120 Degree Phase Shift Osc.

If it's the "intelligent scissors" function they're using to strip the background, I actually took a class back around the turn of the century from the guy who wrote the code for it -- one day he flipped on a few hidden "debug" switches in PhotoShop and demonstrated how it really worked (the debug switches enabled cool little animations where it would show how/where it was searching for edges in this outwardly-spiraling search); I was surprised at just how sophisticated it was.

...which might explain why it works better than the GIMP's...

messagenews: snipped-for-privacy@30g2000yql.googlegroups.com...

No, I consider a solid state dimmer to use a potentiometer to set the trigger VOLTAGE. A traditional dimmer pack was a Rheostat, connected in series with the lamps to set the CURRENT. Two ways of doing the same job. I have worked with studio lighting with both technologies.

They peaked in WW-II. ;)

--
Politicians should only get paid if the budget is balanced, and there is
enough left over to pay them.

messagenews: snipped-for-privacy@30g2000yql.googlegroups.com...

Yawn. You're as smug and ignorant as ever.

--
Politicians should only get paid if the budget is balanced, and there is
enough left over to pay them.

Michael A. Terrell Inscribed thus:

Yes, thats always a possibility though I've only seen a couple of real CCFL failures ! More often its the caps that go high ESR and I wouldn't rule out bad solder joints. Did you see the photographs I posted some time ago of the bad HV cap ?

--
Best Regards:
                     Baron.

messagenews: snipped-for-privacy@30g2000yql.googlegroups.com...

The French and Germans that I work with speak English at least as well as I do. And I can communicate with the Brits and Irish fairly well.

John

Where was it posted? I've seen a lot of photos, and real HV caps that were bad. I've been replacing a lot of different types for 45 years. :)

--
Politicians should only get paid if the budget is balanced, and there is
enough left over to pay them.

(PST),BillSloman

happenedBillSloman

:

stability, and doesn't i=

point out -

be. JA can

out-think you wit=

when you say

who was delighted

he was.

If you can't

sci.electronics.design, you

a 10Hz

cells."

that.

use "a

talking about

implementation

mA at

isn't

done at

Liar. That is *not* what you said.

You're projecting again, Slowman.

More lies. What else is to be expected from Slowman.

happenedBillSloman

happenedBillSlom=

doesn'=

-

can

out-think you =

you say

deligh=

can't

sci.electronics.design, you

the

I can, but am demonstrating that in addition to being a liar, you're illiterate. The part is the 87C51Fx.

More projections.

(PST),BillSloman

(PST),BillSloman

(PST),BillSloman

it happenedBillSloman

:

wrote:

(PDT),BillSloman

wrote:

(PDT),BillSloman

OP's

drawn

They didn't. I got to learn about it from books. It's kind of irrelevant when you are talking about generating a 10-bit accurate approximation to sine wave.

John Watkinson's "The Art of Digital Audio" ISBN- 0-240-512270 -7, first published in 1989, a year or so before I bought my copy, does talk about this subject. The 44.1kHz you are thinking aobut refers to the sampling rate, where each sample originally represented a 14-bit number.

The 1-bit DACs you are thinking about are clocked quite a bit faster.

claims that 11.3MHz (256 times over-sampled), is a representative clock rate for a 16-bit audio DAC, whose frequency output is intended to roll off rapidly above 22kHz.

Cursory reading suggests that if one adopted this approach to Mark Weaver's problem one could get by with 32x oversampling - you'd divide the 10Hz period into 32 roughly 3msec intervals, and pulse width modulate within that period to 10-bits (preferrably not by leaving the output high for the first portion of the 3msec and low for the rest), which would require a 330kHz clock. The maximum step size from sample to sample is about 10% of full scale, which you want to filter down to

0.1% - a factor of 100 - and a two-pole low pass filter with a 3dB point at 30Hz would do this (at a guess - I've not worked it out in detail. Each filter would introduce an appreciable phase shift at 10Hz

- about 60 degrees - so you'd want to make them with close tolerance parts. More poles in the filter would allow you to get by with less over-sampling, but this isn't a practical route to follow.

The two-poles filters with a 3dB point at 30Hz would be a nuisance - I'd jack up the number of samples and the sampling frequency by at least a factor of ten, which wouldn't cost anything worth worrying about, in order to make the low pass filters a little more practical. This jacks up the clock frequency to 3.3MHz.

