isolated DC/DC converter

I never really got the point of tubular PVs. Seems a huge waste of, well, practically everything. Especially taxpayer's money.

And what's this nonsense about DOE "investing" in technologies? A loan guarantee isn't an investment, it's got all the downside and no upside.

All air flow is counter-intuitive. Air has a mind of its own.

We build physical mockups, like plywood and cardboard boxes with heatsinks and fans and stuff, to evaluate cooling. Incense sticks make nice smoke indicators. We fiddle until it works, sort of like designing circuits with Spice.

John

I replaced one side of the enclosure with window-insulating film, then fired 'er up. Easy. Real-life conditions. Perfect visibility. You could even shoot IR readings thru the film, I s'pose.

The flow was pretty darn good actually. Better than I expected.

-- Cheers, James Arthur

Black plastic trash bags are almost transparent at thermal wavelengths. Or FLIR will sell you an IR-transparent window for a couple of K$.

John

John Larkin says, "We fiddle until it works, sort of like designing circuits with Spice."

That's not designing, that's fiddling/hacking. But I guess Larkin's attitude fits with the "designs" he posts... never enough definition to prove/disprove function. What a jerk! ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

You're still pretending to killfile me, and hanging on my every word.

Geezer sicko.

John

Hey, John -

He replied to dagmargoodboat. Even if he has you plonked, he probably doesn't have dagmargoodboat plonked. So, he will see your post through him.

I'm not taking a side here, just pointing out that he may not be pretending.

Exactly. He is debating me while pretending to killfile me.

Hey, he's a sicko.

John

He may not be pretending. Read my post. Read again. Keep reading until you understand.

Just because he responds to someone's messages who quotes you does not mean he has not killfiled you. For an intelligent person, you are emerging as a bit non-worldly (to put it mildly).

Yes, my bad, checking the report it's 60nV, not 50! Sorry :-)

Should be 12Vrms, if I follow you...

I admire you to have so much certitudes. You know nothing about the board, how it was designed,... but still you know it's not going to happen. Well...

Who said it was 400kHz **BW** ? Not me. The figures are coming from an FFT test, with a 78Hz bin width. The 60nV then translates to 6.8nV/rtHz for a white PSD noise, a not that unreasonable figure. Oh, and the inputs were obviously short circuited to measure the board internal couplings.

That's the "incremental" efficiency (or dPout/dPin), that is it doesn't take the PWM controller power into account. Only efficiency related to the switch, inductor and input and output filters. But you're right, it's not 98, but rather 98.2%-98.3% that I measured on the proto board. And it has to be that high, given the working environment and some other constrains.

Oh, and the board also has a 50kHz 1ppm DH2 medium power amplifier along with its ppm level DAC signal source but I'm ready to bet that it's not going to happen too... not speaking of the test setup I had to design...

--
Thanks,
Fred.

Of course he has killfiled me. What's stupid is that, having done that, why doesn't he ignore me? Why play this stupid passive-agressive game with himself?

Well, maybe because he's a senile redneck geezer.

John

Your 50nv was misleading. You did not indicate it was root Hz.

That value has little relation to the noise from the switcher. It makes little sense to measure noise at frequencies below the 400KHz switching frequency.

If there is anything related to crosstalk from the switcher, it could be intermodulation or various mixing products. These are extremely unpredictable and could be well above or below the actual crosstalk value. So they are essentially useless.

You have a 1.5MSPS ADC, and the 400KHz is well within it's bandwidth. So you need to do a FFT at 400KHz and above to see how much noise you are really getting.

In addition, a single measurement method can be highly misleading. You need to make a second, independant test using a completely different method so you can confirm the results.

If the numbers are important, such as a contractual obligation, you might also make a third different measurement to validate the first two.

Mike

Nope! I made that mistake once. Fully optimized board layout, merely

15nH between supply rails. That should "minimize supply inductance", eh? The board rang line a bell at 60MHz, 80% overshoot!

Any attempt to apply diodes to clamp said ringing results in a dull thud of ~100V overshoot, which is real voltage applied across the diode, because it takes time for a silicon junction diode's intrinsic region to fill with charge carriers. Si shottkies are not usable at this voltage, and SiC schottkies are not practical at the current level -- peaks of ~120A mean you get as much voltage drop across the largely resistive chip as you get from the inductive characteristic of the junction diode.

So on the revision, I literally opened holes in the board to "worsen" my layout. Shows what rules of thumb are worth!

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

Try something like this, they aren't expensive anymore:

--
Regards, Joerg

http://www.analogconsultants.com/

Tim, if I recall, you are running high power induction heaters, right? I thought you were using resonance. If so, can you arrange to switch the devices at the point where the load current goes through zero?

Next, how do the big boys handle it? Here is an IRF IGBT 1,200 volt, 80 Amp half bridge intended to run at 20KHz:

Fig. 14 shows the typical switching time vs. IC with Rg=5 ohms and L=250 µH. At 80A, the Tdoff is 300ns. A simple-minded calculation shows

V = L * dI/dt = 250e-6 * 80/300e-9 = 66,666 volts

Obviously, the diode has to turn on and dump the current into the rail, so they have the same problem you are describing.

How do they solve it?

Thanks,

Mike

The one cold morning there's a wee mist hanging in the air, the sidewalks look wet even though it hasn't rained, someone comes into the lab, flicks the button ... *WHADDABAM*

There's easier ways to do HV power transfers :-)

How about LiIon or NiMH? Every once in a while a battery is flung across the barrier by a Roman catapult, smack dab into a holder. Then the depleted one is fired back by a Greek catapult (until they run out of those ...).

Government grant? Technology looking for a home?

--
Regards, Joerg

http://www.analogconsultants.com/

Got to keep the traces and package paths to a minimum. TO220 with long legs isn't the ticket in more demanding situations, all of that represents inductors.

Here is an interesting write-up:

Not sure when that was done (wish publishers would get into the habit of printing dates) so I am not sure to which diodes Jim and David had access. Judging by the sketch of the Tek 7104 it must have been a while. Still they had one diode (figure 8) under 4nsec and SMT wasn't too common back then, at least not for big ones.

--
Regards, Joerg

http://www.analogconsultants.com/

We made one box, a kilovolt impulse generator, where one essential part of its operation was to forward bias a power diode by 48 volts. It took, as I recall, a couple of hundred ns for the current to ramp up to 50 amps or so.

John

Do they make ones that don't explode when you put them across rectified

480VAC? ;-)

We run our inverters from 240 or 480 VAC, so DC is 300V minimum, up to about 700 when it drifts up unloaded. Naturally, we change the diodes, capacitors, transistors and transformers depending on the supply voltage,

240 or 480.

For the old board, schottky really would've been handy, but fortunately, careful design of the snubber will allow regular silicon junction diodes to do a fine job.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

Thanks for the link. The device I was referring to was a bit bigger than a TO220. I picked it since it would be difficult to get the inductance down to 15nh or so. This would aggravate the problem and make it more difficult to solve.

Of course, the IGBT won't switch as fast, but it has a lot more energy to deal with.

But these devices are used in many places where high power is needed. So how do they handle the switching transient?

Mike