TINKERERS POWER SUPPLY

[porkluvr]

Whew, Finally done. As I was making the parts list I kept coming across little things to tweak. Big ones, too.

This power board is designed to use an 11V Li-Po or similar battery (that could have an output ranging from 8.5~14.4V)
and develop a constant +15V to power a PI detector having an N-channel FET and a DD coil. There are also three +5V outputs and one -5V output.
This board can also be used to power a monocoil detector that uses a P-channel FET.

I've tried to keep the schematic and the parts list consistent but there came a time when I had to break the two apart
and the parts list was updated manually even as I was making changes in the project.

There is an explanation in the zipped folder so I won't repeat everything here.

You can open the schematic and board in Eagle 4.16 or anything later (freeware or standard versions will open it).
It was made with a legal freeware version so there should be no DRM issue to mess things up.

A few days into this design I realized that I could use it in my upcoming project because I'll be using a DD coil.

The PCB layout shown here looks incomplete because the ground polygon is not poured. To complete the ground path you
need to open the schematic and board. In the board file do a "Tools > Ratsnest". Turn off all layers except for Bottom, Pads, and Vias and
Dimension (16, 17, 18, 20).
Then, "File > Export > Image" allows you to output the board file either to the clipboard, or to a file folder.
Save it to a file folder if you want to print it out. You have different file formats to choose from that way.
Use monochrome if you want to etch. Be sure to use a high enough resolution, the default 150 is pretty useless.

The polygon-poured picture is just too large to show here, and changing the resolution to reduce the size would make it
unusable - so there is not much point in showing it here.

[porkluvr]

L2, L3 toroid calculation mistake

I made a mistake when I calculated the number of turns to be used when using 26 mix powdered iron toroid cores for L2 and L3. I think the L1 calculation (using a ferrite core) was OK, but L2, L3 are bad wrong. A 26 core (or the Fair-Rite core listed) should not be used because too many windings of would be needed.

I gave myself a crash course in using toroids. I found only a mish-mash of incomplete and outdated material on the Amidon website. That was dissapointing because I always assumed Amidon is sort the Mac-Daddy of toroid cores. (Maybe that title belongs to Fair-Rite. ) The Amidon website's instructional data is purely stone-age. They have separate pdf files of scanned pages from one of their old databooks, probably from the 1980's from the look of it. Very difficult to browse. I tried to fill in the gaps from other sources that were often heads above Amidon's site in presentation, but it was only tonight that I think I grasped the view of the big picture.
Toroid cores have an AL, or inductance index associated with them which varies with their size and composition. When calculating the number of turns needed for a given core from its AL specification, there are three different formulas you might encounter. There are two interchangable formulas for ferrite cores (one relates to nH and one relates to mH) and a third formula for iron core toroids, relating to uH.

The ferrite and powdered iron formulas are NOT interchangable. If you use one of the ferrite AL formulas with AL data for an iron powder core you will get a wrong answer, as I did.

I overlooked the fact that the odd formula is the one that I was supposed to use when calculating the number of turns on the L2 and L3 powdered iron cores. L1 uses a ferrite core and I think I got that calculation right.

I listed some factory wound cores on the schematic page and in the parts list that should be suitable for L2 and L3, but those two hand-wound recipes will come up short.

We could use FT-63A-77 core for all three inductors (I calculate 9 turns for 105uH) but that core is over-size for what we need and it's kind of expensive. A smaller 77 mix core (say, FT-50-77) should work OK because it won't need to handle the same stresses as the L1 coil. I should have planned to use a 77 mix core at the beginning and I wouldn't need to be writing this.

FT-50-77 with 10 turns #18 or #20 AWG should fit and make 100uH. I think that should work for L2 and L3. Let me sit on that for a few days...

