Simple home made analog MPPT contoller

[Solarstefan]

The MS184060-2 you can find on the net. I since finished winding it and it measures 280uh,
actual test is ok but it is a bit lossier and gets warm even so it does not saturate.
Later on i will remove some inductance and see what happens.
Thanks for your input! Looking forward for you calcs.

 

[Warpspeed]

The -8 core will be the one to use of the Micrometals cores, it has quite low core loss, and more inductance per turn than the -2 core.

Now from the Micrometals data book they tell us that 15volt microseconds per turn will give a temperature rise if 15 Celsius at 50Khz with the next sized core down (T184-8). So 15 V/uS is pretty safe for core loss and heating.

Now assuming our input voltage is 35v and our output voltage is 14v and we run it at 20 Khz.
Voltage across the choke while on (35v - 14v = 21v) voltage across the choke while off 14v (ignoring the diode drop).
Total period 50uS. On time 20uS or 40% duty cycle. Off time 30uS or 60% duty cycle.
On volt microseconds 20uS x 21v = 420v/uS Off volt microseconds 30uS x 14v = 420v/uS.

So minimum turns we can use before the core starts to get a bit warm (>15C rise) maybe about 420/15 = 28 turns minimum.
We can run more turns than that, but not less without cooking the core.

Assume one solar panel at this stage, 35v 280 watts to keep it simple. Anticipated output 14v at 20 amps.

If we go for minimum 28 turns and 20 amps of dc, MMF works out to 54 Orsteads (using T200 sized core).
Permeability of -8 material drops to 90% of initial permeability, so we lose very little permeability with only 560 ampere turns.

Initial Al of T200-8 is 42.5nH/turn. 90% of that is 38.25nH/Turn. Multiply that by 28 x 28 x 38.25nH = 29.988uH
So 30uH for 28 turns at 20 amps.
Ripple current 14 amps peak to peak at 20 Khz.
Definitely higher than I would like, but more turns could be added to reduce that, we are way below any risk of dc saturation.

Its then a case of what sized wire will fit 28 turns onto that core ?
Magnetics suggest 11 AWG for 20 amps, which is 1.82mm diameter. That is a fairly high current density for only one strand, so its going to run warm.

See what will fit.
More parallel strands of 1.8 wire would be much more efficient, but putting on more turns on instead, would reduce the peak ripple current.

Stacking two cores would double the inductance and halve the ripple current, and that would be well worth doing, especially if more parallel strands of 1.8 wire could be used as well. 1.8mm wire will be fine for skin effect at 20 Khz.

 

[Warpspeed]

Simple answer for T200-8 core at 20Khz.
Minimum allowable turns 28, more will be much better if possible.
Minimum wire size 1.8mm, that is also about the maximum size due to skin effect and ease of winding.
Don't go thicker wire, use parallel strands of 1.8mm or thinner wire if possible.
Stacking two cores a good idea too.

 

[Warpspeed]

O/k found some data on the MS184060-2.
This is also now a Micrometals core as Micrometals appear to have taken over Arnold.

Its one of the newer much more efficient cores than the old powdered iron types.
Its now in the "Sendust" range, permeability quoted as being 60.
https://cdn.ozdisan.com/ETicaret_Dosya/409227_990653.pdf
https://datasheets.micrometals.com/MS-184060-2-DataSheet.pdf

This will be better than the T200-8 with more inductance per turn and even lower core loss.
At 20 amps permeability drops to about 70% with the same 28 turns.
But the initial Al is much greater 90nH/turn as opposed to 42.5nH /turn for the T200-8, so about double the inductance per turn.

You can put more turns onto that too, but the inductance per turn decreases a bit faster with increasing dc current and increasing turns than with the old powdered iron core.
But it will have lower core loss at higher frequencies than powdered iron. Its a more modern material, and should be a lot easier to obtain than the older T200-8

If you decide to use this MS184060-2 a switching frequency higher than 20Khz would reduce ripple current without horrific core losses.

