I've been reading about the separation of analog and digital circuitry in an effort to improve the noise situation in a project I've been collaborating on.
We have many analog circuit devices: 8 analog pressure transducers and 6 analog thermocouples being processed by 4 ADCs.
We also have many digital devices: 8 BJTs controlled by a microcontroller, a load cell which transmits digital data to another microcontroller which also receives data from the ADCs.
It seems based on this article that the best we can reasonably do without redesigning much of the circuitry is to star ground by supplying analog and digital circuitry with separate power supplies that are joined at their respective ground pins.
My question is whether this is going to have an appreciable effect on the behavior of the analog portion of the system or not given that eventually this system will be deployed in a manner that forces the power source to be:
A gas generator powering a power strip hosting the power supplies for this circuit. Nothing can be done about this.
Without knowing what really goes on inside a power supply as far as filtering goes, my guess is that it would be standard to perform some amount of filtering that makes the power supplies behave independently as desired. But I lack the specific knowledge to be sure.
The microcontrollers are indeed Arduino MEGAs, sorry that was omitted from the first post.
But power delivery requirements exceed the maximum rated currents of these two boards, so we have to use external power supplies. The pressure transducers in particular cannot draw power directly from the Arduinos. The total draw for everything (including the MEGAs and the RPI they communicate with) is approximately 5A.
If I understand correctly, you are saying that if I use the MEGA as the star ground, I would be correctly separating the circuits?
Assuming that beyond the power supplies everything is separated properly and the supplies themselves are the star grounded are the benefits from that configuration lost by using 1 power strip to power both power supplies?
Separating digital an analog power supplies is a good idea (if indeed they can be separated).
Power cables can have ferrite beads added to them to stop noise from radiating, a bead at the input terminal to a block is also a good for noise coming from the block.
Grounds should be connected a one star point.
A/D converters must use a voltage reference, not the Vcc voltage which is the default in the Arduino.
ex: the Arduino Reference pin.
If time allows, a running average can reduce noise effects.
60Hz rejection on analogs can reduce any AC interference.
Example, a dual slope A/D converter adjusted to reject 60Hz.
I am also confused. By my understanding AREF is a way to adjust the scale range of the ADC inside the Arduino. I am using 4 external ADCs communicating values over I2C to the Arduinos. What are you telling me to do with the AREF pin?
This is really a question about filtering circuitry inside power supplies, I think. Whether the output of a 'digital' power supply is pure enough to be considered 'separated' from an 'analog' power supply, when they draw from the same source AC power.
We have implemented running averages in the past, and there is some improvement. I am looking into how feasible it is to do more, essentially.
How would one adjust an ADC for frequency rejection?
When we talk about Arduino, the default reference for its A/D converter is the PCB 5 volts.
5v is full of switching noise and this will effect your readings.
You can use the controller 1.1v internal reference or a good external voltage reference instead of 5v. This will reduce supply noise from affecting your readings.
See: analogReference() - Arduino Reference
If you are having 60 cycle inductance problems on the Arduino, you can average two successive readings separated by 8.3ms to cancel out 60Hz.
i.e the first reading with induced 60Hz is cancelled out by the second reading’s induced 60Hz.
By power strip I think you mean the mains power into the power supplies.
The power supplies should completely isolate the mains input from their DC output, so there is no connection between the outputs of the different power supplies other than whatever connection you make. You should create a star grounding point close to the electronics and connect the negative (usually) from each power supply to the star point, then from the star point to the respective loads.
Separately from that is the issue of the safety electrical earth. Depending on the power supplies this might not be connected at all or might be made available for you to connect as you wish to either the positive or negative output of the PSU. Connect 1 wire from the mains safety earth to the star point.
If you have a metal case or other metal structures as part of the project they too should be connected to the star point and no where else.
As for what to use for a star point, a brass bolt with brass nuts and solder tags for the wires works well. If you have a metal case the attach it to the case.
Yes, thank you. I lack(ed) the terminology to properly express that.
The power supplies should completely isolate the mains input from their DC output, so there is no connection between the outputs of the different power supplies other than whatever connection you make.
This is what I was hoping for.
Separately from that is the issue of the safety electrical earth. Depending on the power supplies this might not be connected at all or might be made available for you to connect as you wish to either the positive or negative output of the PSU. Connect 1 wire from the mains safety earth to the star point.
I have power supplies which look more or less like this. If I understand correctly you are saying that I ought to connect the ground terminal to the star ground point?