Here is a very nice program for calculating toroids but it does not have a complete database, and in fact does not calculate FT-63A cores. http://www.dl5swb.de/html/mini_ring_core_calculator.htm

I hope this hasn't an inconvenience to anybody. Hey, I have some stuff on order for this beast. I'll order some cores from Amidon tomorrow. I'm glad I didn't do it earlier. When they arrive I can build and test this power board, but as yet it is untested.

[Tinkerer]

Thanks porkluvr, you are really putting a lot of work into this power supply design. I am sure the results will be accordingly.

Many thanks for your great help.

Tinkerer

[porkluvr]

Well, in all honesty I came across my math error as I was trying to calculate inductor recipes for my other project. I had a sneaking suspicion but I couldn't put my finger on it.

I found on the Amidon website where there are some up-to-date booklets available, more up-to-date than the (scanned from 1980 databook?) downloadable PDF pages I found so appalling. Man, they need to get rid of that old junk. It makes a bad impression.

Ahhhh, but the booklets are available only through mail order.

That's OK. I like getting things in the mail. It makes me feel important.

(I hope I don't have to pay postage.)

[porkluvr]

Owww

I'm here to make another red-faced confession. The ferrite cores I listed for the L1 hand-wound recipe are not suitable.

And, I read the wrong line in the Fair-Rite catelog. 5977004901 would be Fair-Rite's closest equivalent to the Amidon FT-63A-77. But that is a moot point anyway because a flux density calculation indicates that either one would likely be destroyed working with the LT1171.

I never took physics and flux density is a muddy concept to me. I'll try to figure it out before something blows up.

LT1172 would work well with a 50uH coil instead of the 27uH shown for the LT1171, and that could use about two additional turn on the FT-63-77 core, lowering flux density a little, but this design is still in the brier patch.

What is needed is a ferrite mix with a substantially lower inductance index (AL) using more turns of wire, or else to use a much larger core.

This is serious, and I appologize. It wasn't purely wishful thinking that brought this about. Also, bad math.

Let's see... larger core and/or use the 43 mix material... and/or use the LT1172 with a 47uH inductor.(You gotta do what you've gotta do.) 43 mix might not be as suitable for a 100kHz switching regulator as the 77 mix, but the alternative may be to use a core that is over 1" diameter. Ouch. I do not relish using a 1" diameter toroid core.

I've gotta cancel my order with Amidon because I ordered the wrong parts last night.

I thought I had the Bmax figure calculated and showing a safety margin but I calculated again tonight and the result is scary. I'm new at this toroid stuff and I guess it shows.

LT1172 is probably a better IC for this power supply anyway, but somehow these bits and pieces aren't coming together the way I'd like them.

[Qiaozhi]

I'm here to make another red-faced confession. The ferrite cores I listed for the L1 hand-wound recipe are not suitable.

And, I read the wrong line in the Fair-Rite catelog. 5977004901 would be Fair-Rite's closest equivalent to the Amidon FT-63A-77. But that is a moot point anyway because a flux density calculation indicates that either one would likely be destroyed working with the LT1171.

I never took physics and flux density is a muddy concept to me. I'll try to figure it out before something blows up.

LT1172 would work well with a 50uH coil instead of the 27uH shown for the LT1171, and that could use about two additional turn on the FT-63-77 core, lowering flux density a little, but this design is still in the brier patch.

What is needed is a ferrite mix with a substantially lower inductance index (AL) using more turns of wire, or else to use a much larger core.

This is serious, and I appologize. It wasn't purely wishful thinking that brought this about. Also, bad math.

Let's see... larger core and/or use the 43 mix material... and/or use the LT1172 with a 47uH inductor.(You gotta do what you've gotta do.) 43 mix might not be as suitable for a 100kHz switching regulator as the 77 mix, but the alternative may be to use a core that is over 1" diameter. Ouch. I do not relish using a 1" diameter toroid core.

I've gotta cancel my order with Amidon because I ordered the wrong parts last night.

I thought I had the Bmax figure calculated and showing a safety margin but I calculated again tonight and the result is scary. I'm new at this toroid stuff and I guess it shows.

