DC-DC conversion from 1 Kilovolt DC down to 12VDC (isolated)

Sure. That's the only way to make money.

So you don't have production problems? Interesting. I did

Semiconductors are very repeatable. If you can test parts to destruction, and back off some prudent amount, they will usually be reliable. For example, some microwave schottky diodes that are rated for 2 volts reverse are fine at 5. Microwave parts aren't characterized for things they can actually do well.

An 0603 resistor, rated 1/10 watt or some such, can dissipate a watt if you mount it right. And you can pulse it at 5 watts.

Sometimes you take chances, when the payoff is big. If you take chances all the time, you'll get into trouble. Tantalum capas can explode at half of their rated voltages, and 300 watt reted mosfets can blow up at 200 watts. Testing to destruction can reveal a lot.

John

[on shorting a HV power source]

All I remember is a little 'click' if the geiger counter had a speaker, and a tad of a bounce if a meter needle was the only output. You don't usually have a big filter capacitor on any HV source, and shorting it is almost ALWAYS nondestructive.

Mind, I'm thinking this kind of power is about what a submarine data link's repeaters might depend on, and that means the precautions against shorts might get a lot of engineering attention.

If the voltage ratings by this company are as "optimistic" as their current specifications ("package limited" ), it would be a good idea to use a large headroom :-).

On a sunny day (Sat, 25 Jun 2011 12:53:52 -0700) it happened Joerg wrote in :

Yes I remember Grundig, but not for defective SCRs. It is also possible that in some countries other models were sold. For example I had a multitude of Philips... Greatz, Blaupunkt, some East German sets, some Russian sets, all sorts of things. Philips came also under other names, like Aristona etc.

Yes SCR in H out is really a bad idea, never understood why they did that.

Some were more reliable, but for example Sony worked around the PAL patent so they did not have to pay royalties, they used no delay line either,

3 chroma demodulators for R, G, B, and believe it or not: 'a hue control'. Now that hue control was what was no longer needed in PAL. They sold a lot of that inferior crap though. Regarding that I think Samsung is the new Sony, man that 3D 46 inch set I have is bad. I would fire the whole design department that came up with it, Samsung is Korea no? If so the Mr Kim Il has a job to do.

Yes some years ago I put my analog portable with the garbage, only showed noise on DTV, Maybe these days, where no standard last more then a few years, better get a good monitor and a receiver box, like I have an USB one for DVB-T, and a PC card for DVB-S.

I am dreaming of a new PC with the new AMD graphics-CPUs that come out this year. Those can do full HD HDMI, but if it will make the Samsung look better... I think not.

I've found the same thing. I recently came across one of our products that was using an 0603 (rated at 1/16W) at 250mW. It wasn't even warm. Neither end was connected to a plane, either (differential terminator).

Were you the one here who postulated that all SMT resistors dissipated the same heat (if by itself)?

I've been getting away from Tantalums, not because they blow up (no known problems) but because of cost.

SOA?

Or on a ship, the room where you flush it?

-- Cheers, James Arthur

That's easy--women have to paint them because ... they don't paint themselves.

Grins, James Arthur

things.

Raduga sets from Russia were famous for setting whole living rooms on fire.

On the scope (after replacing the SCRs) everything looked like there's enough margin. So as a kid I often wondered whether SCRs had a finite lifetime that's never reached when operated at 50Hz in a dimmer but is reached after a few years when operated at 15675Hz. Kind of like with avalanche transistors.

I liked Sony sets, good quality. The hue thing was a thing of the past since a long, long time. NTSC did not drift in color. In fact, our TV set didn't even have hue and tint adjustments, only in the setup menues.

is bad.

But not North Korea :-) ... They would never be able to bring something like that to market, plus they couldn't threaten anyone with it.

noise on DTV,

good monitor

year.

think not.

IMHO TV ain't worth it today. Although yesterday we saw a movie with Ronald Reagan that was pretty nice.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Probably me. But it was predicated on the end caps being soldered to copper pours, such that the pour area dominates. Center-of-resistor hot spot temps are about the same, 0603 through 2010.

I like them because the ESR is pretty constant over temperature, especially below 0C. But too much dV/dT will explode them. Derating

3:1 on voltage seems safe.

Probably. Many SOAs have black holes between the DC and, say, 10 millisecond curves.

John

On a sunny day (Sun, 26 Jun 2011 10:13:46 -0700) it happened Joerg wrote in :

Actually we once bought some 'Rigonda's, little portable all transistor TVs, to sell in the shop. Not bad at all.

I dunno, but I think that many of the problems with thyristors come from them not switching off again. The reason is then in defective caps (output tuning change), bad inductor connections. (inductors solder themselves lose in the board because of overheating). SRCs know something called 'hot spots', where only a small part of the si conducts, that then overheats and causes local breakdown. This last thing was an issue to watch out for with the bigger stuff I worked at,

400A sort of things.

Been I while since I worked on NTSC sets, but AFAIK they at one time had some color correction signal (burst) in the vertical interval, elimination the need for a hue control.

The 'trinitron' tube was a bit brighter, and was pushed as sales argument. It was also microphonic :-) Because the guns were in line the convergence errors were smaller, the electronics simpler, say cheaper. But it had stripes ;-(

have is bad.

I was thinking of nukes.

noise on DTV,

good monitor

year.

think not.

Reagan: I have seen a movie where he was sheriff and ordered a building burned down where an alien life-form was hiding. I used to stay up to listen to his speeches when he was president. If the cable company did not turn the equipment off when they went home late at night.

conducts,

at,

Limit the di/dt through the device to solve that problem.

