first box didn't explode

Very cool. Did Protocase engrave and infill the front panel, or did you use FPE or someone like that?

-- john, KE5FX

They did everything, including silkscreen on the rear panel, from a Solidworks file and an ai for the art. They are very nice people. The finish is beautiful. I think we got a few boxes for under $300 each. They even did the fan grille on the side.

Very pretty! What's the jitter spec on the alternator outputs?

Cheers

Phil Hobbs

(Looks like those sticky-up toroids are going to need some hot snot to survive shipping)

I keep telling the kids that they are simulating an alternator! They are going overboard on elaborate sinc3 filters for the delta-sigma ADCs (isolated voltage and current sensors, to measure things and simulate impedances) and DDS stuff, and alternators aren't 16-bit devices.

Why are DSP jocks so hostile to IIR (integrator based) filters? They are so obvious and simple. The classic sinc3 filter terrifies me... a string of integrators (rolling over yet!) cascaded with a string of differentiators, with DC response!

We actually got the real FADEC alternator, chucked it into a milling machine, and spun it and measured stuff. The actual specs are unavailable and proprietary! I suspect we know more now about this alternator, than the people who make it.

They have little round rubber thingies on the bottom sides. We specified that the vendor include the mounting hardware.

Because they can blow up. I got caught out once when the real world presented a waveform that I hadn't expected and the IIR filter locked up. The pathological waveform came from an op-amp going out of common mode range and repeatedly flipping its output polarity. I was using a 16-bit DSP with 32-bit accumulator.

If a corner frequency is very low compared with Nyquist you need a ridiculous number of bits in the filter coefficients to get the expected results. Rounding errors can accumulate to a surprising extent.

On the other hand, FIR filters may be inefficient, but they don't surprise you.

John

Well, of course you have to be aware of limit cases, analog or digital. A decent digital IIR filter won't go crazy if the input isn't railed, and will recover too.

For extra fun, make the coefficient multiplies all right-shifts. That's a whole new class of filters.

I've done IIR lowpasses in uP code too, where FIR wouldn't be practical.

That's where saturating arithmetic comes in handy. If it won't blow up when made of op amps, the IIR version won't either.

Cheers

Phil Hobbs

About 18 yrs ago I did a design for AVAYA (nee Lucent) in Denver and figured out how to design cool AND quiet using open frame power supplies and/or motherboard in a large volume box. This kept it quiet and cool for a 19" rack.

2) diffuse and increase exhaust area to reduce air velocity like a muffler then use laminar airflow exhaust without a grill that causes eddy current noise. Egg crating works well.

4) Most importantly, keep moving fan blades in a laminar flow plenum, far away from grills that create turbulent noise. Plenums can be like any air duct material, cellulose, metal or plastic.

6) use Thermistors to control variable DC-driven fans to enable power and regulate fan speed for hot spots at 50'C (e.g. epoxy thermistor to XFMR)

Summary

- Laminar intake

- Spoiler and plenum to create large eddy currents at max air velocity over hotspots

- diffuse exhaust plenum then laminar output

My inspiration was a DELL tower, circa 2002 with an inline plenum muffin fan to remove CPU heat directly via large hose outside the box, rather than directly CPU heatsink with fan turbulent noise.

That laser-cut grille may be chopping up the air flow multiple times per blade rotation. We could open that up and use a classic wire grille. Or none.

Each class-D amp has a fan on the bottom, blowing right on a heat sink, but that's quiet.

Have you considered also selling it as a car stereo subwooer amplifier?

Probably could push some serious bass

No, at full power the transformers are only good down to 100 Hz. That's 1000 RPM for the type of alternator we're simulating.

And it costs more than a car.

We took some big dykes and cut away the four little webs

to remove the grille pattern. There's a big hole now. The sound is a bit quieter, but much less whiney and annoying.

We have a classic wire grille in stock somewhere, so we'll add that.

onsdag den 17. juni 2020 kl. 23.06.20 UTC+2 skrev John Larkin:

flipping the fan so it is sucking instead of blowing or vice versa can also have a big effect on sound, though it'll also have effects on the air flow

Strangely, there was as much whiney noise coming out the back of the box as the side. In a rack, the noise out the back is most likely to annoy customers.

If you find an inline Dell plenum fan, that will work, otherwise follow my plenum ideas.

If you find an inline Dell plenum fan, that will work, otherwise follow my ideas and beware of panel resonance with RPM.

Hmm. ON STBY POWER ON.

weird layout, sort of like an an IBM midrange system with power switch that looks like it's a circuit breaker with a trip indicator position, but isn't.

Why not use the normal air intake on the front, exhaust on the rear arrangement?

In the datacenter world, a few companies tried the sideways breathing equipment designs. All that junk burned out and wasnt'a problem for long though.

Does that mean you won't be buying many?