Analogue and digital noise

[Timmy Hat]

Hello all,

I am using a 6110 analogue board and a DIO32HS board which are linked together internally to do simultanious aquisition. The application is in-cylinder pressure measurement from a engine. The engine has a pressure transducer which is linked to a charge amplifier (class 2 device, not earthed) which is then fed into the 6110 board. In order to keep track of where the engine is in the cycle the drive shaft of the engine is connected to a shaft encoder which give out pulses every 0.2dec and a pulse for top dead centre and bottom dead centre.

My problem is as follows... The environment is very noisy. There is an inverter nearby which drives the dynomometer which keeps the engine at a constant rpm. There is also noise from the spark coils and discharge. This gives a noisy analogue signal, something which I think I could probably deal with by adding a low-pass filter in the analogue signal path. The real problem is the digital signals. Every so oftern, seemingly at random there is a spike in one of the digital signals. (i.e. a high value when there should not be one). What is the best way of dealing with this. Is it possible to use a low-pass filter on the digital lines as well or will this mess things up by attempting to filter a signal composed of high and low values. How should digital signals be earthed. At the moment the earth from the shaft encoder is connected to the earth on the digital board. All cables are shielded with BNC connectors.

Many thanks, Tim

[Buechsenschuetz]

It is not very easy to remove spikes from a digital line.

First you should carefully check your earthing/grounding scheme. If noise/spikes are coupled to a ground line you will never get happy with filtering. You should also consider using isolated inputs for the digital signals, in many cases this method avoids lots of trouble.

I can't judge your grounding setup. Usually you should tie the reference gnd of analog signal to the analog gnd or reference input.

To remove spikes from digital signals, you can use 'digital filtering' by software (if possible). A quick-and-dirty method would be to add a lowpass filter in the digital line but you will need a Schmitt trigger circuit to recover signal slopes. A more advanced hardware method would be to use a square wave generator with rather high frequency and a counter triggered by the signal line. Signals will only be allowed to pass if they last a certain number of generator cycles, i.e. let the counter count up to a certain state. If the pulse is too short the counter will not reach the defined state and the signal will not pass. Any of these methods, however, will cause a certain delay of the digital signal.

[Henrik_Volkers]

Some ideas for your noise:

Avoid Grounding loops ... (Has to be sad 🙂 )   Is it possible to use optocouplers ?  

Try to put the maximum load resistant of your (digtal) sensor output (see spec) near your (DIO) card.

Think of every signal path (Signal + and -) as an induktor loop and magnetic fields are very hard to shield. You mentioned BNC connectors, what type of cable(s) are you using? Tightly twisted pairs? what type of shielding? 

How do you power your amplifier?

[Timmy Hat]

Hi all. Buechsenschuetz and Henrik Volkers, thanks for your suggestions. I've solved the majority of the problem by coupling the digital inputs with the DAQ card with small (0.1uF) capacitors accross the inputs. This seems to kill the majority of the noise without distorting the signal too much. There is still the odd spike. This seems to be because the output of the shaft encoder only drops to about 0.7volts when low. Therefor it only takes a small spike to push it over the 0.8volt threshold and give a high reading. The cables are shielded with a single core and terminated with BNC connectors (Henrik). What is the definition of a ground loop. I have many digital signal sources each with their own ground which all feed into the breakout box for the DAQ card. Once inside the breakout box all the signals are grounded. Will this cause a ground loop?

Once again, thanks for all the suggestions. Tim

[Buechsenschuetz]

It is difficult to analyze ground loops from the distance, there are too many details involved. There is a good article about grounding schemes on the NI website. It mainly deals with analog signals, but it is no fault to treat digital signals with the same care as analog ones.

One problem on pc based DAQ is that on virtually all pcs the GND line of the main power supply is connected to the case or protection earth - a ground loop cannot be avoided if you connect other devices with the same configuration to the pc, and it will be even worse if the devices are connected to distant mains power outlets.

I think one of your problems is the high voltage of the low level. 0.7v gives a very low safety margin. You might consider using external Schmitt trigger gates, like the 74LS14, which will ignore spikes slightly above the low level voltage until they actually reach the high level input voltage again. NI usually does not define hysteresis on digital inputs so there is some danger that the inputs behave irregular in the 'dead band zone' between 0.8 and 2.3v. You might also design your own comparator circuitry with lots of hysteresis.

My colleagues and I also uses capacitors across the inputs of noisy digital lines. I think it is better to add a series resistor with around 100 ohms inserted before the capacitor which turns them into a 'real' low pass filter. The capacitor should be as small as possible for the sake of signal slopes. I always use Schmitt triggers on digital inputs, or at least bus receivers with a hysteresis, like the 74LS245 (not all versions of the 74xx245 have hysteresis, I think the 74HC(T)245 doesn't).

[Timmy Hat]

Hi, I think I have found the source of my problems... I have now been informed that the ouputs from the encoder are not proper TTL but "Darlington line driver" instead. This is a historical issue from when all the outputs were fed down long wires to a mainframe which had an ADC on it. My next question is: Does anybody know a quick and easy way of converting between the two? Are there any companies out there which make boxes which do this?

Once again, many thanks, Tim

[Henrik_Volkers]

Industrial level shifter : phoenix contact  or weidmueller or schneider just to name some

However, the output of your encoder couldn't be that wrong. It worked with long cables in the past. Just look at the old ADC input circuitry and you might get an idea. If you post a schematic of the output or a link to the datasheet, maybe it's just an additional pullup resistor.

The connector box of your DAQ might provide enough power to drive a little converter (One of those little black bugs, which in the past need a foot stamp to be flat, but thanks to Darwin they are already flat nowadays      

[Buechsenschuetz]

I don't think a pull up resistor will solve the problem with the rather high low state voltage, it will rather increase this level. Pull-up resistors may be a solution when you have outputs with rather high leakage currents or a too low high state level.

As mentioned, 0.7v low state voltage gives a pretty bad safety margin. It is not too difficult to brew your own level shifter, however you might find an off-the-shelf solution as indicated by the previous poster. Don't know whether Phoenix or Weidmueller components are easily available outside Europe but there might be equivalents from domestic suppliers.