How would one make an AC Programmable load?

[calzap]

A network of high-power BeO or AlN resistors, like those made by EMC, should work well.  They are non-inductive and frequency ranges are typically DC-2 GHz.  Available 2-1000 ohms, maybe more, and with power ratings up to several hundred watts.  They do require heat sinking, but a shared heat sink arrangement should work because only a few would be powered at one time.  Real questions are how to interconnect them (manual switches, relays, solid state, ?) and whether to use a microcontroller to achieve a specific resistance.

Mike in California

[Jeroen3]

Thyristors are often used for switching ac.
You'll end up with a ladder of resistors and thyristors. Not quite the flexibility as the dc loads have.

[johnh]

   There not programmable but its robust. Might be of overkill though.

[staze]

Yeah... Was thinking load banks. We use them for commissioning data centers.

Sadly, yes, overkill.

One would think something like a Variac would work (I know someone mentioned it earlier), just run "in reverse". Or is that not really possible? The issue would be audio is not just a constant amplitude sine...

Yeah, guessing still better to do resistors. Was thinking about repurposing my old computer water cooling gear to create a nice heat sink for a power resistor (or two).

Shame there's not a good way to do this. The question was more hypothetical. :/

[quarks]

unfortunately I do not have a DC load.
But I would just try/use a nice beefy bridge rectifier on a big heatsink (+ fan if necessary) and connect it in front of the DC load to make it a AC load.

[tszaboo]

Stepper motor + Rheostat. Nothing beats the simplicity.

[macboy]

Stepper motor + Rheostat. Nothing beats the simplicity.

Let's think about programmable loads for a second.
For DC, you generally would want your load to operate in one of three modes: constant resistance (current proportional to source voltage, load power is then proportional to V²), constant current (load power is then proportional to source V). and constant power (current proportional to 1/V).

Since it is fundamentally a DC load and "DC" is contant voltage by definition, the load typically operates in all modes as a constant current dump, with the current set-point programmed according to the mode as above, and only needing to be adjusted occasionally (whatever than means, maybe between few times per second, or more, or much much more, depending on your requirements). Usually, an analog control loop takes care of controlling the current to match the set point current. For a DC source with AC (ripple or noise) superimposed, the above modes still make sense, as long as the DC>>AC.

For true AC, the only sensible mode of operation is constant resistance. The resistance (not the current) is the 'programmed' parameter. (Think about what would happen near/at zero-crossing with a constant current... the load becomes a short-circuit and still fails to deliver the required current, not what you intend I think). Of course, an AC load could do a modified "constant RMS power" mode where the RMS value (as opposed to the instantaneous value) of the power would be constant, with the AC load adjusting its constant resistance parameter to meet the requested power. Or you could do "constant RMS current" again by calculating the required resistance for the present input voltage, and applying that resistance. But any AC load still fundamentally a constant resistance load.

So, given that, an AC load is going to be very different from a DC load. Certainly, it is not a DC load with a rectifier on the input, whether a diode bridge or an ideal rectifier network.

Obviously the easiest AC load is a resistor, or a variable resistor. A bank of power resistors with a creative array of (mechanical) relays to switch them into various configurations could do the job, but the adjustment will be rather granular. If you want fine adjustment, you need more complexity.

I would imagine that using a power amplifier output at the 'ground' side of the resistor network (instead of direct connection to ground) would provide you the ability to create an finely adjustable resistance by adjusting the output level of that amplifier. It would provide a sort of back-emf for the resistor that would reduce the current through it, effectively reducing the resistance of the AC load. A power amplifier is a four-quadrant device, capable of sinking or sourcing current at any output voltage, unlike the simple single FET/BJT of a DC load, which is a one-quadrant device (it can only sink current with positive input voltages).  A power op-amp-like audio amp chip like a LM3886, capable of a minimum 7A continuous output (within its SOA) could do this duty. The power amp would only be called upon to provide a few volts of adjustment to fine-tune the resistor bank. They could be paralleled for more current when necessary.

[zapta]

For true AC, the only sensible mode of operation is constant resistance. ...

First, it may also be useful to have constant impedance, not just resistance, in case you want to simulate non resistive loads. Second, theoretically it's possible to have also a constant (RMS) current and constant (RMS) voltage.  Something along these lines:

Let's say your source is 60 hz sinusoidal mains. You digitally lock phase on it and now at any given time you can compute the expected current for that phase and adjust your variable load (e.g. MOSFETS) accordingly. You end up with sinusoidal current of the target RMS regardless of input (RMS) voltage.

[mikerj]

This might be a possible application for a magnetic amplifier.

[macboy]

For true AC, the only sensible mode of operation is constant resistance. ...

First, it may also be useful to have constant impedance, not just resistance, in case you want to simulate non resistive loads.

Good point.

...Second, theoretically it's possible to have also a constant (RMS) current and constant (RMS) voltage.  Something along these lines:

Let's say your source is 60 hz sinusoidal mains. You digitally lock phase on it and now at any given time you can compute the expected current for that phase and adjust your variable load (e.g. MOSFETS) accordingly. You end up with sinusoidal current of the target RMS regardless of input (RMS) voltage.

I did mention that already 
But you must understand, that the load will be programmed to operate at a constant resistance(/impedance) according to RMS input voltage and desired RMS operating current (or power for that matter). Yes, the 'constant' resistance will need to be varied over time, as the RMS input voltage varies over time. This is fundamentally different to how a DC load typically operates.