The way I see it then, heat conductance is proportional to:

• Cross section area and area of contact of end cap to copper, each proportional to the square of the resistor's length

• Inversely with heat path length, meaning inversely with resistor's length

Resulting in heat conductance proportional to the length of the resistor.

Or, am I getting something wrong here?

As for heat conduction by air (independent of convection), area is doubled but temperature gradient in the air around the device is halved (for a given amount of power dissipation) if the linear dimensions of the resistor and hot air layers around it are doubled. That makes this heat conductance proportional to the length of the resistor.

As for heat removed by convection for a given temperature difference - that varies a little more than proportionately with the square of linear dimensions, provided the shape of the heat source is constant. Why a little more - the air moves faster, meaning more heat removed per unit area (for a given resistor temperature). Working against this a little is air boundary layers getting thicker (lower temperature gradient) as linear dimensions increase. So, I expect heat removal per unit area to be not far from constant for a given resistor temperature as its linear dimensions increase. However, for small individual parts, convection does not appear to me nearly as significant as heat conduction.

As for radiation - for a given resistor temperature and a given ambient temperature, amount of heat radiated is proportional to area, or to square of linear dimensions if the shape stays the same as size is varied.

But among these factors, it appears to me that heat conduction is the main one for SMT resistors 2010 and smaller. So, I expect power dissipation capabilities of these to be a little more than proportional to their length.

Now, for actual typical power ratings, and exponent for relating change in size to change in power rating, for 2:1 sizes: (1.0K thick-film in-stuck in cut-tape form at Digi-Key)

0201, .6 x .3 mm - 1/20 W 0402, 1.0 x .5 mm - 1/16 to 1/10 W 0603, 1.6 x .8 mm - 1/10 W except for one 1/4 W example 1206, 3.2 x 1.6 mm - 1/4 to 1/2 W, mostly 1/4 or 1/3 W 2010, 5.0 x 2.5 mm - 1/2 to 1 W, mostly 3/4 W

This mostly supports power dissipation being proportional to length raised to a power in/near the range of 1.18-1.55. I also see some trend of this exponent increasing with size, mostly 1-1.32 over the 0201-1206 range of sizes.

======================

Meanwhile, I am not surprised at SMT resistors surviving continuous power well above their ratings if they are heat-sunk by good solder joints to oversized traces, or if used on metal core PCB.

I once had a factory prototype of an LED lighting product (with metal core PCB) come back to me with a few 1206 resistors substituted with 0603 ones - that part of the circuit worked fine. (With generous amount of solder and pads sized for a 1206 and traces .06" wide.) Nearly 1/4 W was being dissipated in 1/10 W resistors, and they were doing just fine.

However, I see this as an example of knowing what parts can be safely operated outside what ones of their datasheet ratings and how to do so, should there be a customer for something that needs "extreme engineering" for some constraint - often a size one.

--
 - Don Klipstein (don@donklipstein.com)

your still on the ragged edge.

I think you've still not grasped the circuit. There is no reason for the core to saturate, certainly no cross conduction, where could it cross conduct to?

I don't smoke anything.

Some igbts maybe

Most resistors are made on 20 mil alumina substrates. So thermal conductivity is proportional to width and inverse on length. Since most parts have similar w/l ratios, most have about the same thermal conductivity end-to-end, irrespective of size. We measured the central hot-spot temperature of resistors from 0603 to 2010, all at the same power dissipation, but soldered to lots of copper. They were essentially the same.

Look at these!

John

?

That what I'd probably use if I'd have to design this.

If there is capacitive coupling into the transformer it'll be ok. Split primary with direct drive may not be ok. Cross conduction can happen with a half-bridge, through the two transistors or FETs. But there's driver chips that handle this.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

20 mil alumina substrate in all sizes from 0603 to 2010? That would make heat conduction through the ends in watts roughly constant for a given central temperature rise.

WOOF! As in A250N50X4? Dissipate 250 watts into something 3/8 inch by

3/8 inch by .04 inch? That has to be attached to a heatsink. Otherwise, its 2.2 square centimeter total surface area would dissipate 250 watts mostly by radiation - at a temperature of a little less than 2112 Kelvin, 1839 degrees C (going by blackbody formula), producing maybe a few hundred lumens of light. That part is only rated to 200 degrees C.

--
 - Don Klipstein (don@donklipstein.com)

That's what we measured.

That would

Right. We measured that, too, with a Flir thermal imager.

I think they need heatsinks! AlN rocks: it conducts heat almost as well as aluminum.

John

It's fortunate that you don't.

e

e/

t

OM*

e.

You still haven't explained what your talking about here.

Now what are you talking about?

Split primary with direct drive may not be ok.

Why not?

Cross conduction can happen

There are drivers but they only go upto 600v busses, even at this voltage most comercial designs don't use them.

The reason I don't use PowerMESH is that I haven't yet needed quite that much breakdown voltage performance. But if I had to I would, unless I find a better deal :-)

See below.

Ok, a picture is more than lots of words:

If you have an ever so slight timing imbalance in the gate drive of Q1 versus Q2 then two things can happen:

a. The rapidly spiking sense node CS is recognized fast enough and U1 curbs power.

b. It is not sensed fast enough ... *KABLAM*

If you couple capacitively and the cap is small enough there won't be a DC offset concern.

Right, I usually don't either. But mostly because I turn into uncle Scrooge when doing circuit designs for clients. These days you can get gate drive transformers for well under 50c. Ok, at that price they'll be coming in from Outsourcia, of course. Sometimes I couple capacitively.

--
Regards, Joerg

http://www.analogconsultants.com/