Hello all

I am facing a problem, I have a feed back signal from a sensor that will "travel" by a cable in a noisy 50hz environment.
I am looking for a solution where no microcontroller is involved, well, at least no software involved.
The solution will be integrated on two pcb
My guess is that that the signal will pick up a lot of noise and interference from outside the cable (ordinary flat cable) so therefore i do not want to feed the receiving with my fairly high impedance feed signal side.

My idea was first to use a PWM set but all they all suffer from either high prices for the chips, a lot extra components and a 2 order RC filter on the receiving end, causing long response time and still have ripple.

My second thought was to isuse VCO, Voltage control Osc and a Osc controlled Voltage but they all suffer from high prices for the chips.

So my idea is to connect an ADC directly with an DAC, but i can not find a good solution, I want the ADC to continuous feed the DAC without any control logic involved. (ADC101S021 seems to do the ADC trick, no protocol or i2c/spi commands) just feed it with a clock and it output data)


Is there already a good solution to this problem?
What is the best way to do this ?
Is there any tips of self operating ADC/DAC that can do this?

 

hi Stefan,
What is the specification/parameters of the signal you are planning to TX.
ie: Frequency range, amplitude etc.
Cable length?
E

 

You've left a lot of important things unspecified or extremely vague.

You talk about causing long response time, so presumably response time is important. But what response time do you require?

You talk about a concern over it still having ripple. But what level of ripple is acceptable?

You talk about high prices being a concern, but give no clue as to what kind of budget is acceptable.

You give no indication of what kind of signal this is or how well you need to be able to recover it at the receiving end. Recovering a 100 mV signal with 0.1% accuracy is far different than recovering a 10 V signal with 1% accuracy.

If you go the ADC->ADC route, can you tolerate the latency involved if you can't tolerate the processing of a PWM signal?

Why are you so resistant to using a microcontroller? Simple ones are dirt cheap.

 

What do you mean by "feed back from a senior" ?.
What is the sensor ?
Are you fixed upon the cable ?
And all the above questions .

In general
Differential is better at noise rejection
If single ended , screening is good and watch for the ground return .
If you can amplify the signal at the sender and attenuate it at the receiver , external noise is also attenuated.
Also in general , one can filter the noise if it's outside your band of interest.

 

As above, we do not know anything about your requirements for the signal or signal fidelity. You're direct, A/D-to-D/A approach can work, but you still might need a reconstruction lowpass filter at the D/A output. This depends on the ratio of the A/D clock frequency and the frequency and shape of the analog signal. Without it, there will be "stairstepping" in the output signal. That is a form of distortion caused by the quantizing of the original signal by the A/D.

Cable distance?

ak

 

noisy 50hz environment

How often do you need to sample the data? If you need the data very slowly, then adding a low pass filter to remove noise above 10hz will help. But if you need the data without delay a low pass filter will not work.

Normally I send the analog data and a reference voltage down two different wires. On the receiving end I use a differential receiver that subtracts the two signals. Often the noise will be the same on both wires and will be subtracted out. OR. Send the signal and the inverse of the signal down two twisted wires. The differential receiver will hand you 2X the signal with the noise removed.

 

What type of cable is used for the sensor and how long is it?
If you're dealing with high resistances, the requirements for the cable are even higher.
The cable must have a good quality shield, and the shield is connected only at one end.
In some cases, you have to use a galvanic isolator, but they aren’t cheap.
To give some specific advice, you need to describe the circuit in more detail

 

Another approach is to filter out the 50 Hz noise with a comb filter at the source.

ak

 

Another approach is to filter out the 50 Hz noise with a comb filter at the source.

ak

At the source?

How will that filter out 50 Hz noise picked up in the lines between the source and the receiver?

 

How will that filter out 50 Hz noise picked up in the lines between the source and the receiver?

Put the filter on the receiver end.

 

I was thinking of one at the input to the A/D. Re-reading post #1, I see an alternate meaning to his description, that the noise is more radiated than conducted.

ak

 

If you‘re running cable add a couple of extra cores for a power supply, and amplify the signal at source, use a balanced line and make sure that the amplifier can drive the line. At the other end receive with a differential line receiver such as INA500, with built in matched resistors.

And disregard all this advice is your signal is not analogue between DC and 100kHz.

 

Hello all
Thank you fro all response.


The basic problem is to transfer a analogue signal (like 0-1V) through a ordinary flat cable through a noisy environment.

It is actually a feedback signal for a DC-AC converter so the the sensor is a photo transistor in a opto-coupler.
Which is close to the 120-230V outlet and the cable will run along the transformer up to the sine wave generator and
is used to correct the outgoing voltage.
This mean the the photo transistor output and the cable will pick up noise and be affected of the 50/60 Hz interference and make it more difficult for the sine wave generator to track and correct the outgoing voltage.


> You talk about a concern over it still having ripple. But what level of ripple is acceptable?
Very little, it is a feedback signal used for voltage correction, so the relationship is almost 1:200
feedback signal 1v , tracking a 230V

> You talk about high prices being a concern, but give no clue as to what kind of budget is acceptable.
less than 4€ or complete solution, the simple solution is often the cheapest ^^.

