4-20ma, hz, dwell, inverse slope.

[bernie_carlton]

ghriver - to visualize the relationship fire up EXCEL and do the following. In the first column (labled DWELL) enter 1 through 100 (use Edit - Fill Series). Make a second column (labled SPEED). In the second column's first cell enter "= 100/A2". Drag B2 down to fill the whole column.

Nopw make a chart of these two columns. The DWELL line is a linear change. The SPEED line is the result of dividing a constant by a linearlly changing value.

 

[kamenges]

I think your difficulty is that you are falling prey to the very human desire to accomplish the problem in the fewest possible steps. You believe there is a direct conversion from time to command to be had somewhere. There is. the probelm is that this conversion will completely obscure what is happening in the process. You are much better served breaking the mapping/scaling into all of it's components and figuring out each one individually.

First you have a scaling that converts desired dwell time to motor speed. Second you have a scaling that converts desired motor speed to a current value. Third you have a scaling that converts a desired current value to an analog output command. Of these three scaling operations, only the first is an inverse proportionality. As I said above DO NOT TRY TO DO THIS IN ONE STEP. You will simply confuse yourself in the process because you obscure the obvious relationships between the elements.

So lets use the values I came up with before to see if the 45 second value you listed makes sense. According to the equation I posted, a 45 second dwell should require a motor speed of 23.78Hz. Also based on the current command equation I posted, the current command would be 5.512mA. Two posts ago you listed the information I need to convert current to analog output command: 0 = 4ma; 30840 = 20mA.
This results in the current to analog command scaling of :
= (1927.5 bits/mA * mA command) - 7710 bits offset.

Using this equation I figure a 45 second dwell should take a bit command of 2914 bits. This is off by about 10%, probably the result of the initial constant I came up with and cumulative measurement errors.

Bob O suggested using Excel to give you a trendline value. You can do this but you will need many more points to give you a good fit. I would suggest at least 10 and would lean toward 20. If you do an X-Y scatter plot you don't need to worry about consistent sample spacing. You will also need to use the 'Power' equation type. You will find that the equation will be something like:

y = 1000x^-1, with the 1000 just a number I picked out of the air.

x^-1 is the same as 1/x, which makes it an inversely proportional relationship.

Keith

 

[ghriver]

I got a handle on conversion steps 2 and 3 it was the 1st step of converting dwell time to motor speed that is troubling me.

 

[bernie_carlton]

You have a system that is transporting something a certain distance. The time it takes to go that distance is dependent on the speed of the transport.

D = R * T (Distance equals Rate times Time)

REWORKED TO:

R = D/T (The needed speed will be the distance (some constant) divided by the time)

So you use DIVISION. That's the whole basic concept for step 1.

 

[russrmartin]

My fault

I feel obligated to jump in here, as I think that my previous error and mentioning y=mx+b has confused the original poster. The best thing to do here is ignore y=mx+b completely at first.

That said, think of this exactly how these two have set it up. Your formula is V(apparatus velocity) = D(distance traveled which is constant) / t(desired dwell time).

Once you have determined V(apparatus velocity), you can then complete your scaling as Kieth has described. It sounds to me like Kieth is right on, and you're trying to come up with one equation in one shot with which you can calculate the scaled value.

Once you have your linear function to scale your 0-32767 to 4-20mA, or whatever you want your function to produce, you can plug V in as the X value in your linear function. Thus, I think your final equation will look something like this:

y(scaled value)= m[slope of scaling function] * X[which is comprised of the formula for V] + b[linear scaling offset].

You other guys may want to check my work on this, as apparently my brain is convoluted and I messed this up to begin with.

Russ