Transformer for a DC-DC converter

[jdbakker]

Hi all,

I need a low-noise DC-DC converter for my portable HDD recorder, to get +/-12V, +/-24V and P48 for the analog bits. Some searching has turned up the LT1533, a push-pull converter with a shaped drive waveform to reduce switching noise to low levels.

So now I need a transformer. This is my first transformer-based switcher, and the data sheet and app note both refer to a non-documented Coiltronics-part (with no hints on finding a replacement other than 'ask the manufacturer').

I would prefer to use a ready-made part available from Digi-Key or the like. What I'm looking for is a transformer with a 1:2 ratio, with center taps on both primary and secondary. Power would be ~20W, to be on the safe side. Switching frequency is either 100kHz or 200kHz (depending on whether I choose to support 192k sampling. The best I have found so far are the JW Miller PM600/610/620 series and the TDK PCA series, which all consist of six-winding transformers (1:1:1:1:1:1). However, all these only specify winding inductance, how do I translate that to switching frequency ?

Plan B would be to wind my own. Any suggestions for (toroid) cores for 20W/200kHz ? Could I just use an existing toroid or common-mode choke and add a secondary winding ?

I don't care about RoHS, since this is DIY use only. I do care about EOL status, since I'd like others to be able to re-build the design. I also care about affordable shipping to Europe (which nowadays sadly excludes Mouser).

Any suggestions ?

JDB.

 

[Svart]

How about moving to an inductorless converter or one that uses an inductor instead of a transformer?

The Transformer you are looking for is a coiltronics ctx02-13664 which somewhat fits the equation given in the document as well as being listed in a table of acceptable transformers.

There is another equation that should help you in the datasheet, one that give you the ideal turn ratio for the given input/output voltages and current draw.

I assume using parallel converters is out of the question due to the space limitations of the project?

The switching frequency is only one aspect to look at, the higher you go, the smaller the magnetics can be and still be efficient. What you want to do is choose a magnetic and then find your optimal switch frequency for the saturation and then choose your optimal pulse width. Simulations are damn handy for this as you can measure the temp of your ferrite.

Let me look around for some of my past study material to find more info for you.

 

[Svart]

LTC3204-5?


We need to know what your intended current draws on each voltage rail are going to be as well.


EDIT: This was a good paper and poses some of the reasoning behind my questioning of your usage of a flyback..

http://www.maxim-ic.com/appnotes.cfm/appnote_number/661

 

[jdbakker]

[quote author="Svart"]We need to know what your intended current draws on each voltage rail are going to be as well.[/quote]
Good point.

I have battery packs of 10 NiMH cells each (this is flexible), so:

Vin = Vbatt = 10...15V, but 12V most of the time.

I need:

VA1p: +9...+12V @ 700mA max
VA1n: -9...-12V @ 500mA max
VA2p: +15...+20V @ 200mA max
VA2n: -15...-20V @ 200mA max
VAP48: +50...+60V @ 100mA max (unregulated Phantom supply)

VD1: +5V @ 1A max
VD2: +3.3V @ 2A max

Let's not worry about VD1/VD2 yet; I think I'll use standard (filtered) step-down switchers, likely the LTC3431.

VA1p/VA1n and VA2p/VA2n need not be tracking or symmetrical. I don't even need them to be strictly regulated, since all channel strips have local regulators where needed.

[quote author="Svart"]How about moving to an inductorless converter or one that uses an inductor instead of a transformer?[/quote]
My main priority is regulator noise, with efficiency a distant second. The biggest issue arises with the negative supplies; standard inverting topologies are notorious for having strong switching noise on both in- and output, and (from what I can tell) Cuk regulators aren't much better. None of the capacitor charge pump designs I've seen have enough current output, and most of them are not synchronizable to my sampling clock.

I had originally planned to use a SEPIC to get the negative lines (as described in this Design Note), but those don't work too well when Vin is close to Vout and the regulator duty cycle is close to 100% (as is the case for VA1).

I've done some measurements on earlier inductor-based switcher designs that I have lying around (some Maxim parts, some LT ones) and all are too noisy for my liking for the analog supply lines.

[quote author="Svart"]The Transformer you are looking for is a coiltronics ctx02-13664 which somewhat fits the equation given in the document as well as being listed in a table of acceptable transformers.[/quote]
Like I said, it's also not documented (not anywhere I can find it, anyway). Tried the Coiltronix website, several US and foreign Coiltronix distributors and Google, but no joy. Do let me know if you have a link.

