# Phase shift

Discussion in 'Electronic Basics' started by Ahmad, Mar 3, 2008.

Hello ,
i am trying to combine or mix three signals with the same frequency
together in one channel. i have no idea hw i can do that. The signals
are sin waves. I think if i chnaged the phase of each one and then
combine them together it will work. but i have no chip to circuit to
implement that. could u plz suggest me something
this is my email also

thanks

2. ### Bob MastaGuest

It's not at all clear just what you are trying to do.

If the 3 signals are sine waves of the same frequency, they will add
up vectorially to produce a single sine of that same frequency.
Depending on their relative phases, the amplitude of that sine
may be anything from zero to the sum of the input amplitudes.

If that maximal amplitude is what you are looking for, you have to
insure that all 3 waves are in phase. If they do not originate from
the same source, it may be rather difficult to adjust the phases since
they will tend to drift. If they *do* originate from the same source,

but have passed through independent phase-shifting paths before you
sum them, then adjustable phase shifters may do the job.

Perhaps if we knew what your ultimate goal was, we could suggest
the best approach. For example, if you are trying to measure total
amplitude, it might be simpler to rectify and filter each sine
separately and then sum them as DC.

Best regards,

Bob Masta

DAQARTA v3.50
Data AcQuisition And Real-Time Analysis
www.daqarta.com
Scope, Spectrum, Spectrogram, FREE Signal Generator

3. ### Tim WescottGuest

Why do you wish to do this, and in what way do you wish to combine the
signals?

If you are planning on simply adding them, and you want to use them to
communicate three different channels, then you'll find you're out of
luck. With a perfect phase reference (which generally requires a second
channel anyway) you could put two quadrature signals on a wire and
decode them at the other end - but you'd need that second channel.

Could it be that you have three signals that fall into the same
frequency _band_ that you wish to multiplex onto one wire? If so, tell
us and we'll have some suggestions. If not, tell us what you're really
trying to do.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

hello,
Actually, i have 3 signals as a sine wave from different sources as a
measured quantities from 3 different sensors. i want to transmit these
signals in the same channel. if i just added them it will be added
together. what i want is to tranfere these signals in the same
channel. Is that possible to do by delaying (adding different phse
shift) each signal with different time delay. Knowing that i don't
want to use analog multiplexer (time division multiplexing). If the
phase shift is ok...do we have a circuit to do that except (All pass
thanks

5. ### Tim WescottGuest

That's not possible. Phase shifting only gives you two degrees of
freedom, and then only if you have a solid phase reference.

You need some form of multiplexing. I'm not sure why you don't want to
use time division multiplexing, but it may be the way to go; if your
signals aren't too high in bandwidth and your accuracy requirements
aren't too severe then it's not that difficult. The range of
possibilities are endless; two analog methods that spring to mind are
frequency-division multiplexing, by heterodyning your signals up in
frequency at the transmitter and down at the receiver, or
amplitude-modulating some carriers with your signals at the transmitter,
then filter/detect the signals at the receiver.

This is 2008: the best way to do this in many ways is to encode your
signals digitally as close to the sensor as you can, then send them over
a network. Even if all you do is regenerate them back into analog at
the other end, digital is getting easier and easier with respect to
analog every day. Without knowing your situation I can't insist that
this is the best way, but you shouldn't discard it as a possibility.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

According to your answer i have to explain my case in more details.
I have a 4 x 4 array of capacitive sensors. I have already used
Frequency Division Multiplexing (FDM) with different frequencies to
excite the capacitive sensors for each colomn. The output of the
sensors are the common rows of the matrix. Therefore, each row in the
matrix consists of 4 signals with different frequencies. The output
signals now are AM modulated signal which have frequencies as the
carrier frequency (1 Mega HZ to 4 Mega HZ). At the output side, i want
to combine all of the signals of each row just in one wire in such a
way that transmitting the signals will be easier and less hardware i
can use. I am thinking of adding some time delay in each row different
from the other rows. Another idea i have is to do AM modulation again
(frequency shift) to higher frequencies. But the later choice is
little bit difficult because i have already high frequency signals and
i don't want to go higher which will cost me more expensive hardware.
i am now stucked at this point. otherwise, i have to make for each row
its own detection and ADC circuits. For higher order arrays, this will
be difficult and more hardware i have to use. On the other hand, using
time multiplexing and analog multiplexer will affect the bandwidth of
the sensors and decrease the measuring time therefore, less SNR. Hope
this is clear and hope to get an idea for this. If the phse shift will
work, plz let me know if any chip is available with these high
frequency signals.

