MEMM

[PaulLowrance]

Hi kingrs,

Yes, LTSpice is amazing. It has shocked me several times so far in its accuracy. Too bad it can't predict Naudin's metglas MEG.   Make sure you get the latest update by clicked Tools -> Sync Release.  One thing I love about ltspice is FFT spectrum displays. On a graph just right click and select FFT. You'll see a window full of options. Default is usually good. Sometimes you'll want to bump up the data samples. Click OK then you'll see the FT graph. The default display is DB vs. Log.  Just left click (not right click) on the DB or KHz numbers to change this. I like linear in Cartesian display. This will give you the typical Amplitude & Phase.  Another great option is testing for component noise. I used it to accurately design a 3 stage FET amp that had less than 1nV/Rt noise. In noise mode it will tell graph the voltage noises inherent in parts. LTSpice knows about all the different types of noises from thermal noise to shot. Lots of goodies.

I added an extra 1.6Ohm resistor to the circuit in series with the input coils and it seems to make little difference to the current, which seems odd.

Actually I already placed the 1.6 R in the inductor. Just right click on the L1 or L2 and you'll see it has 1.6 ohms for the series resistance field. Same goes for L3 & L4 except it's 37 ohms.  So you don't need to add another 1.6 R.  Also you'll see all the inductors have parallel R and parallel capacitance.

The inductors are coupled with the Kx commands. I could have just did K1 L1 L2 L3 L4 0.99 for example, but I broke them up to give each coupling a unique coefficient so as to represent the MEG core since the two secondary coils are appreciably separated from the actuators.

This is going to be interesting to see the results.  I think Naudin's 100K is closer to 35K, but people are saying it varies in resistance.

BTW, I tried replacing BUZ11 with IRF540 and what a difference. The irf540's are a lot slower than buz11's so the sim is a lot faster.

Regards,
Paul Lowrance

[MeggerMan]

Hi Paul,

The BUZ11 is probably older than Methuselah and is very slow response wise.

The IRF540 is a lot quicker from start to end it takes just 120ns where the BUZ11 takes a painful 570ns, time to make a cup of tea, walk the dog, have a nap and its still just firing up.

So you would be far better off using a more up to date MOSFET like the IRF540, and of course there are better and faster mosfets than even the IRF540.

http://www.irf.com/product-info/datasheets/data/irf540n.pdf

rather than the old BUZ11

http://www.farnell.com/datasheets/45278.pdf


The IRFU3707 start to end is just over 100ns but is only OK for 30V so it would need protection from that back  emf.
I have been thinking it may be worth protecting the device by straping a zenner diode of say 10 to 20V across the input coil.

Regards

Rob

[PaulLowrance]

Hi Rob,

My datasheets show the BUZ11 turns on twice as fast as the IRF540-- 30ns vs. 60ns.  Although the IRF540 turns off a lot faster-- 180ns vs. 50ns.  I'm thinking the turn on speed is very important in the MEG.

So don't discount those BUZ11's, lol.

I've been doing some FEMM simulations on the MEG. Due to the PM's the core field drastically drops at around 15 volts. I made an exact copy of the Metglas 320 core, 100 turns on actuator (MEGv21). Naudin's FEMM simulations are incorrect because he assumes the core's permeability is the same at 30KHz vs. DC. That Metglas core at 30 KHz has roughly the same permeability as FEMM's iron material. Naudin's power charts show a larger rise at around 20 to 25 volts input, which seems to agree with my FEMM simulations. Also anything after 25 volts does not help as much, which is probably why his power chart shows a huge decrease in efficiency at 30 volts.

I did sims on both Neo and ceramic magnets. The neo's will probably work well at higher input voltages and the ceramics work better at lower input voltages. This is because the neo's introduces a stronger traverse field near where the magnets and core touch. Traverse fields can hurt permeability. Ceramics might be a best first choice since they work better at lower input voltages and that's what Naudin used.

Regards,
Paul Lowrance

[PaulLowrance]

Actually I have it backwards. It's the turn off time that seems to be more important since the PM reverses the process. Still lets not discount the BUZ11 since that's what Naudin used and it is twice as fast turning on that the IRF540.

Paul Lowrance

[MeggerMan]

Hi Paul,
Thats interesting, looks like every one and his dog makes a different version on the BUZ11.
RS have discontinued them all 4 of them.
They all appear to have different properties, and I bet if you bought two of the same and tested them they would show different speeds.
Price of the part seems to vary between manufacturers and it looks like its being discontinued all over the place.

Can you post your femm files.

Did you put in the material for the Metglas core (B/H curve), I had to do this the other day for a ferrite and it was not easy?

Core should arrive tomorrow and I will post a picture of it so you can see what it looks like.
I still need to order the bits for the pulse circuit.  Are you getting a TL494 chip?
I noticed that the TL494IN is the one to go for as it works at -40c to +85c, (brrrrrr to damm thats hot!).

I need to check my scope still works and look at sorting out the interface to the PC for it.
I'm tempted to get a Stingray usb scope but the input burns out at +/- 50v and the potential offset probes cost more than the scope.


Regards

Rob