[quote author="Svart"]I assume using parallel converters is out of the question due to the space limitations of the project?[/quote]
Space is no problem at all. This thing is not going to be tiny (current plan: 45x30x10 cm).
[quote author="Svart"]The switching frequency is only one aspect to look at, the higher you go, the smaller the magnetics can be and still be efficient. What you want to do is choose a magnetic and then find your optimal switch frequency for the saturation and then choose your optimal pulse width.[/quote]
For optimum rejection, the switching frequency is tied down by the nulls in the digital filter in the converter, which are hard-linked to the sampling frequency. On the other hand, the higher you go, the more energy ends up at the sampling rate of the first delta/sigma stage (6.144 MHz), which is Not Good either. So I'm stuck with a switching frequency of either 96KHz or 192KHz (or 384KHz, if I really need to). 192KHz is probably the best choice, as a compromise between parts size and efficiency.
[quote author="Svart"]Simulations are damn handy for this as you can measure the temp of your ferrite.[/quote]
OK, I'll look into that. Still nothing beats bench measurements, since simulations don't usually show things like magnetic field noise (and my recorder is small enough to have to worry about that).
[quote author="Svart"]LTC3204-5?[/quote]
Um, are you sure that's the right part ? Its voltage range is way too small, and there's no way I'm going to hand solder that ;-)
[quote author="Svart"]EDIT: This was a good paper and poses some of the reasoning behind my questioning of your usage of a flyback..

http://www.maxim-ic.com/appnotes.cfm/appnote_number/661[/quote]
Yes, but I'm not trying to do step up/down, and push-pull efficiency is higher than the standard flyback described in that paper. Even so, I don't really care about efficiency. The reason for the LT1533 is its ~100uV output noise caused by its controlled switch current ramping (which, incidentally, also lowers peak pulse current).
[quote author="Svart"]Let me look around for some of my past study material to find more info for you.[/quote]
Please do !

Thanks,

JDB.

 

[Svart]

JDB,

Check out this:

http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1003,C1042,C1033,P1093,D1479

They spec the Dale LPE4841 transformer which is much easier to get than the coiltronics. This will get you the flyback you wanted.

I used these a few times to build a supply for an RF/video device.. it's VERY quiet. I had forgotten the part number I used and looked through my old papers and found it. It was a reference design, pretty much worked the best when built just like the datasheet.


Look further into the datasheet and you'll see the differential rail regulator, very handy in your situation. I would definately make a separate supply for the 5, 3.3 logic though.

 

[jdbakker]

[quote author="Svart"]Check out this:

http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1003,C1042,C1033,P1093,D1479

They spec the Dale LPE4841 transformer which is much easier to get than the coiltronics. This will get you the flyback you wanted.

I used these a few times to build a supply for an RF/video device.. it's VERY quiet. I had forgotten the part number I used and looked through my old papers and found it. It was a reference design, pretty much worked the best when built just like the datasheet.[/quote]

Thanks ! I'll have a closer look at the part; I might end up using a Cuk converter instead since simple 1:1 transformers are so much easier to source (in which case I'll use either this part or the LT1534 which has a similar config but slew-rate limited ramps for extra noise suppression).

Any suggestions for sources for the Vishay/Dale part ? I've tried the usual suspects (Digi-Key, Mouser, findchips.com) but no-one seems to have this part in stock, or orderable in qty < 600.

Thanks again,

JDB.
[slowly coming to terms with the fact that I might have to roll my own transformer... seems to be a popular pastime around these parts]

 

[Svart]

you can get them from Newark (the USA farnell) via special order, or you can contact one of the local distributors, check Vishay's website and it will give you tons of local companies.

Good luck! :thumb:

 

[jdbakker]

[quote author="Svart"]you can get them from Newark (the USA farnell) via special order,[/quote]
Er. Yes. With a 600-piece MOQ. On >10 weeks lead time. Maybe not...
[quote author="Svart"]or you can contact one of the local distributors, check Vishay's website and it will give you tons of local companies.[/quote]
Wish I could, but I'm in the middle of a tax audit for my company, and one of the areas where I got my wrist slapped is for ordering parts for private projects through my company (which consists mainly of me), even though I (privately) reimbursed my company for the parts, and even though several of the private projects lead to commercial spin-offs. So I won't be doing that again anytime soon, and from previous experience I know that Spoerle, the local Vishay rep, really doesn't want to deal with private individuals. This is why I was looking for a US-based distributor with parts in stock and an EU-friendly shipping policy...

Oh well.

Plan B is a Cuk, for which I can get the parts.
Plan C is (ab)using one of those new Power Over Ethernet-transformers; they operate on the right frequencies, have the right pri/sec inductance and almost have the right winding ratio.
Plan D is winding my own transformer.

Thanks,

JDB.
[reading up on PoE, and on various toroid core materials]

 

[Svart]

Let me do some contacting, I may be able to get some through one of my distributors, I would like a few myself actually.