7. ### Tim WescottGuest

There are very few commercially available chips that are designed to do
things that are technically impossible. Those that are designed so
usually aren't commercially successful.

work. If you don't believe me, get out a pencil and some paper and
study the behavior of sine waves.

I suspect that the bandwidth of the actual signal you want is
significantly lower than the bandwidth of your carriers. You may find
it very worthwhile to find a way to reduce the carrier/bandwidth ratio;
this will allow you to pack more information into a pair of wires that
are good to some maximum frequency. You may also find that the best way
to do this is to move your ADC/demodulate step close to the sensors, so
that you can send the data digitally for the long distance.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

8. ### Sjouke BurryGuest

I beleive that the phase shift is not working.if the delay is
small..but what i am thinking in that to delay the signal with a time
equal or little bit more than the bandwidth of the first signal...it
is just kind of Time Division Multiplexing (TDM) but without loosing
any information from the first signal by using the traditional analog
multiplexing techniques...using the ADC in the early stages will not
be the best choice..because the amount of the hardware which should be
used...regarding the demodulation..i wanna to use DSP processing to
recover the signals and not the traditional way in using
demodulator....

10. ### Bob MastaGuest

The simplest solution (given what you have so far) would seem to be
time-division multiplexing, where you connect each row to the output
line in sequence. The problem then is telling which position the
multiplexer switch is in at any given moment. You could add
a DC signal (or another frequency) that encoded the switch position.
(Or, if you insist on doing something with phase, you could phase
shift each row differently... ugh!)

There are capacitive array sensor chips that are designed for
touch panels and the like. You won't get continuous output from each
sensor, but rather the chip's estimate of where the touch point is.
If this is what you need to do, the interfacing will be much simpler.

Best regards,

Bob Masta

DAQARTA v3.50
Data AcQuisition And Real-Time Analysis
www.daqarta.com
Scope, Spectrum, Spectrogram, FREE Signal Generator

11. ### whit3rdGuest

OK, that's easy. First, you need to isolate the combined signals
from the sensor matrix outputs (so, this is a job for ... an
amplifier).
The combination can be a simple summing junction (four amplifier
outputs connected to the summing point).

Four one-transistor amplifers, with their collectors all joined
to a single pullup resistor, and some base and emitter biasing, will
do it. If you wish, a signal transformer at the collector can be
useful
and at 1 MHz needn't be an expensive component).

12. ### Jon SlaughterGuest

If they are true sin waves then you could just take the amplitudes and put
them through an orthogonalization process.

f(x) = A1 + x*A2 + (3*x^2 - 1)/2*A3 is unique.

This is the legrendre orthogonal polynomials.

You basically sweep through x from -1 to 1 sending f(x) as the signal.

You can then recover A1,.., A3 by the orgonalization properties. (requires
an integrator)

This is not necessarily the best way but I think it could potentially be
done with just passive components and maybe a few transistors and/or op
amps.

Probably easier just to frequency shift the signals and then filter them
back out or simply time domiain multiplex them.

You won't be able to simply add them directly because there is no way to
exact them back out of the sum. Its like if you take 3 + 3 = 6 then try to
recover what was added to get 6. It doesn't work cause there are an infinite
number of ways to do it and not just 1.

I think though simply exacting the amplitudes(pretty easily done) and send
them in some way, even a asynch serial like protocol would work fine.

For example, simply send the amplitude of the first, then ground the line,
the amplitude of the second, ground the line, and the amplitude of the
third, then ground and repeat.

_____
| | _____
_____ | | | |
| | | | | |
---+ +---+ +---+ +---

The only issue is if they get out of sync but this I think you could throw
something together with minimal components here. (just a mux/demux, a clock,
and some passives)