LT1172 is probably a better IC for this power supply anyway, but somehow these bits and pieces aren't coming together the way I'd like them.

I hate to see a grown man cry.

Try this -> http://www.intusoft.com/Mag.htm
You can download a demo version that might be good enough for your needs.

[porkluvr]

Thanks for the link, Qiaozhi. I looked at the demo. The commercial version would be immaculate but I am afraid the demo is really not geared towards what I need right now. I'm not building a transformer coupled flyback design, and that's what the demo calculates. Maybe when I get some time I can glean some good information from the demo - and in fact the instruction manual has some good educational material. But it is not direct and to the point at a time when I need some very pointed information.

In my ignorance I thought that 77 material would be the perfect choice -and it is, but only if we are willing to use relatively large core sizes. I'll try and compare the relative merits of using some of the lower inductance index materials, looking towards minimizing the core size but still do the job efficiently. It's not such a simple process as I first imagined.

My PC desktop gets pretty cluttered trying to juggle the information I'm trying to filter, but I have a hardcopy of the Fair-Rite catalog and should have some Amidon information in a few days. Unfortunately, their products mirror each other to a high degree and what I would like is more variety.

Here is an exerpt from Linear technologies' AN44 where they present the core selection process in a new light. I'll work through the design process presented.
The Mini Ring Core program is helpful but it is nowhere near complete.

[Qiaozhi]

Another useful reference is "SMPS Simulation with SPICE 3" by Steven M. Sandler.

It's published by McGraw-Hill. ISBN 0-07-913227-8

The book was published in 1997, and seems to be selling for an inflated (stupid) price on Amazon. However, Steven Sandler apparently has a newer (cheaper) version available, dated 2006, and I can only assume that it's an updated version of the above. There is no price label on my copy, so I've no idea how much it cost originally.

[porkluvr]

I found some likely good choices for the toroid inductor cores. My first estimates were pretty much just plain wrong. 77 mix is absolutely not a good choice for the L1 core because the very high inductance index of that material would result in a very low turns count, and flux density would be excessively high. I don't understand and so can't explain the math behind calculating flux density, so don't ask. If anybody would like to give an explanation or if anybody sees some error in my calculations, please, please, feel free.

Unless a 77 material core was excessively large, if used for L1 it would likely get hot. I suppose heat is the result of excessive flux density. (Have I mentioned that I am stoopid when it comes to magnetics?)

So, we don't want to make L1 out of a 77 ferrite toroid, unless we want L1 to be a big one - and that would go against my better sense. All two of them. Now, the L2 and L3 toroids could probably be made out of 26 material, but I have found something better. Besides, I am darn irritated about the different conventions that are used, and how using the wrong equation for the wrong material will give you the wrong answer. I need to reconcile these different formulae to the rules of
algebra, but I don't feel like revisiting my math right now - not without some further tutelage. You would think that the toroid AL formulae are all interchangable but from my experience that just aren't so. Maybe I keyed the numbers into my calculator in the wrong order, but doh, (!) there it was. (Waaah!) On that note I would like to introduce a third category of core material, called "powder material".

I first learned about powder cores from examining the Linear Technologies AN44 which listed some various materials that might be used for the LT1070 familty of SMPS controllers. LT107x are almost exactly the same as the LT117x family that I am using, except for their using a 40kHz switching frequency insead of 100kHz. There is definately an overlap in the technologies involved in core materials so far as ferrite, powder, and powdered iron materials are concerned, but I will not try to explain the similarities or differences. Just like I won't try to teach algebra, for which I am supremely challenged at times, or grammer and speling.

For the LT1172 SMPS energy storage inductor (L1) we need something that can handle a fairly wide voltage swing and at the same time pass somewhere around 1/2 amp DC, and do so at a 100kHz switching frequency. After doing some waffling, I have settled upon the LT1172 instead of the LT1171 as the IC of choice. LT1171 would do well at
higher power levels, but at 4-1/2 watts or less I believe that LT1172 is definately the better choice.

