Protection circuitry

[breakshift]

First off, let me explain that this question is related to the same project as this thread:

https://www.electro-tech-online.com/threads/petrol-generator-question.116051/

Basically, we're going to be measuring the output of the generator (6kW), voltage and current, and feeding this into a sbRIO for the purpose of a control loop. Clearly, the generator output will be scaled down to a suitable range before it's fed into the sbRIO. Although I'm not entirely sure which sbRIO model we're using (I can easily find out if required), here is a link to something similar:

NI sbRIO-9612/9612XT - Embedded Devices with Analog Input and DIO, 2M Gate FPGA - National Instruments

I'd like to design some circuitry to protect the sbRIO from the possibility of some failures causing the generators full output to become present at the input of the sbRIO. I think the following crowbar circuit would be appropriate:

Axotron

This would protect the sbRIO from overvoltage, and also overcurrent. The fact that the circuit permanently breaks once triggered is not an issue - in fact this could be desirable as it would make it very obvious that something has gone wrong in the first place (possibly a failure in circuitry that steps down the voltage...).

What do you think?

Is the generator likely to produce voltage surges/spikes? If so, why?
Also, would the Crowbar circuit respond quick enough to overvoltage?

I'd appreciate any opinion on the subject!

Thanks.

 

[Jaguarjoe]

You can buy analog opto isolators. That's all the protection the I/O module would really need.

 

[mbarazeen]

for over current/ short circuit MCCB cab give you enough protection. for under / over voltage & phase failure there are relays available for industrial application, you can directly installed it to the lines.

 

[breakshift]

You can buy analog opto isolators. That's all the protection the I/O module would really need.

Ok so the I/O voltage input range is 10V. I have circuitry in place to step down the generator output to this range. Then I place an optocoupler in between this and the I/O module. So I'd need the optocoupler to have a current transformation ratio of 1:1, and place a suitable resistor at its output terminals to establish the voltage again.

EDIT: Had a look on RS Components and all their optocouplers have a maximum input voltage of below 1.6V - I need one with 10V. Are they made with this kind of input range?

Edit: How about this? https://www.electro-tech-online.com/custompdfs/2011/02/65236.pdf

Is this device analogue input and output? And I assume it's a 1:1 transformation ratio?

 

[Jaguarjoe]

The Agilent part is digital. It's "4 and 20" ma, not "4 to 20" ma.

If you have a 10v signal trying to feed into a 1.6v input you would make a voltage divider. Vin/(R1 + R2) = Vout/R2. Pick R2 to be 1/10 (or less) of the input resistance of the 1.6v device. Solve for R1.

The RIO box datasheet says that AI range can be 10, 5, 2, or 0.2 volts. This should be easy to work with.

I found this at Digikey:

Digi-Key - 516-1691-5-ND (Manufacturer - HCPL-788J-000E)

It has a 200mv input so a voltage divider will let work with 10 v signals. The output range is scaled by Vref, a voltage you provide. $9 is pretty cheap.

Note that there are 2 power supplies for this part, one for the isolated input side and one for the output side.

What did you do for current measurement?

 

[breakshift]

Looks like that part will do the trick. However I'm in the UK, so I found it on the RS Components website, and also found a couple more parts which I think might do the trick:

https://www.electro-tech-online.com/custompdfs/2011/02/0900766b807796e0.pdf
https://www.electro-tech-online.com/custompdfs/2011/02/0900766b807796df.pdf

I have a question about resolution and accuracy here... if I'm stepping down 230V to, say, 200mV for one of these devices, surely I'll be losing some accuracy of measurement? A 1% variation in the 230V will cause a variation of 200mV*0.01=2mV. This is small, so wouldn't noise in the signal cause errors in the voltage measurement? I appreciate that a variation of 1% isn't very important, but you see what I mean.

So in this case, maybe the first of the two parts I suggested would be better, as it's designed for noisy environments?

Edit: Sorry, for current measurements we have a Hall-effect sensor, so that's isolated already.