 

[Solarstefan]

Thanks Warpspeed for your detailed reply, I have copied all new info and saved it on a separate notepad,
so I can study and digest the info in detail. Many thanks

 

[Solarstefan]

looking at your calc, and making progress on the T200-8 core you got 15V/us from a graph? I looked and can not find that.
To calc H you need a mean length?

I wound a T200-8 inductor a full single layer has 44turns which measures 80uh. 42.5nHx44x44=82uh.
Great that works and then 90% due DC bias. effective 74uH

For the second core, if I dont have the V/us, ??
This MS inductor has 40turns and measures 180uH

 

[Warpspeed]

looking at your calc, and making progress on the T200-8 core you got 15V/us from a graph? I looked and can not find that.
To calc H you need a mean length?

I have a very old Micrometals catalogue "Iron powder cores for power conversion and line filter applications" here.
Could not find it on the internet, its a pretty old catalogue.
You could order a copy, I presume its free ?

http://www.iec-international.com/micrometals/micrometals/catalog.html

In that on page 23 is a table of "volt microseconds" with various sized -8 cores that produces a 15C temperature rise.
That is where the 15 volt microseconds came from. Its actually for a T184 at 50Khz -8 material.
Not all core sizes are listed, but T184 is the closest size in the list.

In the same book on page 8, all the core physical sizes and materials are listed in sequence smallest to largest.
Physical dimensions are shown, outside diameter, inside diameter, core height, and also the "L" value in one column, which is the effective magnetic path length around the toroid. Its sort of a mean path length.
T200 core is shown as L = 13.0cm magnetic path length.

I wound a T200-8 inductor a full single layer has 44turns which measures 80uh. 42.5nHx44x44=82uh.
Great that works and then 90% due DC bias. effective 74uH

First you need to work out H, the magneto motive force in Orsteads for the turns and dc current yo plan to run. That will dc bias the core.
0.4 Pi times amps, times turns, divided by magnetic path length in cm.
For 44 turns and twenty amps, its 0.4 x 3.142 x 44 turns x 20 amps, all divided by 13.0 cm = 85 Orsteads.

Then you look at the magnetisation curve for -8 material. It has Orsteads on the X axis and % permeability on the Y axis.
Draw a line up from 85 Orsteads until it intersects the -8 curve. It looks like about 85% of initial permeability.
Magnetisation curve is here, top right hand curve for -8 material.
https://www.micrometals.com/products/materials/-8/

Now if you just test your 44 turns without any dc, it should read as having 42.5nH x 44 x 44 = 82uH.
But with 20 amps of steady dc, you will be down to only 85% of that, perhaps about 70uH

For the second core, if I dont have the V/us, ??
This MS inductor has 40turns and measures 180uH

All this is complicated enough, but what makes it all far worse is there is no standardisation of how the information is presented, as well as the mix of imperial and metric units. Core loss versus temperature rise is particularly difficult, because its non liner with frequency, and the physical geometry (surface area) effects the cooling. So milliwatts per cubic something or other is not very helpful in estimating temperature rise. Unfortunately milliwatts per cm cubed versus frequency and flux density is now becoming the standard way of giving core loss figures.

I used volt microseconds per turn, given for a specific core size, shape, and and material, as a much more direct way to judge safe temperature rise, which my thirty year old catalogue fortunately provided. Easy, simple and convenient.

The more complicated way is to regard core loss in terms of the ac voltage across the winding and the frequency.
Faraday's (transformer) equation links ac voltage, turns, frequency, and core cross section in terms of flux swing in the core.
Then you use the loss curves which are straight lines drawn on log/log scales for a particular material to give a core loss figure in mW per cm cubed.
That tells you the core loss, but not the resulting temperature rise.
Its the top left hand curve:
https://www.micrometals.com/products/materials/-8/

Its a much more cumbersome and round about way, and it still does not tell you how hot the core is going to get.

 

[Solarstefan]

Update, while still waiting for some decent sun, I have been experimenting on the bench with a DC supply and sofar with the right inductor i can easly achieve 25A out at around 83% efficiency. I can see from all the test that Vin should be lower for best efficiency and least heat. The stranded inductor performs best , but due the lower inductance, the frequency has to be higher, i am running at 60khz atm. You can see at lower loads that the loss due higher osc freq is present.