Molypermalloy powder (MPP) is my first choice for the L1 core material. High-Flux and Sendust are also suitable and better in some respects, but both of those have higher core losses. MPP is heavier than the other two, so if we were going to need a very large toroid core then MPP might lose its #1 spot. If you want to read up on these 'powder' materials, then mag-inc.com is a good place to go. If you want to purchase them in small quantities, then go to cws-bytemark.com. Surplus Sales of Nebraska has some cores suitable for L2 and L3, but not for L1 (unless you want BIG, or something that might glow). For L1 toroid core you should go to CWS-bytemark.

Tony (Tinkerer) asked me to work on a power supply for his project. After realizing that I also could use a similar power supply I was only too happy to do so. The attached PDF file is a compilation of the better choices that I have found along the way. I have yet to actually own any of the listed products, and so the power supply remains unbuilt and untested, but I thought it best to post some update here.

The Bmax number in my table is not set in stone, but for reliability the lower the number the better. 500gauss is a reasonable starting point for use at 100kHz SO FAR AS I KNOW. Hi-Flux and Sendust might be safe at a substantially higher number, and they cost less, but my primary goal was efficiency so MPP material is suggested .
So, with that, I'm toast for tonight.

[porkluvr]

power supply update

I've sent Tinkerer a rough draft of a new schematic and PCB layout using a more suitable core for L1. The new design uses a CM270060 toroid core that is about 1.1" in diameter. A smaller core could possibly be used but the ones I looked at would make it difficult to fit enough turns of 18AWG as I want. #19 wire might be OK and fit on a smaller core but for a prototype I would rather start out a little too big and maybe work down rather than to try and do it the other way around. Skin effect is in play at the LT1172 switching frequency (100kHz) so larger wire might be a waste, and at the same time smaller gauge wire exhibits substantially higher resistance than at DC.

There is only one good source for the CM270060 that I know of (CWS Bytemark) but the other inductors are much less critical. In fact, solenoid coils might be effectively used for L2 and L3, and take up considerably less room than toroids but the layout has space for toroids. There is room for T65 or T68 Toroid cores for L2 and L3 but I'm thinking that FT50 size ferrite could probably work. I'll need to research that. L2 and L3 need to handle a maximum of about 300mA (we're going loaded for bear) but there is essentially no AC voltage on them. L1 handles the supply current, PLUS, up to 12VAC, so it needs some bulk.

I'll look at CWS's stock of ferrite parts and try to pinpoint some "best" cores for L2 and L3, but I've already ordered Magnetics 55118 cores (or something similar) from Surplus Sales of Nebraska, and that's what I'll use. The closest replacement from CWS is the CM172125. Although the CM part has a larger outside diameter, it has both lower inductance index (meaning less inductance per turn), and a smaller "window area" meaning fewer turns will fit. Either one should should be good enough, but there is probably something better. It's just a matter of looking (maybe tomorrow).

I figure that to wind about 68uH with either of the last two cores would require using 21AWG, and that is as small as I want to go. 20AWG would make me feel better but I may only be able to obtain about 50uH. Either way, it's not worth losing any sleep over.

l need to take a few days to clean up the layout and at the same time try and weed out any glaring design errors. The power supply's design is not particularly complicated, but the devil IS in the details. There is added a battery low LED warning based on 1/2 LM393.

The board is a single sided design so there is no bottom ground plane - I can only hope that any radiated EMI will be of low enough power as to not mess up the processor or analog circuits.

If anybody can point me in the direction of a dealer who will supply small quantities of Magnetics powder cores (MPP material, or maybe High-Flux) I would surely appreciate it. The CWS product line is good, but they stock only a limited variety. Another source would be welcome.

edit: more on L2 and L3 cores. I have been concentrating so much on problems with L1 and how the more common ferrite materials (75, 77, etc.) are unsuitable, that - I have almost forgotten how those ferrites ARE probably suitable for L2 and L3. Ferrites toroids are available from Amidon, but with their extremely high inductance index (compared to "powder") they are a world apart from powder materials like MPP,and High Flux. It's late. I'll look into this TOMORROW.