 

[Jaguarjoe]

The voltage divider can be implemented with 0.005% resistors if need be. They're about 15USD each. Be sure to keep power dissipation in mind too.
THe 16 bit ADC will be a considerable challenge. If you do very well you'll end up with 14 bits, not so very well could land you at 12 bits. Getting to 16 is like Monty Python searching for the holy grail. Some tricks include installing the ADC all by itself in its own separate enclosure. Isolate the I/O with opto isolators everywhere. Never let any noisy digital lines near the ADC. Power supplies should be clean and stable. Last but not least, deep pockets always help

 

[breakshift]

THe 16 bit ADC will be a considerable challenge. If you do very well you'll end up with 14 bits, not so very well could land you at 12 bits. Getting to 16 is like Monty Python searching for the holy grail. Some tricks include installing the ADC all by itself in its own separate enclosure. Isolate the I/O with opto isolators everywhere. Never let any noisy digital lines near the ADC.

I'm not following you... what do you mean by if I do well I'll end up with 14 bits? Are you talking about the I/O module or the optical isolator?

 

[Jaguarjoe]

16 bits is 65,535 step resolution. That's better than 0.002% which is practically impossible to obtain. Say that the ADC is set for 10 volts full scale. If you applied 5 volts to it and if everything was perfect, you'd get exactly 32768 counts. In reality it'll be constantly dithering at 32768+/- 4 or maybe worse. I was surprised to see that RIO device with everything together on one board.
The opto's are strictly analog devices so resolution isn't a concern.

 

[breakshift]

Yeh the RIO is a very nice piece of kit. I think we were talking about different things - you meant the resolution of the ADC, but what I meant was the fact that, purely in the analogue domain, if I'm stepping down a 230V signal to 200mV, noise is likely to influence the accuracy of my measurements. I don't know what typical +/- mV noise tends to have on signals, but because 200mV is so small, the noise on that signal will be scaled along with the signal to whatever the output voltage of the isolator is set to.

 

[Jaguarjoe]

The 230:0.2 divider will be resistors. Using good metal film or wirewound pieces will generate ~1uv noise. OTOH, the HP 7800 isolator generates 300uv of noise mainly due to the chopper stabilized amplifier and the sigma-delta ADC inside of it. Most of that can probably be filtered out though. If this is a DC generator, the noise from the brushes and commutator will be quite significant, an AC machine will be quieter but it will require a rectifier to get to DC to suit the ADC and maybe the HP part too. At normal levels
(+/- 10v) precision rectifiers are easy to build. I'm not aware of any that can be store bought but I've never been in the market for one either.

GEN-->230:10 divider-->precision rectifier-->10:0.2 divider-->HP 7800-->16 bit ADC-->pub

Do you have an oscilloscope that can see submillivolt levels?

 

[breakshift]

The generator is AC - it's output is household mains voltage - 230V at 50Hz. What's the sampling rate of this ADC? If it's above 100Hz then I believe we don't need the rectifier, as it can handle the incoming 50Hz AC, am I right? I was under the impression that the HCPL-7800 was AC-in and AC-out?

Having said that, it shouldn't matter very much if it's AC or DC - at the end of the day I just need an accurate reading of the peak voltage coming from the generator, not the whole waveform.

Yes I do have access to such an oscilloscope - it's a Uni project

 

[Jaguarjoe]

You're in good shape. The HPCL part has a bandwidth of 85kHZ, the last page of its data sheet shows its response to 20kHZ sines and square waves. 50HZ is a walk in the park.
The RIO board has 4us response. Good there too.

 

[breakshift]

Question: The Vref pin on the part that you suggested, is that just for limiting the output voltage to something below Vdd if you so require? It doesn't allow the output voltage to be raised above Vdd?

 

[Jaguarjoe]

According to the data sheet, 4<Vref<Vdd+0.5 volts maximum. 4<Vref<Vdd volts recomended. Vdd is 5.5 volts max. Very, very few circuits tolerate inputs above Vdd or below Vss. Some don't even like getting close.

 

[breakshift]

Yes I see that - but what's Vref for then?

 

[Jaguarjoe]

Vref sets the upper value of Vo, the output. It is determined by a gain formula for the output circuit: Vo = Vref/0.504
Vref also sets the upper value of ABSVAL, the absolute value of the input. The input is bidirectional so it can measure positive and negative values. This is of value when measuring current that can flow into or out of a circuit, like a car battery provides current to start the car then accepts current from the alternator as it recharges.

 

[breakshift]

I see. Could prove very useful in certain circumstances.

Well, thanks very much for the help, been very helpful!

Cheers,

Mike.