Power supply wisdom

I'm designing a photodetector for moderate-quantity production (probably

1000-5000 units over its lifetime).

It needs a few power supply voltages inside, like

+3.6 for the micro, < 50 mA +- 15V for the analogue stuff +50-75V for the photodiode bias.

It only dissipates a few watts total, mostly from +-15.

Total BOM cost including box and board, excluding supplies, is about $220 at most. DC-DC converter modules are expensive enough that they'll add another $100 to this, which stinks.

Sooo, it looks like I can build my own DC-DC with several outputs, and post-regulate them. Not awful, but will need a custom transformer.

However, it occurs to me that I could very easily use a 12-14 VAC wall wart, and do it the old-school way with a couple of half-wave regulators and a Cockcroft-Walton tripler for pretty cheap--mostly the cost of the caps, which should be under $10 altogether in 100s.

One drawback is needing different wall warts for different markets, but that isn't a terrible problem in this instance.

Alternatively, I could use a +18V wall wart, one inverting switcher, and a charge pump / Cockroft Walton.

Suggestions? Pitfalls? Input protection advice?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs
Loading thread data ...

l

I think all of the ADSL modems I have seen use an AC adapter, I haven't checked but I assume they do it to easily get a bipolar supply

lots if those usb chargers are 5V/2A switchers, an ldo gives you

+3.6V, something like this:
formatting link
+/- 15V this:
formatting link
75V

don't have to pick Maxim ;), they were just the first that popped up on google

-Lasse

Reply to
langwadt

[snip]

I can remember, about a year ago, 10W boost, from +5V to +24V, was ~$20 from Newark in single quantities.

(Uncased, PCB with 4 pins, to mount to your mother PCB) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
 Click to see the full signature
Reply to
Jim Thompson

Do you *really* need +/- 15V? Much analog today is done in 5V, or even less. A lot of our stuff is going to 3.3V.

Reply to
krw

Companies like Pulse and Wurth have transformers with 6 independent windings which would allow for converting 5V into different voltages.

5V is very common for USB gadgets so adapters are extremely cheap. Even though custom transformers don't have to be expensive I wouldn't design a mains PSU unless someone is going to pay for agency approvals.
--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
 Click to see the full signature
Reply to
Nico Coesel

I lose more than 20 dB of dynamic range by going to 3.3V, so yes, trust me.

Thanks

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

Personally, I'd go with the 18V-output switching adapter for this kind of thing and derive the rest from switching supplies. I'd look at 12V and 5V ones too.. they might be a lot more availablec/cheaper in relatively small quantities and thus worth adding a bit extra on the rest of the BOM. Maybe you can reduce the analog to +/-10 or +10/-15 or some such thing.

Be careful about the inverting switchers.. some topologies draw massive start-up currents which might drive your SMPS adapter into current limit.

Switching adapters make the product MUCH easier and cheaper to package and ship-- the linear transformers are a bit of a battering ram in the package which requires some care. Assume the box will be drop-kicked off a truck tailgate and you'll be right some portion of the time.

What kind of DC-DC modules add $100 to the BOM!? A VASD2-S5-D15-SIP is under $6 in 100's and provides isolated +/-15V at 2W from 5V in. An advantage to using 5V in is that you can use a USB-style plug and adapter. Getting the PD bias could use a boost converter (which only requires a standard inductor). Or use 2-3 of the 1W isolated +/-15V version in series for 60-90V but that will run another $12. Those numbers can be beat for higher quantities, that's just Digikey prices.

Reply to
Spehro Pefhany

Sure it's not 3.3V?

Plus you never know what their EMI footprint will be or when the day comes where the mfg has this brilliant idea to save a couple pennies, and then all hell breaks loose in your signal processing.

Yup. Provided the +18V are robust (as in "laptop supply") I'd do it thaddaway:

a. Buck switcher IC with internal switch to deliver the logic supply voltage. Doesn't have to be a sync buck.

b. Sync buck to deliver +15V but use a dual inductor (often called SEPIC inductor). For the analog ... ahem ... analogue stuff. Regulate this converter from the positive output.

c. Hang the 2nd winding in reverse, plus a Schottky diode, this will make -15V. Well, actually -14.5V. If it really has to be -15V use a FET but then the switcher must have external FETs so you can drive it. Important: Force the sync buck into continuous mode so you negative supply won't droop if the positive rail isn't loaded enough. Although this negative supply is unregulated it tracks quite well.

d. Hang a Cockroft-Walton onto the switch node. 3x or 4x step-up because you'll have 18V swings there. Alternatively you could use a flyback transformer but that usually costs more.

So in the end all you have is two switchers, one makes the logic voltage and the second one makes the other three voltages.

Laptops supplies come with all the certs and universal input, and some even with a variety of power plugs.

Well, careful with EMI, as usual.

Get a good quality laptop supply, not the "fell off a truck in Shanghai" kind. If you want to be extra good put a TVS across the DC jack. Also some LC filtering it passes EMC.

--
Regards, Joerg

http://www.analogconsultants.com/
Reply to
Joerg

It'll tolerate more than 3.6, and using 3.6 makes sure it drives 5-V logic reliably.

Yup. The previous version of this box was in production for almost 20 years.