Discussing separate power strips is pointless. Discussing separate facilities circuit breaker lines
is another thing. That being said , forget about the power strips and just use separate switching
power supplies and have a common ground (star is good). Beyond that , sufficient filter caps for
both analog and digital circuitry is recommended.
The schematic doesn't help much because the device labels are unreadable.
The lack of any filter caps anywhere begs the question "How much circuit design experience does
the designer of this circuit have ?"
Do not connect the ground terminal on the power supply to anything.
The "ground" we are talking about here is the negative of the supply, and it has incidentally confused some people that the power supply has three positive and three negative terminals. In each case, all three terminals are the same, only used to provide alternative connections when using the maximum output current of the supply.
All this talk of "separate supplies" and "star grounding" can be somewhat misleading and counterproductive. What is needed is an adequate analysis of analog signal paths and ensuring each ground path closely follows the signal and does not carry any other current.
If you open the image in a new tab you should be able to zoom around. It is a high pixel-count image, so everything should be readable.
Not a whole lot. This is a student project. Our electrical engineers abandoned us. There are filtering caps on the sensing circuits, and so far we haven't had issues with our digital circuits. I can certainly try to find a way to add filtering caps elsewhere. Where do you recommend placing them?
I should, once again, be more precise. The ground I was referring to is the one with the ground symbol in the image I posted and in your cropped version, where main power enters the power supply. Not the V+/- outputs. Is that not the "earth ground" which @PerryBebbington was referring to?
That is the ground wire of the power cord. It would be very troublesome if you started connecting things to it as it connects to your house wiring and may carry all sorts of odd currents.
As I said, the tendency is to refer to the negative supply of the circuits we are using, which is generally the return of most circuits, as "ground", but this is only a convenience. "Return" is more accurate.
I mean that what is important is to define each signal circuit which involves a signal line and the "ground" which is its return.
So the most important part is that they travel together and where practical, separate to other circuits especially power suppy as permitting them to take different paths is inviting some other circuit to introduce interference or spurious voltages.
And simplistically arranging "star grounding" may actually cause such a problem.
The ideal situation is to have analog and digital power supply and ground lines. The digital lines will service digital circuits and be laid out in such a way that branches are formed by individual circuits joining a single return line or pulling from a single supply line. The analog lines will be the same for analog circuitry. This provides a single easy path for current to return to the power supply. Before it does so, the analog and digital ground lines must meet at a single star ground point to tie the reference voltages together. This should be out of the way of the current return path so as to only tie the reference voltages together.
It's not really a full schematic, to be fair to you. It's more of a hookup guide. But the full schematic, when it is made, will show a mixture of both.
The thermocouple boards are AD8495 breakout boards. The relay is just a relay. The ADCs are ADS1115 breakout boards. The HX711 is a load cell breakout board.
The analog vs digital issue is the digital changes state very quickly. To do so the change generates frequencies well into the MHz. An analog circuit can react to these frequencies in a number of ways. So as you realize they should be kept as far apart as possible.
In addition there is what is referred to as common mode noise. This is noise the comes in on both the Positive and negative leads of the power supply OR on the thermocouples and pressure transducers.
We have many analog circuit devices: 8 analog pressure transducers and 6 analog thermocouples being processed by 4 ADCs.
Also I don't believe two power supplies are needed and may be detrimental. Looking at your schematic perhaps a 12V supply then from that a 5V supply.
I can suggest some approaches but I need more information.
What are the LPG_PT_Breakout boards?
I see there are 6 12V solenoids and one 24V. These are a source for voltage spikes and need to be carefully connected and their physical layout must be thought through.
Can you describe the housing construction? Not in great detail but: are they all in one box? Is the box metal? How far away are the thermocouples, pressure and flow sensor.
How "Bullitt proof" does this design need to be? aerospace, remote commercial, commercial, hobby.
What will your assembly wiring be: custom PCB, handwired with terminal strips, breadboard etc.
What is happening to create this noise? For TCs/PTs how do we know this noise isn't the actual signal?
The last missing piece is a Raspberry Pi which provides power to the Arduinos. I am relying on the Arduinos to provide 5V to the rest of the circuitry. My next question for @Paul_B is going to be how to connect that to the star ground... I need to spend the time to make a fully complete schematic, clearly.
The housing for all of these circuits is made using 3d-printed PLA parts mounted using bolts to pegboard. There is something of a nest of wires along the underside to get signals to where they need to be on the output screw terminal blocks. Could that be a problem?
The entire assembly will rest on rubber feet in a plastic box fitted with convective air cooling and reflective tape to keep temperature down and will be operated in desert conditions for the purpose of controlling a rocket engine.