted load).

You seem to have no current limit as a part of your actual buck topology, that's one of the big issues (another one is layout). So of course it's blowing up.

How do you know what's your maximum inductor current, and is it exceeding the FET limits? Here, it's even irrelevant whether the inductor is saturating or not; you are not looking at the instant current, so even if the current is rising nice&slow (a few amps per microsecond maybe), you are going to exceed the limits. Saturation would make reacting to it more difficult, but as you are not reacting anyway, it doesn't make a difference.

What I mean with current limit, in this context, is that you measure the current way faster than it can rise, given the inductance of the inductor, and stop the pulse if it exceeds the limit. Say, you could design your components so that the inductor saturates at 80A and your FET dies at 100A, and then you'd stop the pulse right away if 50A is exceeded. This guarantees that the inductor simply cannot saturate, in any condition. Once you control this current limit, you have a so called current mode control, which is the only sane way of constructing a buck converter in 2000's, not only because it's robust and doesn't blow up randomly in corner cases or with "difficult" loads, but also because it makes sense and it's easy to understand and analyze for a beginner, as it's very well understandable only with the basic inductor equation.

If you just want adjustable current, and if the voltage is irrelevant, and if the current doesn't need to be exact (but maybe +/- 5-10% of the full range is OK), all you need is this quick "low level" current mode setpoint; you don't need any kind of PID "control loop" at all, you just bang the FETs on/off based on whether you are exceeding this current limit or not. If you want to control voltage, or need a more precise current limit, then you add it on the top, by implementing it as a classical control loop (typically PI) so that the output parameter is not the PWM duty cycle, but the low-level current setpoint!

Hysteretic current mode controller - i.e., measure the inductor current with fast opamp, compare with fast comparator with some hysteresis, use the comparator output as a FET driver input - might be the easiest-to-understand you can go for, especially if you need over 50% duty cycle which are prone to subharmonic oscillation with fixed-frequency PWM current mode control.

In any case, this is not an easy project, and there are a lot of important details I didn't go into at all.

Layout is paramount, not only in gate drive signals, but on the power wiring, especially how the DC input bypass caps are located with respect to the FET bridge.

[T3sl4co1l]

When the controller is attached the gate wires are under 2 inches which was as short as I could go without a pcb.

The gate leads aren't really a big deal, if it's not being driven hard.  The load current path is the biggest area of concern.

You can always do better -- here's one way, with copper strips and free air (point to point) wiring:



Note that the gate leads aren't particularly short (like 10" to the driver circuit), which is fine for the drive strength.

(There's a formula for that, by the way: what lead length is acceptable, given drive strength or resistance, and transistor ratings.)

Note how short the load current path (supply capacitor to drain to source to diode back to supply capacitor) is.  That's how little wiring length you can afford!

The inductor is not gapped as far as I am aware. I just wrapped 10 awg wire around the EE core.

Hmm, seems like maybe 50 turns?  At 40A, that's 2000 amp-turns, which is enough to saturate even a distributed-gap core, like a T300-26D (the toroid used here https://www.seventransistorlabs.com/Images/Buck3.jpg ).  You'll need, probably around 1/16" of air gap on that thing, to keep it linear.  (There are formulas for this as well; to use them, you'll need to measure the size of the core, particularly the cross-sectional area of the center limb the wires are wrapped around.)

My control loop uses the pots as setpoint only which get read by the adc and which ever pot is higher is what controls what sensor is looked at so when I want CV I look at the voltage dividers and when I want CC I look at the shunt. Never both at the same time.

Whoa!  So, if it's sitting there, working into a high resistance load (so the voltage is nominal and the current is small), set at some voltage, then you raise the current setting past the voltage pot setting, and suddenly it jumps up to maximum output (full voltage, nominal amps), because it's trying to drive current now?

Holy moley!  Let me off this rollercoaster!

What you need is more like: feedback <= max(voltage_sense, current_sense).

But that still wouldn't make sense, because you'd be trying to set both with one pot.  What you actually need is: two control loops, one for voltage and one for current; each one receives its respective inputs (voltage setpoint pot, and voltage sense; current setpoint pot, and current sense), and they output two different "error" signals.  You set PWM <= max(V_error, I_error).

