Hypertrend has suddenly starting showing decimals.

[WaterWorks]

I have Hypertrend looking at several  modbus addresses. (eg.  40048,40050,40054)

The trend has suddenly started to wander on only one of the addresses, as if incoming data is straying.

Hypertrend has always shown the numbers as whole numbers (ie 85 %).  But since it has started to wandering (down 15% of full Scale) It will show decimals in the strange data(ie 65.567 %), all data before and after this is whole numbers.

In the data member configuration I am doing no scaling and the filter is set at 1, and the log to historical database is set at 1.

I thought when the filters were set to one that no decimals could get through.

The incoming data appears solid but there is a chance that it has started to wander. 

I need to rule out the Hypertrend numbers first.

[WaterWorks]

Anyone with ideas on this?

[Ryan.S]

"resolution 1" doesn't mean to round it to integer. It only guarantees that the whole number is exactly right, but the decimal part is not accurate.

For example, the value is 1.2345. If you set resolution to 1 and log it, when you retrieve it from database, you might get 1.xxx, which does have the inaccurate decimals.  "xxx" can be any value.

The filter set to 1 doesn't mean only whole number can pass it. It means if the new value has the deviation of 1 from the old value, it pass the filter. For example, the last value was 1.2, only when the new value is bigger than 2.2 or smaller than 0.2, it will pass it.

So, your configuration does not do exactly what you want.

I think you need to use round().

[Bob_L]

WaterWorks,

If your source registers are integers, you can only read and therefore log whole numbers. A new value is logged only when the device is polled and then only if the value changes by at least the deviation specified. (1 in your case)

The hypertrend object plots those values and draws a straight line between values. The rate of change is assumed to be linear. When the cursor on a hypertrend is positioned between 2 logged points, the displayed value is interpolated from the adjacent 2 known points. That interpolation result may very well contain a fractional component even when the known points are integers.

If you logged a value of 85 at 1200hrs and 86 at 1300hrs, the implied value at 1230hrs would be 85.5. A Hypertrend would display this value even though it was never logged.

Regards,

Bob

[WaterWorks]

Thanks Ryan and Bob,

It makes sense that I am seeing weird decimals when I think of the data being linear between two points.  Normally I don't see such strange data but because I am getting such a wide swing in my data in one poll to the next I will see this.

I think I was sleeping on the filter idea. I knew that but was not thinking that thru.

Now to find out where my strange data is coming from.

Thanks for the help.

WaterWorks