1. Home
  2. Electronics Design
  3. 120 Degree Phase Shift Osc.
  4. Page 3

120 Degree Phase Shift Osc.

uld

Except that you then have to filter the PWM waveform. which means extra phase shift in the filters. And 10Hz filters use tolerably big capacitors, which makes it difficult to find 1% tolerance parts.

DDS is a rather better way to go. You still have "stair-case" noise on the DAC outputs, but it's at a much higher frequency (depending on the width of your DAC), so much less phase shift in the filters and a better chance of being able to buy close tolerance capacitors for the filters.

Analog Devices parts may cost more than PICs, but they do do most of the job for you (filters excepted).

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman
Loading thread data ...

e:

k

tion.

I am

y to

rt from

age,

uch

he

ough

e four

s.

f

ves

Hill

d it

S

-dds...

a

gic

Hz

red to

degree

3,4
0).

three

ch,

4

er,

art of

ts

on of

n PWM

ving

=A0 =A0 =A0Vcc

=A0 =A0 =A0 =A0 =A0 =A0 |

[DIODE>]-+ =A0

=A0 =A0 | =A0 =A0 =A0 =A0 =A0| =A0

-[COIL]--+ =A0 =A0

=A0 =A0 =A0 =A0 =A0 =A0| =A0

DE>]-+ =A0

=A0 =A0 | =A0 =A0 =A0 =A0 =A0| =A0

=A0 =A0 +--[COIL]--+ =A0

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0| =A0

--+-[DIODE>]-+ =A0

=A0| =A0 =A0 | =A0 =A0 =A0 =A0 =A0| =A0 =A0 =A0 =A0 =A0 =A0

=A0 +--[COIL]--+

|

=A0 =A0 =A0 =A0 =A0 =A0 =A0

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0

John Larkin practising the irrelevant comment defence. If you can't beat them, change the subject.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

He isn't, but regular reeds are good for 10 million operations, which is about ten days at 10Hz, and mercury-wetted reeds are good for 100 million operations, which is roughly three months.

The fact that he wanted a sine wave drive was a clue to the fact that he isn't actually interested in relay closure, as I obliquely hinted in my comment early on in the thread.

Reply to
Bill Sloman

te:

=A0 =A0 =A0 =A0 =A0Vcc

| =A0 =A0 =A0 =A0 =A0 =A0 |

-+-[DIODE>]-+

=A0 =A0 | =A0 =A0 =A0 =A0 =A0|

--[COIL]--+

=A0 =A0 =A0 =A0 =A0 =A0|

DIODE>]-+

=A0 =A0 | =A0 =A0 =A0 =A0 =A0|

=A0 =A0 +--[COIL]--+

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0|

---+-[DIODE>]-+

=A0 =A0| =A0 =A0 | =A0 =A0 =A0 =A0 =A0|

=A0 +--[COIL]--+

|
3-
p
8

at

All it needs is the right shift register, a scheme for intialising the contents of the shift register - as you pointed out - and (for people like me who believe in Sod's Law) a scheme for detecting when the shift registor contents go wrong and correcting them.

In a micro, its a watch-dog timer ...

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

em

s

could

n

If the PWM is 20KHz, the filters are trivial.

-- Cheers, James Arthur

Reply to
dagmargoodboat

If the PWM is 20KHz, the filters are trivial.

-- Cheers, James Arthur

_______________________________________________________

And since the phase shift is the same for all three outputs, that's no issue. Plus, if three op amps are used, the drive buffering can be tailored to whatever the application requires.

tm

Reply to
tm

A sine wave is not exactly the best drive for a relay coil; the lower the frequency, the worse the pullin and dropout will be. Three One-shots in series loop would do the job..

Reply to
Robert Baer

On a sunny day (Fri, 5 Nov 2010 19:03:50 -0700 (PDT)) it happened Bill Sloman wrote in :

You do not need that, the PWM would be kHz, so 10 Hz needs only a simple fitler. The phase shift introduced by the filter would be the same for each channel.

The minimum component config I can come up with is a 20 pin PIC, using 3 8 bit output ports, with each 8 resistors in range 1:2:4:8... summing into one point. Soft reads with three pointers from sinewave table. RS232 frequecy controlled, and phase controlled too if needed. Needs some transistors to drive the coils if much voltage is needed, else none. Total cost < 4 $ excluding hardware like pcb, case, connectors. If you have the money for more resistors or only a few values available use R2R ladders. Some decoupling caps too.