That way, whichever sensor is closer to the set threshold, is the one doing the regulation.

The best method, however, is to cascade the loops, so PWM is set by the current controller (inner loop).  Current setpoint is then set by min(voltage controller output, current pot), and voltage setpoint is set by the voltage pot.

This way, the inductor (and therefore switch) current can never exceed the set current, and the voltage cannot exceed the set voltage for longer than it takes the voltage controller to respond (which is typically in the milliseconds).

Note the current shunt needs to be on the inside (i.e., before the output filter cap), otherwise it isn't measuring only inductor current, and the control will be terrible.

I do understand a little of the post you provided and it sounds like for the inrush current was to high and that was blowing up the fets but if I did my math correct my inrush. current should be limited to 50 amps by the inductor. Unless you think it is saturated in which case that would be bad as I never thought I would be pushing the inductor that hard. I didn't think I would need a real control loop(pid or feedfoward) because the anodizing process is so slow.

Inductors don't limit inrush current.  They can soften it, but not usually much.  Normally, SMPS need electrolytic capacitors large enough that the amount of softening isn't significant (i.e., peak current is limited more by capacitor ESR).  You can get by with less C than this, but you need special capacitors (usually film types) that are more bulky, and expensive; and you can only do so much, anyway, before the control loop becomes unstable.

But more importantly: no matter how much inductance you have, it won't protect against a short circuit!  Whether it's for lab or electrochemistry use, you will need short circuit capability sooner or later!

The inductor is definitely saturating.  It's probably saturating after just a few amperes, so it will be doing almost nothing over most of your desired operating range.

Tim

[MDM3D]

Fair enough back to the drawing board/ground zero. 

When I first started this I received some help from a EE student and he helped me setup this spreadsheet.

https://docs.google.com/spreadsheets/d/1GK_Qn1LtYD4ijKI7h_XIKCIbcDyMb96peTBUXUTSaS0/edit?usp=sharing

I have done most of this using Buck converter chip design documents but most of them are (low) power designs (500w max) and my background is ME.
Do you know of any books or design guides for high power SMPS?

[MDM3D]

[/quote]

Inductors don't limit inrush current.  They can soften it, but not usually much.  Normally, SMPS need electrolytic capacitors large enough that the amount of softening isn't significant (i.e., peak current is limited more by capacitor ESR).  You can get by with less C than this, but you need special capacitors (usually film types) that are more bulky, and expensive; and you can only do so much, anyway, before the control loop becomes unstable.

But more importantly: no matter how much inductance you have, it won't protect against a short circuit!  Whether it's for lab or electrochemistry use, you will need short circuit capability sooner or later!

The inductor is definitely saturating.  It's probably saturating after just a few amperes, so it will be doing almost nothing over most of your desired operating range.

Tim
[/quote]

So the main problem is that my inductor is saturating and loosing impedance correct? I think I am starting to understand why the buck topology isn't used at this power level normally. The inductor starts to get really big and complicated thus leading to other SMPS designs.

https://www.mag-inc.com/getattachment/Design/Design-Guides/Inductor-Design-with-Magnetics-Ferrite-Cores/InductorDesignWithMagneticsFerriteCores.pdf?lang=en-US

[oldway]

The goal is to be able to hard coat Type III anodize aluminum.

For this kind of application, I would choose a simple and robust solution:
A motorized three-phase variac with isolation transformer, a three-phase rectifier bridge and a three-phase circuit breaker.

The control of the motor would be slaved to keep the current constant.
The current measurement is made by 2 current transformers 100/1 mounted in open delta at the AC side of the rectifier bridge

The ripple of the rectified voltage is very low and there is no need to use a filter capacitor.

Isolation transformer should have a relative high short circuit impedance for stability of current regulation.

[schmitt trigger]

Re: gapped inductors.

I see that you are in the US. You can find use a Micrometals core. These are powder cores with distributed air-gaps. So you don't have to be concerned with gapping, which is difficult for a beginner to do it correctly.


Micrometals have on-line tools to choose the proper core size and type, and to calculate the winding.

http://www.micrometals.com/software_index.html