[David Hess]

I would imagine that using a power amplifier output at the 'ground' side of the resistor network (instead of direct connection to ground) would provide you the ability to create an finely adjustable resistance by adjusting the output level of that amplifier. It would provide a sort of back-emf for the resistor that would reduce the current through it, effectively reducing the resistance of the AC load. ...

This is a common configuration when it is desirable to terminate a transmission line into a lower value resistance so the voltage compliance of the termination can be raised without raising the supply voltage.  For instance on the driving side, a source terminated driver normally requires a gain of 2 and twice the supply voltage.  By reducing the resistance and adding feedback, the output impedance of the driver is artificially raised to make up for the lower resistance of the source termination and the necessary gain and supply voltage are reduced.

[zapta]

I did mention that already 

Sorry about that ;-)

But you must understand, that the load will be programmed to operate at a constant resistance(/impedance) according to RMS input voltage and desired RMS operating current (or power for that matter). Yes, the 'constant' resistance will need to be varied over time, as the RMS input voltage varies over time. This is fundamentally different to how a DC load typically operates.

What I describe earlier allows the momentary current to be readjusted immediately, no need an over time RMS measurement, assuming you know the shape of the signal (e.g. sinusoidal) and you are phase locked to it. 

In a sense it is a generalization of what DC loads do, they assume some signal shape (flat), derive the expected momentary current (always the same) and adjust the actual current.

[Marco]

Let's say your source is 60 hz sinusoidal mains. You digitally lock phase on it and now at any given time you can compute the expected current for that phase and adjust your variable load (e.g. MOSFETS) accordingly. You end up with sinusoidal current of the target RMS regardless of input (RMS) voltage.

You will have to adjust that MOSFET really hard to get it to the negative resistance necessary for the part of the cycle where the current is flowing against the voltage ...

You need a powered amplifier to emulate non resistive loads.

[robrenz]

A bipolar operational power supply like a Kepko BOP or the Lambda BOSS (thread here) could possibly be a good start on a AC load that could be set up to follow and load an AC input signal.

[eneuro]

You could use 2 back-to-back MOSFETS as an AC variable resistor, but the drive & loop control would get somewhat tricky, more so as frequency increases

Do you mean AC mosfets switch not fully open?
This is low power version of such switch, so only 3 AC switches in pararell- 6 mosfets needed for this prototype 1inch2 PCBs each 

It is up to designer to choose right amount of pararell mosfets and its RDSONs to be able avoid power loses in mosfets body diodes-keep them not conducting even when switch is fully open at desired maximum power

[zapta]

You will have to adjust that MOSFET really hard to get it to the negative resistance necessary for the part of the cycle where the current is flowing against the voltage ...

You need a powered amplifier to emulate non resistive loads.

Yes, that's right. My description above was for resistive constant current or constant voltage AC loads.  If you want non resistive load with arbitrary phase, you may also need to source current to achieve the momentary target current or voltage.

[zapta]

A bipolar operational power supply like a Kepko BOP or the Lambda BOSS (thread here) could possibly be a good start on a AC load that could be set up to follow and load an AC input signal.

Dave mentioned once a 'four quadrant power supply'. Is it a similar concept?

[tszaboo]

Let's think about programmable loads for a second.
...

Excellent case for my one line! Anything which rectifies the AC will not be a good load, as diodes conduct or they dont. You cannot pull current from the rectified AC when your bulky capacitor's voltage is above the mains voltage. It needs to be a resistive load, or an inductive load, analog, something which works down to the diode drop level and below.
Now there are some really nice rheostats out there, that is why I was mentioning them. And it is not that hard to make a mechanical system from off the shelf parts, shafts, timing belts (hope that is the right term). And it will work nicely, the resolution can be even higher than a normal PWM or a 12 bit DAC would make. A  a 200 step/rev stepper is alone 7+ bit, even that the rheostat could make is a bit non linear.
I mean sure probably there is some twisted PWM circuit with active, separately powered PFC and huge inductors and transformers... Or you can just put some 60W incandescent light bulbs on relays and make a stepper controlled rheostat for the resolution. One is a weekend project, in the other getting rid of the stability problems could take forever.

A bipolar operational power supply like a Kepko BOP or the Lambda BOSS (thread here) could possibly be a good start on a AC load that could be set up to follow and load an AC input signal.

Dave mentioned once a 'four quadrant power supply'. Is it a similar concept?

Four quadrant is something which can source or sink current with positive or negative voltage

[robrenz]

A bipolar operational power supply like a Kepko BOP or the Lambda BOSS (thread here) could possibly be a good start on a AC load that could be set up to follow and load an AC input signal.

Dave mentioned once a 'four quadrant power supply'. Is it a similar concept?

Yes, but these are more like a mega op-amp that can follow an bipolar input up to about 20kHz depending on the V and A rating. Check the link out.

[skipjackrc4]

I use a 12A variac followed by a space heater.  Works well enough for my purposes.  There is a small amount of leakage inductance from the variac, of course, but this is usually small enough to be negligible at high currents.

[ivaylo]

A bipolar operational power supply like a Kepko BOP or the Lambda BOSS (thread here) could possibly be a good start on a AC load that could be set up to follow and load an AC input signal.

Dave mentioned once a 'four quadrant power supply'. Is it a similar concept?

Yes, but these are more like a mega op-amp that can follow an bipolar input up to about 20kHz depending on the V and A rating. Check the link out.

The Kepco will actually do this out of the box. The current modulating input is already inverting, you just feed it the same AC source you want to load (so positive voltage generates negative current and vice versa, i.e. load), scaled to the 10V input the Kepco takes. Probably through a transformer for galvanic isolation and all. Good idea!