Reply to
Jan Panteltje

Generate and buffer a sin(wt), cos(wt) quadrature pair and then use the identity cos(A+B) = cos(A)cos(B)-sin(A)sin(B)

summing them with B = 2pi/3 Hence your three phases are

cos(wt) (cos(wt) + sqrt(3)*sin(wt))/2 (cos(wt) - sqrt(3)*sin(wt))/2

It is left as an exercise for the reader to find resistor ratios for

0.500 and 0.866

Regards, Martin Brown

Reply to
Martin Brown

TM, I wish you'd learn to post. Your replay doesn't go after the sig separator.

Even if they aren't, it's a simple matter of feedback and programming.

Reply to
krw

ladders.

--
How about this: ;)

                                         
                        
                      +---------+       +---+  +---+
                    +-|256x8 LUT|--/8/--|DAC|--|LPF|-->PHASE 1
                    | +---------+       +---+  +---+
       COUNT        |                  
      +-----+       | +---------+       +---+  +---+
CLK>--|>   Q|--/8/--+-|256X8 LUT|--/8/--|DAC|--|LPF|-->PHASE 2
      +-----+       | +---------+       +---+  +---+
                    |           
                    | +---------+       +---+  +---+
                    +-|256X8 LUT|--/8/--|DAC|--|LPF|-->PHASE 3
                      +---------+       +---+  +---+
Reply to
John Fields

On a sunny day (Sat, 06 Nov 2010 11:13:19 -0500) it happened John Fields wrote in :

R2R ladders.

Yes, I did propose that with a 555 for clock, a 4040 counter, 3 EPROMs, and 3 R2R DACs, see one of my previous posts. But I am not sure about EPROM prices these days, 2716 was an old small one :-)

What you propose here can be done in software too, if you use a PIC that has 3 8 bit output ports, and use an interrupt tick as clock, at each tick increase a pointer in a sine lookup table for each output, pointers 120 degrees apart, send value to ports. Sort of a circular buffer idea in the lookup table. That way you use only one chip. There are many ways to skin this cat so to speak.

I think what Bill Showman does not realise, is that even with his DDS or whatever, he still needs a micro to program it.

Reply to
Jan Panteltje

uld

een

It's only the same if the filters are identical. "Close tolerance" capacitors are +/-1% and you can only get them in a restricted range of capacitance.

Meaning what?

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

seem

his

t could

een

Depending how you do the pulse width modulation. The temptation is always to divide the 100msec period into the smallest number of discrete periods you can get away with, and have the output either high or low for as much of each period as you need to get the right average output voltage. The average cheap micro has usually got built- in timers that make this easy

You can do a lot better by having one accumulator for each output which mimics the output voltage and comparing it with an accumulator that accumulates a finite diffeence approximation to the desired sine wave output - the output of the digital comparator sets the drive high or low for the next 50usec (or whatever) and increments or decrements the mimic accumulator appropriately, but the last time I put that forward, we ended up needing to use a transputer (it was a while ago) to handle all the arithmetic involved in the time available. Programmable logic is usually quicker, and allows you to set up real parallel processing, but it is even lower level programming than assembler, and the software guys wouldn't even think about it.

Of course this switching the drive outputs more often, and the switches get correspondingly hotter. At 20kHz this isn't often a problem, but 20kHz is just fast enough not to give an audible buzz, which may not be compatible with the customer's idea of an adequate approximation to a sine wave.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

nnel.

umming

se none.

use R2R ladders.

  • =A0+---+

-->PHASE 1

=A0+---+

=A0+---+

-->PHASE 3

  • =A0+---+

e :-)

whatever,

DDS chips do need a serial bit stream to load program, but you don't actually need a micro to generate that bit stream - serial PROMs (of various sorts) have been around for years to handle this and similar jobs (setting up the classic Xilinx programamble logic chip comes to mind).

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

What you have here is the op-amp equivalent of the well known Scott-T transformer. These have been used for years to convert two-phase resolver signals to three-phase synchro signals.

--
Virg Wall, P.E.
Reply to
VWWall

none.

R2R ladders.

--
Around $3 for On Semi 2KX8.
Reply to
John Fields

On a sunny day (Sat, 6 Nov 2010 10:08:39 -0700 (PDT)) it happened Bill Sloman wrote in :

I have not used DDS chips, so, but does it provide a clock for that? Still, even if it does, that makes 2 chips!

Reply to
Jan Panteltje

First generate the two required frequency sine waves in quadrature. Then use the op-amp equivalent of the Scott_T transformer to convert these to the desired three phases.

You can even find the real transformers if you want a passive solution.

--
Virg Wall, P.E.
Reply to
VWWall

On a sunny day (Sat, 06 Nov 2010 10:49:00 -0700) it happened VWWall wrote in :

10 Hz means big transformers.
Reply to
Jan Panteltje

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.