120 Degree Phase Shift Osc.

Photoshop Elements or PSPro X will probably do what you want and have modest prices and shallower learning curves cf Full Photoshop.

You can usually get the N-1 th version at a huge discount. So unless the new one has something you *must* have buy one version behind. The older version will also have more bugs already fixed.

Regards, Martin Brown

If

new

ue

p,

ec

oman.

n d

This ignores clock speed required to do the pulse-width modulation, which does suggest that you didn't read the post in any kind of detail.

I was expecting someone to query my choice to do the pulse-witdth modulation with 10-bit - 0.1% -resolution - the quantisation noise introduced by using a slower clock and coarser quantisation would also be attentuated by a low pass filter, if the noise were random when in fact it is totally predictable, and would repat exactly from cycle to cycle.

-- Bill Sloman, Nijmegen

--
Now you're trying to pull a Sloman!

My reply was in response to your: 

"I am now more mystified than ever as to whether the term rheostat
still has currency in electronics across the pond."

and proved that the word does have currency on both sides of the pond.

Your thrashing about does nothing to disprove that.

Not if you synthesis the sine wave by exploiting the relationships

d sine x/dt =3D cosine x and d cosine x/dt =3D -sine x.

and set up a finite difference approximation to sine x and cosine x. You'd need an accumulator that was at least 14-bits wide to synthesise a stream of 10-bit sine and cosine values and you'd need to bias the starting values within the "guard" bits to prevent rounding errors accumulating to the point where they were observable in the 10 most signficant bits. The "average" rounding error per accumulation is the LSB divided by root 12, 0.287 bits, and if you treat this is as a random error (which it isn't) you'd expect an accumulated rounding error of 13 bits over 2000 accumulations

When I used this technique for the digital synthesis a hyperbolic function, I modelled the rounding error in Fortran, but today I'd do it in Excel. You'd need to reset the accumulators at at least the 360 degree points, and this is easier if you use third register to keep track of the number of operations, but in principle you could reset (to eight different pairs of values) every 45 degrees using a digital comparator that kept track of the states of both accumulators.

It was a well-known technique when I was young, and when I had to do a

8051 programming course, this was the process that I elected to program.

-- Bill Sloman, Nijmegen

em

aw

c

on

e h d

gh

d d l e

Not if you know what you are doing. Finite difference approximations can be handy.

-- Bill Sloman, Nijmegen

m w n h

Sure, that's fine, it just seemed like you were proposing to do a DDS.

I thought of synthesizing the waveforms in the PIC similarly, maybe using Bresenham's techniques, but I was too busy being hen-pecked about providing exact part numbers and dynamic power consumption and such to want to get into that.

Of course if you have a uC you might as well just use a table--it's easy, and free.

To get the 3 waveforms spaced at 120 degrees I think you'd need three generators, wouldn't you? A 120=BA delay line would take too many flip- flops if implemented as a shift register.

-- Cheers, James Arthur

In terms of power efficiency you do not want sine waves driving the relays.

If you did want sine waves then one approach would be a "bubba oscillator" but with 120 degrees per stage instead of 90.

Tim.

--
5 only if you do it in hardware:

news:cpoqd65emr24k3771dp1cr36knp9bhdfvd@4ax.com

--
He doesn't want to use them as relays, he wants to excite the coil
with sinusoids in order to get a smoothly rotating magnetic field.

You only need to generate a pair of 120 deg phase signals A, B as the third phase can be generated by the linear combination C = - (A+B)

Regards, Martin Brown

Sorry, I couldn't follow the link, but, come to think of it you don't need shift registers, just two extra 10-bit registers clocked at 120=BA intervals.

-- Cheers, James Arthur

w

eem

a

raw

ic

w

ion

ve

th

nd

ugh

ld

nd

el

, t

me

It's not incompatible with DDS - though obviously not as flexible as a systemt with a single angle counter and look-up-tables.

-

Probably. You only need to save the increments out to 60 degree behind the main generator, but even that's 334 samples, which wouldn't fit in a cheap pld. Three generators is a lot less macrocells.

-- Bill Sloman, Nijmegen

Yes, there's that too. Kewl.

-- Cheers, James Arthur

--
Yup, you can't access binary newsgroups from google groups

Can you explain that? It makes no sense to me.

John

If you can use a CoolRunner driving DACs to make a usable 10-bit,

3-phase sinewave generator, I'd love to see it. Even a logic sketch and a rough cell count would prove the point.

Do you know what you're doing?

John

Two generators would give sin/cos, from which the 120 degree things can be done by analog mixing post-DAC. A double-integrator sine generator gives you cos for free, so you only need one algorithm. Some of the CoolRunners *might* have enough horsepower. Tbis would of course not be DDS as it's generally understood.

FPGAs, on the other hand, make this sort of thing easy.

We had a lot of fun doing that one. Brought back all the university signals+systems+math stuff, not to mention high-school trig. Surprisingly, the hardest part was the post-DAC filters and amplifiers, keeping things flat to 32 MHz (half of Nyquist!) with real parts, dancing around aliases, and keeping the THD down.

John

No, I mostly crop, tweak colors, and resample. I leave the more artistic things to minions.

John

rote:

(PST)) it happenedBillSloman

ote:

wrote:

llSloman

,BillSloman

DT),BillSloman

y f

ew

e , c

oes

but it clearly shows that 20MHz is way overkill for a 10Hz sinewave

The coils them self will also filter the signal, whats the induction of a 500Ohm relay coil? you might not need a filter at all

I didn't mean you should try to implement a brickwall filter, it was meant as a thought experiment

-Lasse