30V from 24V

No, because the output cap is big (1000 uF) and the driver is a current source. The edges look fine.

I don't want a constant load on the supply, because that'll waste more than 10 watts, and severely complicate the heat sinking. (This thing has to go inside a small waterproof NEMA box, so I can't just hang a fan on it.)

Cheers

Phil Hobbs

Americans eat about 75 billion eggs per year. If you got a royalty of

I like your thinking. ;) Now all I need is to convince the hens.

Cheers

Phil Hobbs

Ignore them, they work for chicken feed...

How about an HF burst going into a voltage doubler, with an inductor to mod erate the current ? After all, LEDs want constant current more than they wa nt constant voltage.

You would still have the ripple problems, but 4,700 uF should do it.

Crude as hell, in fact probably cruder, but then you are only making the vo ltage when you want it. Component count is down. Just have to choose the in ductor right to get the range of ouptut you want for an acceptable range of frequencies.

Trying to solve future problems by using a fixed frequency does not always work. Use what you saved on a DC convertor to take care of that.

Might or might not work, only you can do the math. You said some sort of sq uare wave, but isn't it going to be like a short pulse ? It seems like the eggs will be moving and they want a quiksnap of the internals. Like a shutt er speed on a movie or something, but you can't control the speed of the eg gs like a film, it would scramble them all in their shell.

I think you did say what the spec was but this thread is long enough now it is a PITA to look back. How many mS or uS do you need and how often ? Hell a simple pump charge dealio with an inductor might doya.

BTW folks, that font is called jolt and it is a TTF. You might want to use it sometime just for the hell of it :

Don't tell them where you got it lol.

Nope, 1 kHz square wave. Total measurement time is about 50 milliseconds per egg. It uses one of my trademark boxcar lock-ins, so that the acquisition transient is always exthe same shape and there's no ripple to worry about.

Doing that at 20 watts isn't that easy.

I decided to go with an LT3580 boost regulator, because that allows me to optimize the required headroom. With the inverting regulator approach, I'd be vulnerable to sags in the 24V supply unless I did something fancy.

The 3580 runs really fast, so I can save some dough on the inductor, and has built in soft start and undervoltage lockout, which makes it simpler to use. If it starts up too soon after power is first applied, the big reservoir caps won't have had a chance to charge up all the way, so it'll take several times longer to reach steady state.

There's a 2.5 A two-terminal current limiter on the input, just in case some wiseacre someplace in the backwoods where the power is unreliable decides to run it off two car batteries in series. Connecting 6600 uF of low-ESR cap directly across that source would be very hard on the switches and connectors, if nothing worse.)

With all the filtering required, it's a bit on the clunky side, but not too bad for what it does. Most important in this instance, I can be sure it'll work without having to be prototyped. (The customer is in Malaysia.)

Cheers

Phil Hobbs

On a sunny day (Fri, 07 Nov 2014 13:01:14 -0500) it happened Phil Hobbs wrote in :

I would not dare sell a system design I had no working prototype of. It is also good for fault finding.

That's just a bunch meaningless gobblygook, there's a bunch of engineering, physics and biology information missing here. I didn't take you for the idiotic type, so you must be blowing smoke for fun.

It's actually a reasonably complete description of the spectroscopy, except that I didn't say that the Easter egg dye is part of the published calibration procedure of a competing instrument, which it is.

What part didn't you get?

Cheers

Phil Hobbs

I prototype the parts I'm not sure about, which in this case was the optical system and generally trying out the measurement, using a PARC lock-in.

Cheers

Phil Hobbs

On a sunny day (Sat, 08 Nov 2014 10:50:57 -0500) it happened Phil Hobbs wrote in :

NASA for example has models of the mars landers . When a problem arises, they can test on those models to see what could cause it, and what remedies can be remotely applied. Malaysia is easier to reach than mars, but still expensive if you have to go many times. Also provide a remote way to update firmware (internet connection) I did that some time ago with some far away computer system, just telnetted to it and fixed it. You do not need much extra hardware for that, any Raspberry will do.

On Sat, 8 Nov 2014 02:37:26 -0800 (PST), snipped-for-privacy@gmail.com Gave us:

Look, folks... Here is bloggs yet again proving that he is an abject idiot.

And he is a goddamned Usenet retard for the retarded interface he and the other retard uses.

wo

ou

ng,

Take a look at your useless posts suggesting 30 year old technology with li ttle to no understanding of the any of the engineering or science. You coul dn't put together a run-of-the-mill switcher by the numbers. -Less than pat hetic- probably congratulate yourself for achieving 5% comprehension of an app note abstract- Unitrode no less.

The absorption characteristics aren't the only thing smeared here. The numbers you cite are all statistical, and no mention of the sensitivities, or pushing factors, and whether the associated control parameters are to be stiffly regulated or adapted.

Yeah, he should post all his research and his entire design, before he sends it to his customer.

Something like that has occurred to me. The relative obscurity of Usenet is probably a good thing...

(I know the downloaders in the binary groups certainly think so)

I have no idea what you're talking about. Do you want to apply statistical process control to _hens_?

It's an egg grading machine--you look for cracks, dirt, and blood spots, sort for size, and pack in crates. Rejects just get dropped into a bucket, which is emptied and washed each shift and the contents goes for pet food.

If it were tracking ICBMs, the error rate would have to be pretty low, but nobody much cares if you chuck out one good egg in 1000.

Cheers

Phil Hobbs

If they have problems, they can mail me one. That's a lot cheaper than paying me to do full-up protos, and it's usually unnecessary. I did prototype an entire transcutaneous blood glucose & alcohol setup a year or two back, including dead-bug-constructed SMPSes.

Cheers

Phil Hobbs