I'll probably just use an LM317 for this. The power loss isn't too horrible since the MCU is asleep most of the time. It's just there for supervisory stuff and self-calibration.

If I'm using a 19V laptop supply, I can use a cap multiplier driving an LM317, which will come out just above the 317's dropout voltage.

So out of curiosity, what's the downside of doing the retro 60 Hz thing? All I can really think of is the size of the caps, and needing different supplies for Europe, and it sure seems like it would make the EMC problems easier.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

I've been out of the certification loop for a while, but isn't that a blessing? Someone else paid for all the certifications for radiation and safety for line-connected stuff...in all those different markets.

Reply to
mike

Just be careful, I wouldn't scrape along too close under the abs max value.

[...]

Now that won't get a green-tech sticker :-)

Just keep in mind that the usual LM317 wants 10mA min load.

This I don't understand. The laptop supplies delivers DC, and the LM317 wants DC.

If you want to do a cap multiplier I'd use an inverting gate driver with Schmitt inputs that can at least do 25V and run this as an oscillator. Then use a 3-4 step cascade to make the high voltage and a two-step to make the negative voltage.

It's ok but you'd need a fairly big wall wart. Afterwards, the fun starts here:

US: 120V Japan: 100V Europe: 230V Except in the UK where it's 240V North Africa: 220V (sometimes ....) Caribbean: Whatever Felipe can squeeze out of the generator Australia: Anything from 205V to over 270V.

Electrolytics age. When they are on your board it'll be a service case. If they are in a laptop supply your client can just send out a new laptop supply. This can be major, especially if you are talking countries with onerous customs rules.

--
Regards, Joerg

http://www.analogconsultants.com/
Reply to
Joerg

Only if noise scales with voltage. I haven't found this to be the norm.

Reply to
krw

...or I mean only if noise *doesn't* scale at all with voltage.

Reply to
krw

At least put a bridge on the input so you don't need a particular center-neg or center-positive coaxial plug. Most CUI stack wall warts come with various world plugs. These are switchers too, so filter accordingly.

The non-switchers seem to only come in 120/60hz.

Cheers

Reply to
Martin Riddle

less.

No, if noise scaled with voltage, there wouldn't be a problem--the noise would always be a constant percentage. Optical measurements are characterized by extreme dynamic ranges, mostly because all optical detectors are square law devices--electrical power is proportional to the square of the optical power. Dropping the power supply by a factor of 10 loses you 20 dB.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

less.

I meant the opposite (correction post made).

Assuming you need the entire dynamic range without amplification. Droping the voltage scales the same, modulo some noise (nothing for nothing).

Reply to
krw

less.

Dynamic range is dynamic range.

In photodetection, your SNR is generally limited by the Johnson noise current of your feedback resistor and the input noise voltage of your amplifier, which gets differentiated into a noise current by the photodiode capacitance.

Even with a noiseless amplifier, in order to get to within 1 dB of the shot noise of your photocurrent, you have to drop at least 200 mV across the feedback resistor. (Proving this takes two lines of algebra, starting with the Johnson and shot noise formulas.)

With +-15V supplies, you have 20*log(15/.2) = 37.5 dB of headroom when you get to that point, whereas with a single 3V supply, it's

20*log(1.65/.2) = 18.3 dB. Not to mention the horrible increase in photodiode capacitance if you only have a volt of reverse bias, which makes it horrible squared.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

There's SMPS quiet, and then there's instrument quiet. I could probably live without the LM317, but certainly not without the cap multiplier. Most of the time, this box is going to be dissipating maybe a couple of watts, unless somebody insists on driving +-10V into a 50-ohm load, at which point it'll rise to three or four watts.

The laptop power supplies that I know about don't have such terrific regulation--their outputs can rise at light loads.

It's a piece of high sensitivity lab equipment, so I wouldn't be that worried about it. If the power is good enough to run the ion accelerator, it's probably good enough for my purposes too. Also, a laptop power supply is good for what, 80 watts? That's a factor of 10 higher than I need, at least. The 10k hour@105C electrolytics ones run about a buck apiece (for 3300 uF @ 35V), and they should be okay for a couple of decades at rooom temperature, I would expect. The crappy ones, not so much.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

Is that one of the little Pico ones? I had to try fixing somebody's laser locker that had one of those in it--it sprayed harmonics of 50 kHz _everywhere_. I'd need the kind with good shielding, toroidal power transformer, ripple-nulling topology, the whole nine yards. If I could find that for $20, it would be a no-brainer.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

Pico's pricing is insane. Murata/CUI/Vinfinity sip DC/DC converters come in single and dual outputs, for about $4 or so each. We use tons of them. You have to be careful about noise, but that's manageable.

I've made my own low-power converters with an IR fet driver chip and ISDN and/or dual-winding power inductors used as transformers. Here's one:

formatting link

The unregulated HV runs about 120 volts, which the depletion fet limits to a bit under a mA into the zeners. This is all cheap. The gate resistors in the inverter section soften up the switching, trading a bit of efficiency for noise.

The 60 Hz thing wouldn't be bad, especially if you really care about noise.

--
John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
 Click to see the full signature
Reply to
John Larkin

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.