> Why are you so resistant to using a microcontroller? Simple ones are dirt cheap.
It involves supporting software, track it, store it and maintain it, just a hazzle for a simple problem.
An "off the self solution" would be nicer. But even if you use a CPU for ADC and DAC, you still have the flat cable connecting them.

> How often do you need to sample the data?
Fairly often, it is a feed back signal for a sine wave generator, I want to avoid self induced oscillations.
it regulate a 50/60 Hz signal so 500 times a second should be enough.

> You give no indication of what kind of signal this is or how well you need to be able to recover it at the receiving end.
> Recovering a 100 mV signal with 0.1% accuracy is far different than recovering a 10 V signal with 1% accuracy.
photo transitor of a opto coupler so 0 - 1V, there is space for an op-amp.

ronsimpson and drjohsmith Is on something, a sort of balanced solution.

> I send the analog data and a reference voltage down two different wires.
> On the receiving end I use a differential receiver that subtracts the two signals.
> Often the noise will be the same on both wires and will be subtracted out.
> OR. Send the signal and the inverse of the signal down two twisted wires.
> The differential receiver will hand you 2X the signal with the noise removed.
Sounds like a winner


Another idea I am testing is to use a GreenPak chips (these chip cost only some 1€ each) from Renesas and simply make an ADC-SPI-SPI-DAC solution. Run it at 2MHz could easily give an update frequency of 30000 samples per sec and be sufficient tracking speed.
Perhaps with balanced lines mentioned above (GND, VDD 2xCLK and 2xMISO) and with termination caps like in LVDS to get ride of DC offset.

But
> OR. Send the signal and the inverse of the signal down two twisted wires.
Sounds very simpel

Regards

 

if this is the feedback for an inverter, is this the signal that sets the amplitude of the output? If so, I’m guessing that it gets integrated over a number of mains cycles? If so, the intergator will remove any differential noise, and all you have to worry about is common mode noise, and that’s reasonably easy to remove with a differential amplifier with decent CMRR.
To ask the obvious question. Why a 0-1V signal, why not a 0-10V signal, or even the whole 230V sinewave?
[edit] I would avoid messing about with the signal too much in case you introduce phase shifts that might send the whole thing unstable.

 

Another idea I am testing is to use a GreenPak chips (these chip cost only some 1€ each) from Renesas

If you use these exotic devices, they also need some kind of programming like a simple µController. Therefore this also means for you:

It involves supporting software, track it, store it and maintain it, just a hazzle for a simple problem.

From that point of view, there is no difference between a µC and the GreenPak. Therefore I would prefer a mainstream Cortex M0+ controller like the STM32L0x2 series with ADC and DAC (or many other similar controllers from other manufacturers). They are usually in the sub-€ or sub-$ range, so no cost advantage for the GreenPak. Configuring and programming of STM32 controllers is quite comfortable due to extensive support by tools like CubeMX and CubeIDE. Simply read out the ADC and send it via serial interface to the other controller and there write the received data to the DAC.

 

But
> OR. Send the signal and the inverse of the signal down two twisted wires.
Sounds very simpel

You had specified in your original post that it was flat cable that was to be used.

If you can use twisted pair (and they make, or at least used to make, ribbon cable that consisted of adjacent twisted pairs) that, alone will make most of the external noise disappear, even on a single-ended signal.

Try using a twisted pair in the current implementation and see if that solved things. If not, amplify your signal by something like 10x and/or convert it to a differential signal.

Also, is this a solution looking for a problem?

Are you asking for a solution to a demonstrated problem in an existing system, or are you envisioning things that might be a problem in a system that hasn't been implemented yet and trying to eliminate them up front?

 

If it is a FEEDBACK signal, then the only place it is going is to an error amplifier. All error amplifiers have a pole at f=0, so will filter out any hf Interference.

 

Without the requested information there is no reason to spend any effort towards a solution to the TS problem. The right questions are asked in post #2, although I would ask also as to if the signal is "ground referenced", or is it "differential" with no "ground" connection. That makes a very big difference in the aproach.

 

Actually, the TS stated that " It is actually a feedback signal for a DC-AC converter so the the sensor is a photo transistor in a opto-coupler." That is a moderately low impedance, hopefully not "gounded" at the sensor end. That matters a whole lot!!!
The best plan, for an UNSHIELDED cable, is to use an isolated pair of conductors in close proximity to each other, and only connected to the sensing signal circuit. Most of the noise will be coupled into both conductors, and if both the voltage source and amplifier input are similar impedance values, much of the noise will cancel. That is a "cheating trick" from many years ago. The TS may need to do some experimenting to achieve the same voltage source inpedance as the amplifier input mpedance, but it should be quite possible. With a good match the noise voltage in the two wires should cancel each other.

 

You can deal with “grounded at the sensor end” by running a twisted pair, and inserting a resistance to ground and an equal resistance to the signal. That gives you a balanced (but not differential) signal. Any differential line receiver or other differential amplifier will happily amplify the difference (the signal you want) and ignore common mode noise which should be picked up equally in both wires of the twisted pair as they are now both at the same impedance.