"Electronic" load - different approach (power resistor with DC/DC stage)

[Miyuki]

Hi folks,

Im preparing to build big electronic load, but not with dumping power to transistors at huge heatsink but to use big brake resistor. It is rated 2500W/400V so have low current and can be placed anywhere on long cable and can dissipate this power without any active cooling very reliable as it is industry thing.

I plan to build buck-boost stage with constant input current mode which will dump power to this resistor and will be used for higher voltage situations like 50-500V
And then second pre boost stage for lower voltages, coupled inductor boost for voltages bellow 50V

Two stage design because at high voltage stage is better to use IGBTs and low voltage stage can use high speed MOSFETS with schottky diodes

Will be used to test DC/DC stages and so
Just wonder have anyone electronic load using big power resistor ?

[nctnico]

One of my products is a DC load for battery discharging. I'm using power resistors (switched in various configuration) on a heatsink with a fan. The most powerful version dissipates 2500W maximum (continuous!) in a 3U case of about 40cm deep. The big advantage of resistors is that you can run them much hotter compared to transistors.

[T3sl4co1l]

Most of the cost is in the resistors; you aren't saving much by reducing the semiconductor dissipation towards zero, versus a modest fraction of the total.  You also spend a lot more in design cost, at least, and also inductors and support circuitry.

So, I usually do a power-DAC structure, using a unary ADC like LM3914 (or an MCU, now that it's going obsolete), and a small analog sink to fill in the gaps.

Tim

[Miyuki]

I also thought that big resistors are expensive, but than found that this braking resistors are really cheap even new and almost free when used


I want to build it also to learn something new
I think I can keep BOM cost up to 100$ as I try to optimize costs of this type of projects

[David Hess]

I have done it the opposite way with a higher voltage source driving a lower voltage resistor bank made from water filled copper pipe.  Another thing I did once which worked well was to use a bank of rechargeable batteries as a load.

Maintaining stable operation over a variable wide voltage ratio is a problem as usual with switching regulators if both continuous and discontinuous conduction modes are supported.  A tapped inductor or additional transformer can be used to help with large transformation ratios.  I never tried it with a flyback transformer but I have used an inverter stage before.

[T3sl4co1l]

I have done it the opposite way with a higher voltage source driving a lower voltage resistor bank made from water filled copper pipe.

Hah, and around 10kHz or so, it's its own buck inductor, too -- low enough Q to be useful as a load.

Just don't put any (non-water-cooled) metal near it... BTDT.

Another thing I did once which worked well was to use a bank of rechargeable batteries as a load.

I did one of these for a battery conditioning project: a couple car batteries supply a synchronous buck that cycles power in and out of smaller batteries.  With a few converters and batteries running in parallel (out of phase), a float charger is all that's needed to make up for charging and supply losses.

Makes a rather nice bench supply too, for when you need +/-100A.

Maintaining stable operation over a variable wide voltage ratio is a problem as usual with switching regulators if both continuous and discontinuous conduction modes are supported.  A tapped inductor or additional transformer can be used to help with large transformation ratios.  I never tried it with a flyback transformer but I have used an inverter stage before.

Yeah, DCM-CCM transition kinda stinks.  That's not too bad to smooth over, say with an average current mode controller, but it gets ever rougher at lighter loading.

At very light load, the modulator gain is just too high, and often nonlinear, and the control loop bandwidth has to be even lower.  Or, if not, it hunts around near-zero, which isn't necessarily a bad thing:

Say the PWM generator has a certain minimum duty / pulse width.  Then, the modulator output goes from zero (in cutoff) to minimum, over a short span of input, i.e., it has high gain (dOutput/dInput) near saturation.  This makes the loop gain very high around the setpoint, and effectively moves the PWM generator from the comparator to the error amp, which is oscillating because of the high loop gain.  If the error amp has integrator behavior (with good linearity), it in effect becomes a delta-sigma modulator.  Anything else in the system needs to have low enough bandwidth to average over the D-S output noise (or you add more error amps to implement "noise shaping"), down to whatever ripple margin at whatever smallest output you need.

For an electronic load, you don't have much room for ripple: Fsw is set by passive filtering.  Pulsing periodically, at a rate below cutoff, will just draw gulps of current from the source, and now you're testing an impulsive load rather than a steady-state one.  (It may not be enough to add more filtering, because that changes the dynamic impedance.  That's the thing I like the least about a switching load -- the AC impedance is not CC or CV or variable, it's fixed by the impedance of the input LC filter.)

If you need still more dynamic range, probably better to put several converters in parallel, of different size.  Use the smallest one at light load; probably keep running it at high load, but just have it saturate to whatever maximum design current it should run at.  Then throttle up the next bigger one, and so on -- the sequence should be something like a logarithmic IF amp, where the number of saturated stages corresponds to floor(log(setpoint)).

Tim

[Miyuki]

Its a good point about frequency dependent impedance from input filter and loop, but every load even real load have some nonlinear impedance
I hope it wont be much an issue as I need mostly steady state or slow changing load

I want to use digital controller because of this DCM issues as it can calculate duty in open loop from input voltage and just do little adjustments

[David Hess]

One way to avoid the DCM problem is to use constant-off time control or T3sl4co1l's favorite, hysteric control.

[T3sl4co1l]

They both have the dynamic range filtering problem.

A synchronous converter can avoid DCM by forcing CCM (allowing inductor current to go negative for part of the cycle).  You then need to make sure there's no zero offset error, which can be challenging in the presence of high frequency ripple.

Tim

[duak]

I've also been thinking of a PWM variable load and have experimented with a different way to implement it for a different application.  A friend of mine wanted a variable load for a 400 Hz generator he was testing.  He had rigged up a variac with a load resistor that worked well for the resistive portion of the load.  It turns out that to get maximum power transfer out of a generator the series inductance has to be compensated for, usually with a series capacitor.  I had a Copley Controls PWM servo handy so I designed a circuit using a phase shifter and multiplier to synthesize a reactive load with a leading power factor.  It worked by monitoring the voltage and developing a current drive command signal to the servo.  I also designed an energy dumping circuit that bled off excessive power from the servo's DC bus to prevent it from rising too high and damaging the servo.  This thing worked great for DC and low frequencies - I could basically set any current I wanted for any voltage (even minus) within range of the servo and energy dumper.  The servo didn't quite have enough bandwidth so it was unstable at some settings at 400 Hz.

One thing I found was that at high levels of regeneration, the servo was dissipating far too much power.  My memory is hazy, but I recall that the output devices were MOSFETs, probably FRED devices, and that during regeneration the recovered charge at the high bus voltage was causing the heating.  I tried some faster recovery rectifiers with series diodes to isolate the MOSFET body diodes, but it wasn't a big improvement.  This is when I found out about SiC devices that apparently work much better in this mode. ie., high ratio boost.

Anyway, I'm thinking of resurrecting the jig for myself but I'd like to see how this project goes.  I think that for some tests, eg., low voltage, high current this type of design could be super handy because it could compensate for the series resistance of the cables and connections and present a truly low impedance to the PSU.

BTW, I understand that the really big programmable loads actually regenerate power back on to the mains.

Best o' luck!