[porkluvr]

Tinkerer is making progress on his main board and I have been sitting on this for long enough.

I worked up a through-hole supply based on the LT1172 that should be OK. It's main problem is that the toroid core that eventually found its way into the solution for my mathematical analysis is kind of big. 30mm in diameter and weighing about 40 grams (after winding) is not small. I have not yet built the power supply because I started working on an SMD version update and I'll wait until I finalize that design before I order the rest of the parts for both designs.

The next power supply is based on LT1372 which bumps the switching frequency from 100kHz to 500kHz. What I am finding out that although the toroid core can be smaller, less expensive, and easier to wind - some of the capacitors need to be considered more carefully. If I had a bunch of money to throw at the problem then I could maybe just buy tens of the highest quality parts available, and be done with it. But that is not the case. I'm taking a little longer to try and avoid expensive or over-sized components where they are not necessary. It's a headache because they all look so much alike.

I'll find time to post the LT1172 circuit and PCB, probably Friday. It hasn't been built and tested, and to be honest the math I used to determine the toroid core is suspect. There are different schools of thought when it comes to allowable flux density for toroids. For the SMD version I am working on, to select a toroid core I used a "Bmax" number that is one-half of what I am using for the thru-hole version, so in effect there is twice the safety margin in the newer design. Yet, the newer design uses a core that is approximately 1/3 the mass of the first one. Because of the difference in operating frequency I had to draw a chart and interpolate from Bmax=250gauss at 100kHz (and 30G at 28Mhz) and arrive at Bmax=190G (approximately) at 500kHz.

Instead of 250G, the allowable Bmax number I used for the LT1172 core was 500G at 100kHz. I do not remember where I got the number 500 from, because Amidon tech notes suggest that Bmax= 250G at 100kHz. This number is not set in stone and different core materials can handle more flux density than others, but I cannot go into detail at this time. At first I thought it was a bad thing that toroid core sizes jumped appreciably in size in my range of interest; so that I had to choose between either a core that seemed too small, or one that seemed too large. I chose the larger size. Looking back, if I had chosen the smaller core, and considering that I started with one-half the safety margin that Amidon suggests, the first design could run into trouble with core over-saturation at higher power levels.

As it is, I am not too worried about possibly having under-specified the first toroid because I made Bmax calculations to cover a wide range of input voltages and input current combinations, and excepting the fact that the 'target' number I used may have been inappropriate, there was some safety margin afforded by the large core size. And besides - if a larger core than what I selected is required, then I am not too interested in that solution anyway. (Sour grapes, anyone? )

I'll need to look over the older design and will make a post most likely on Friday, depending on how many 'adjustments' I see as necessary (but umh really, I don't want to have to adjust anything).

I am new at choosing magnetics for use in power supplies, so over the course of this thread if anybody has correction or suggestions to offer I will not take offense if you do.

The table and data used to calculate median points in the chart below, was taken from here:
http://users.catchnet.com.au/~rjandusimports/index.html
Although not expressly stated in the text, I believe the data is originally from the Amidon Corp.

[porkluvr]

An Eagle 4.16 PCB layout, schematic, and comprehensive parts list are in the zip file. This is my latest thru-hole +15V power supply, using the LT1172.

There are only a handful of toroid core manufacturers on the planet, and still fewer have what is needed for L1, and will sell their products in small quantities.
CWS Bytemark does have a suitable product line that is available in small quantities.
I chose their CM270060 core which is approzimately 1.1" outside diameter and .4" in height before adding wire.
It is made from MPP 60u material. With 8V~12.5V input and 15V 300mA output, L1 has a big job to do.
My original estimation of the toroid size that would be required (I guessed 0.68" diameter) was rediculous.
I still don't know a lot about magnetics design but I knew next to nothing about it when I made that estimation.

