Re: Finally Heisenberg's is dead!

Twelve watts per output is pretty pathetic.

It's clearly a specific response to a specific customer. Vanity electronics.

The only negative impedance circuit I can recall posting here was a Philips-developed scheme for running DC motors at constant speed by driving them from a constant voltage through a negative resistance that compensated for the positive resistance of the windings. I think it was used in cheap tape recorders. The positive feedback that made it work was frequency limited to prevent oscillation. I think it depended on a single cheap op amp. I did identify where it came from when I posted about the circuit.

It was used in the EL3302 series, there is a full description of it in the Philips Technical Review. There were no op-amps, it just used discrete components with one special temperature-compensating resistor wound with copper wire.

I have used negative impedance at low frequency to overcome 'cogging' in a motor driving a parallel-tracking gramophone pickup carriage - and at audio frequencies to compensate for voltage drop in a long loudspeaker cable.

The audio installation was going to be in the Science Museum in London with ordinary twin-and-earth house wiring cable connecting an amplifier to a horn pressure driver of about 6 ohms impedance (A Western Electric

I arranged for the power to be supplied on the usual 'live' and 'neutral' conductors but specified that the earth wire had to be connected to the neutral at the loudspeaker end of the circuit. The returning voltage on the earth wire, which was equivalent to the volt drop on the 'neutral' conductor, was doubled and fed into the power amplifier stage to compensate for the total cable losses. Any change of resistance due to length or temperature was assumed to affect both power conductors equally.

I'm old enough to have used op amps made up of discrete transistors, mainly where off-the-shelf integrated circuit op amps couldn't do the job that needed to be done.

It did use up board space.

Inputs, not outputs. An arbitrary impedance is usually used as a dummy load, and a load absorbs power from the customer, an unlimited amount of energy over time.

We're mostly simulating relay coils and solenoids and torque motors, and eight channels at 12 watts is good. I'm designing a higher power

Each channel has a bidirectional class-D amp. Usually it accepts power from the customer and pushes it uphill, into the pair of ceramic power resistors. But we also simulate an inductor that has to return (fake) stored energy back to the customer for a while. We can't really store his energy because we've burned it up long ago; we fake it.

A real, even ideal, gyrator would have to store inductive energy in a cap, and the cap would not be a reasonable PCB component when simulating 10 henries.

The concept is fairly simple. The actual circuit isn't. The real challenge is making it wideband but unconditionally stable. And stuffing eight channels into the available area.

Things like this keep engineering interesting. Learning about real motors and things. Then circuits, thermals, magnetics, PCB design, all tangled.

You have some objection to customers?

Yes. They have been building big boxes full of real resistors and real inductors, faking specific loads. But the concept looks fairly general. I don't know of anyone who sells programmable R+L loads. Well, a few people will build you a box full of relays and caps and inductors.

OK, say something nasty now.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Obviously.

Most circuit design has lots of tedious constraints, which you can't boast about getting around because they are tedious.

Obviously. If you learned a bit more about magnetics you might be able to get into designing your own transformers, rather than insisting on using parts that you can buy of the shelf.

None. Most of yours seem to have trouble understanding their problems clearly enough to be able to go out and buy off the shelf solutions.

Faking specific loads with real components is a much safer option.

The sort of people who want a neat and compact packaged fake load have to have more money than sense. Great customers if you can find them, on the principle that you should never give a sucker an even break.

Buying 90 cent surface-mount transformers make sense. Right now we are waiting for another batch of kapton things from JLC, windings for fast high-voltage planar transmission-line transformers.

Avtech Electrosystems abruptly went out of business and we've had some enquiries about high-voltage pulsers.

We like to do stuff that has no competition. That's interesting and avoids bidding wars.

Real inductors and resistors? That has big problems.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

If they do what you want. Winding your own - or getting your own windings printed - gives you a lot more flexibility.

But does restrict you to niche markets, which aren't big.

If you don't know enough about transformers to design your own, you may well run into problems with inductors. Resistors are rarely problematic

- they do get hot, which does have to be dealt with - but everybody knows how to deal with that.

Big enough, but more importantly fun. Yes, jet engines are a niche market.

It's easy to buy a hunk of iron and some power resistors and stuff it all into a big rackmount box with some fans. Just apply time and money.

We are developing the synthesized RL loads precisely because a biggish organization got tired of doing that a zillion different ways. It turns out that rack space is very valuable to them too.

The problem with real power inductors is making them programmable. One winds up with nightmares of tapped inductors switched in series or parallel with a mess of relays. The electrical issues are ghastly. Play with it some.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

The old way to obtain variable inductance that could be changed under load would be a variometer, a rotable coil inside a fixed coil, wired in series.

What kind of power are you dealing with?

Jeroen Belleman

One board is eight channels, each programmable 10mH to 10H. It simulates small solenoids and relays and torque motors. The customer only needs to go to 2H, but it's easy to go up to 10.

I'm designing a higher power version, 4 channels with a CPU cooler. A good copper cooler bolted onto a board can dump 150 or maybe 200 watts, depending on how hot you allow the parts to get.

Torque motors are interesting. Many are 2-part, one bit bolted to a rotating shaft and a fixed coil assembly outside that. Sort of a

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Bean counters do get excited about rack space. They have to pay rent on the space to accommodate the racks.

Curious. I've used linear differential transformers as position sensors. They depend on moving the core inside a complex winding to vary the coupling. You can move it continuously. There are angle sensors that rotate the core to achieve the same effect.

I've not seen the approach used to make variable inductors. Inductors are cranky enough without trying to make them variable, so there wouldn't be a mass market for the sort of off-the-shelf parts that you'd like to buy.

Some of these test systems go into control rooms or out on a factory floor, where real estate is valuable. Some go into flying test beds, even more valuable.

Saving, say, 2U can mean the difference between two racks and three.

GR used to make motorized Variacs, before electronics was invented I guess.

I guess they won't be on the shelf at Walgreens.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

I don't suppose you're actually switching tapped inductors, are you? I'd rather take a power amplifier driven by a DAC and with output current read through an ADC, and with some FPGA code in between to simulate the actual impedance. Switching real inductors is nasty, as you say. You can't open the circuit because it would spark, and you can't short sections because the energy stored in them would vanish.

Talking of relays, the coil current on pull-in is interesting: The current actually briefly drops. And when you switch it off and the armature moves to open, the current briefly rises. Will your device simulate that?

The torque motors I'm familiar with were just DC motors driven with a constant current. They were used as tape tensioners in a computer tape drive. Pretty low-power stuff, is true.

Jeroen Belleman

That's what we're doing: digitize the input voltage. Go into an FPGA and apply that to a math model and see what the current should be. Make the PWM to drive a class-D amp to get that current.

Easy. Just make it fast and always stable.

We could, but we currently don't. The motion of an armature vs time could get complex.

This is a typical frameless brushless torque motor:

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

A Variac is a tapped auto-transformer, and only incidentally an inductor, and the ones I ran into depended on a moving brush to adjust the position of the tap.

That would work to make a variable inductor, but the brushes don't last.

Moving the core would work quite a bit longer.

For small changes in inductance you can buy gapped RM cores with a central hole, and adjust the inductance of the wound part by screwing a ferrite adjustor into the gap. You get a smooth and continuous adjustment, but it isn't fast and the mechanism would wear out if you used it often. Ferro-fluids might last longer.

Newark would be more the kind of broad-line electronics distributor that you'd prefer to rely on.

There were high-power ones o the cable-laying ship "Monarch", for keeping the cable tension constant. Interestingly they were run from a constant-current main that fed them all in series (and you shorted each motor to switch it off.