We aren't talking three order of magnitude here - closer to the square root of that, and less in a practical circuit.

If you had more than a fuzzy grasp of the subject, you wouldn't have claimed three orders of magnitude - you obviously understand sigma- delta conversion even less well than I do.

-- Bill Sloman, Nijmegen

To you, I suppose. It's just a word. If we connect to two pins of a

4mm surface-mount pot, we still call it a pot.

As someone said, what you call them doesn't matter.

John

z.dk"

te:

ST)) it happenedBillSloman

e:

rote:

Sloman

illSloman

),BillSloman

s

It's one way that would work. Sigma-delta does require that you space the pulses nearly optimally, and filter the pulsed output. The 1-bit audio DACs are clocked at about 11.29MHz - over-sampled by 256 - to offer 16-bit resolution at 44.1kHz, and their low pass filter rolls off the output rapidly above 20kHz.

Doing the same sort of thing to get a 10-bit sine wave at 10Hz suggests - to me - a sample rate of about 320Hz and a clock rate of

330kHz, but that would require a two-pole filter rolling off from about 30Hz to keep the quantisation noise out of the coils. The filters would probably cost more than the rest of the circuit, so you'd probably go at least a factor of ten faster to make the filters smaller and cheaper.

This is the extrapolation to a much higher order filter and a much lower over-sampling factor.

The filter would be a finite impulse response filter made with a - say

- 32-stage shift registor clocked at - say - 320Hz and a bunch of resistors on the taps of the shift register summed into a low pass filter to get rid of the 320Hz edges. Not a perfect brick wall, but - with close tolerance resistors and the right window function - pretty good.

Three sets of resistors summed into three separate low pass filters - all driven by the same 32 taps - could give you the three phases. I've done something similar. It isn't a cheap filter.

Getting a micro or a pld to do more of the work would be cheaper and easier.

-- Bill Sloman, Nijmegen

w

man.

11.3 MHz =3D 256 x 44.1KHz, i.e., 256 x the Nyquist sampling for a 22KHz signal. 32x oversampling x 10Hz x 2 for Nyquist =3D 640 Hz. Might not even need the extra x2, I'm not sure offhand.

FWIW, 3.3MHz / 640Hz =3D 5,156--almost 4 orders of magnitude.

-- Cheers, James Arthur

You'll need a 256-entry 10-bit sinewave lookup table, right? Won't that be difficult to fit into a 128 macrocell part?

-- Cheers, James Arthur

ry in

here

f the

esign

hink

, PICs

d

han

also

u

Thanks for the feedback Joel. ARMs are overkill for most of what I envision--it seems like a sin to waste all those extra bits!

James

I've used the HC05 and HC11 families. Nice code, but lots of clocks per instruction would be one of my gripes.

I dimly recall Motorola trimmed that with a pipelined version, starting with the 'HC08 was it?

-- Cheers, James Arthur

Yup. CPLDs don't generally have block rams.

A dollar or two of uP is the obvious way to generate polyphase sine waves at 10 Hz.

John

I was sloppy--it's actually slightly worse. He's talking about outputting samples at 20Khz, which is 2,000 samples per 10Hz cycle.

Even storing a 1/4 wave table, that takes 500 10-bit entries, not 256.

James

Depends whether or not you care about injecting noise on the floating terminal. Rheostats in high current stage lighting wouldn't care.

In a low noise circuit or at RF it would be bad news.

His statement was ambiguous and I don't rate him as an engineer.

Rheostat just isn't used as a word any more here. Same with condenser. Unless you really are talking about expensive high power wire wound components for repairing seriously geriatric kit or antique radios.

If you are talking about precise definitions of English language but the word just isn't used any more. It's a variable resistor or pot with a high power rating if it ever comes up in a conversation.

Rheostat really sounds antediluvian over here forsooth it does.

Out of 5067 pots. A whopping 3% (that *is* higher than I expected). And they are mostly US stock with a surcharge for delivery. If you click the button to exclude US Stock "extended range items" you get only 4 in stock. My preferred supplier RS has 48 stocked out of 3251 pots

There's quite a few plug-ins for it, though, including some for image editing. Ever taken a look?