As I stated in my previous post, I may have used a number for maximum flux density that does not allow a wide safety margin at the 100kHz switching frequency.
Fortunately, my calculations of generated flux density did not max out at this value (500gauss), so I "think" the design should be OK. (Famous last words?)
Some manufacturers suggest 250gauss to be a safe maximum value to use. Maybe not justification, but, that number is only a guideline - the
correctness of which depends on the specific core material in question, and, the manufacturer where I got that number from does not make MPP60 anyway.

I honestly do not know how I decided to use 500G, but as I stated the other day, "if a larger core is needed, then I don't want this design anyway". Ahem.

I will eventually build and test this power supply design, so you the reader may want to wait to see what happens to me before you build this device for yourself.

After spending some more time researching the magnetics morass today, I have given trying to remember where and how I determined that 500gauss
should be the maximum flux density for L1. There are various charts available regarding core loss vs frequency for specific toroid core materials
and I will try and make some better sense out of what is available. I'll have to figure it out later.

If L1 gets hot the most sure fix would be to use a larger size core or else glue two smaller cores together. Neither option seems very attractive
so I will keep my fingers crossed and see what happens.

Most of the ground connections on the PCB are made through Eagle's polygon pour feature, which was not turned on for the jpg below.
The picture's resolution would need to be higher than is practical to be posted for this forum.

The PCB was designed with Eagle 416 lite (freeeware) so any version after that should be able to open it for editing.
All you need to do to complete the grounds is go to Tools > Ratsnest on the Eagle PCB editor menu bar.

The trip voltage range of the battery low flasher can be modified if needed by changing the voltage sense circuit resistors, but the battery input
for this power supply should not exceed the 8V~12.5V range. With the original components the trip point is adjustable between 8V and 9.3V.

I tried to list a variety of options for the L2 filter choke. The exact value of L2 is not particularly critical but using a choke there is
probably a good idea even though there is a voltage regulator following it. You may note some inconsistency in my suggested recipes; that is
a reflection of the variety that is possible. There are many more possibilities for L2 than what I have listed.

There is some leeway for L3, but that choke should have a value of around 75~100uH. If your L3 inductance is lower, then C3 should be increased
to keep the transmitter load from affecting the VR1 output. I wanted to smooth the input to the various +5 and -5V regulators, but a few tens
of millivolts of input ripple should not make much difference on those 5V outputs.

That's all for now.

I had to reload my XP operating system today because my primary system crashed hard and my back-up system had slowed to a crawl - so bad that
I could not use it to make this post. It was so much fun, and I will probably do it again tomorrow to get it "right". Wish me luck.
Now, I have to install Acrobat Reader... .

Bye

[porkluvr]

I took another look at the power connectors and may have made a mistake on the parts list.
I would not suggest that PWR1057 from BG micro be used. BG does not give a complete description of that part and it does not look like what I have, so don't try and use it.

I thought I had purchased some, but that is not what I have. Use the PWR 1249 or else the WeidmŰller part from Digikey.

PWR1249, at 9mm, is slightly wider than the WeidmŰller 1760510000, and that is what I used for the PCB layout.

[Mustafa]

Hi,
What about LM2575 or LM2675.
It has wide input voltage range: 8V to 40V.
External shutdown pin allows deactivation during critical sampling processes.
Below schematic is from http://www.mino-elektronik.de/12VDC/...h_aus12vdc.htm "Multiple power supply from 12V-DC"
+18V@50mA may be a problem for high TX pulse width.

[Tinkerer]

Mustafa,

thanks for the contribution.
The power supply with the multiple output is interesting. It will need some tweaking, change the 8V outputs to 5V is easy enough.
The 18V to 15, or leave it at 18, but it will need about 150 to 300mA, depending how much pulse power one wants, or how deep the penetration should be.
The shutdown pin is a good idea, I wonder how much time the power supply takes to build up again?

Tinkerer