witches' brew

I'm surprised someone decided to get "fancy" and mount a capacitor at a 45 degree angle...

On your board they look rather thick -- I take it the transformer is what makes them so thick? (Since the PCB you show there sure isn't...)

---Joel

Yuck, hard switching into a capacitor, too? I'd rather use one of = these. Regulated, current protected, and only twice as many = transistors.

Tim

--=20 Deep Friar: a very philosophical monk. Website:

"John Larkin" wrote in = message news: snipped-for-privacy@4ax.com...

Regulated, current protected, and only twice as many transistors.

Lots-o-parts.

The CUI is probably...

ftp://jjlarkin.lmi.net/VASD1_sch.JPG

which is the classic circuit. The resistor has to be small enough to furnish the worst-base operating base current requirement, which fries transistors if you short the output. A couple more parts would fix that, with a smaller startup current and then let the oscillation feed the bases as it grows.

The FR4 substrate doesn't cool the transistors like ceramic would.

They actually work pretty well. Regulation is good (for certain definitions of "regulation") and the 1-watt units work fine, barely warm, at 2 watts. Noise isn't bad at all. It's not worth making your own for $4, and the sip package uses board height and saves footprint. They also protect those tiny JTAG headers.

This business is, increasingly, about putting boxes together.

John

It has been said, for a car, that addition of a 50 cent part increases the consumer price by 5 dollars. So...the addition "of another part or two" would almost double the price. "QED"

Maybe not; series inductance might be designed into the transformer-looking gizmo.

I once had an ... amusing... half hour trying to explain to a colleague about a constant-voltage transformer. He kept thinking it was a transformer. Not true, the magnetic circuit is much different, but the lack of a distinctive name got in the way of any understanding.

It's a tiny toroid in the CUI thing, so I'd suspect insignificant leakage inductance.

John

Does it conduct much switching noise back into the supply?

Funny, I made a chopper recently, having need of isolating a few hundred = milivolts. Quite noisy, but it's low bandwidth so it all filters away.

Off topic: doesn't use toroids.

The interesting thing is, since the chopper is shorting-mode commutated, = it *doesn't* work with a bypassed supply. It astonishes me how many of = these circuits are out there, with 2N3055s and iron core transformers, = and they'd work *so much better* with just a little series inductance in = the supply!

Tim

--=20 Deep Friar: a very philosophical monk. Website:

milivolts. Quite noisy, but it's low bandwidth so it all filters away.

They are small, efficient, and apparently cheap. But harder to wind at home. I once worked for a company that had a small toroid winding machine, and we could make our own. That was sort of cool. I can't imagine how they wind those tiny bifalar ones in the dc/dc bricks, or Ethernet magnetics. Maybe by hand with starvation-pay suicide-prone labor?

*doesn't* work with a bypassed supply. It astonishes me how many of these circuits are out there, with 2N3055s and iron core transformers, and they'd work *so much better* with just a little series inductance in the supply!

Yikes!

John

I didn't measure that. I did include a ferrite bead in the +12 input and an extra cap in the converter side.

Since I'm doing isolated channels, I care more about common-mode noise current. The unloaded output shows about 600 mv p-p common-mode noise on a grounded scope, not bad. It has 83 pF input to output, OK but higher than I'd like.

John

milivolts. Quite noisy, but it's low bandwidth so it all filters away.

Opened up a 1Gbps combined ethernet connector/line interface (like what's on a PC mobo)? Seven or eight little toroids, some fancy connections, all interconnected... Hard to imagine them being made by a machine. There must be some really boring jobs out there :(

*doesn't* work with a bypassed supply. It astonishes me how many of these circuits are out there, with 2N3055s and iron core transformers, and they'd work *so much better* with just a little series inductance in the supply!

Interesting if ones thinks of improving the thing by slapping a big cap on input? Oops.

Grant.

In the real world there is no correlation between them, other than price >

cost. Put another way, if I shave $.50 off the cost of a product there is no corresponding drop in our price. Price is set by a completely different set of equations than cost.

If we add a part, the price goes up 4x the part cost. If we delete a part, the price doesn't change.

John

That's because you have no idea where to price your products.

That's often true. There is theoretically a curve of total profit versus unit selling price, a sort of inverted parabola. Many small companies have no idea where they are exactly on that curve, and have no good way to find out. If you have a lot of competition and a lot of history, you know; the flip side is that, in that situation, margins are usually low. When you have a high-margin niche product that is IP intensive, and no direct competition, it is hard to set pricing.

I recall reading some studies that suggest that most companies tend to set their selling price below the peak point, because more sales make them feel better.

We have lately been adding sales reps all over the country, and we're asking them what they think of our pricing. They are better placed to research that than we are, since they talk to a lot of potential customers. The only feedback so far is the the e/o stuff looks cheap.

If you sold breakfast cereal, you could do experiments: increase the price 5% in 5 cities and wait and see what happens. We don't have enough statistics to do anything like that.

John

Last place I worked at, raw margin was 8x parts, but they didn't count software development. On lower sales volume devices, software cost around $400 per unit, but a common high volume part had a mask chip that meant software cost was almost free. Lots of variation, and it was a small company in a niche market.

Grant.

price.

no

My company averages parts cost about 22% of total sales. A lot of that is pc boards, sheet metal/panels, and expensive analog ICs. We define a thing called "direct cost" which is parts and unburdened assembly+test cost. Selling price can be 2.5x DC (marginal) to as much as 6 or even 8x DC, great if you can get it. At the end of the year, we make a few per cent, so it must be about right, or maybe we're pricing a little low. Engineering, sales, management, are all overhead to be covered by the markup on DC.

John

Nevertheless, car companies do. Price and cost are totally separate (noting the obvious exception).

"e/o"? "Cheap", as in inexpensive for the market, or "cheap", as in they don't like your labels?

I didn't say you were wrong to price the way you do. Obviously you're successful. Car companies don't price a car up five bucks because they spent another $.50 on a bolt. OTOH, they might price it up $50 because the $.50 chrome bolt looks spiffier than the $1 brushed bolt.

...and it builds long-term value as well, since not only will that many more people will know your company's name, but they'll think of your products as a good value.