Armature Feedback vs Tachometer

[Thedroid]

I'm working with a weighfeeder that uses a 3hp 180V 1750rpm DC motor controlled by a Bardac speed control. The Bardac always used tach feedback for regulation, and we have been instructed to remove the tach and switch the controller over to armature feedback. This weighfeeder regulated the amount of coal for the combustion process, and has to be very accurate. The weigh feeder uses a Merrick MC^2 to control the Bardac. Historically the tach's have proven very accurate, but are slowly being phased out as they fail on other weighfeeders.

Is armature feedback an acceptable form of regulation for a critical process like this? We performed three material tests today and hit 1t/hr on the money all three times. Tomorrow we will remove the tach and try again, but I would like some guidance on this.

 

[Thedroid]

Yesterday's tests with the tacho hooked up were all right on the money. Today's tests using just armature feedback were all off by 1-2%. Are tachos alot more accurate than armature feedback?

The weighfeeder uses an encoder mounted on the tail pulley of the converor to determine the the beltspeed and then uses a 4-20 to adjust the DC speed controller.

 

[DickDV]

Armature feedback usually results in around 4% speed error when the load goes from 5% to 95%. An analog tach takes this error down to about 1% over the same load range. A digital encoder takes it down further to near zero---one pulse per revolution non-cumulative.

Note that the speed error is related to load change. If your load doesn't change or only changes over a small range, all of the above speed errors are reduced in the same proportion.

A coal conveyor probably doesn't vary much in motor loading thus the armature feedback error is quite low.

 

[Compositepro]

A "weigh feeder" uses an actual weight signal as feedback and thus it should not require a tach(although you will get better short term accuracy). The weight signal is integrated to get a total weight over a certain time. If the speed is the only feedback signal then it is a volumetric feeder and tach feedback will improve consistency. Volumetric feeders will not compensate for changes in material density or flow characteristics, however.

It is very important to clearly define your sample size. It is far more difficult to hold +/-1% from second to second compared to hour to hour.

 

[Thedroid]

There are two seperate controllers involved. The weighfeeder, which has an encoder on the tail pulley, and also a load cell. The weigh feeder sends the speed reference required to the DC controller. Your right the weighfeeder calculates the speed based on the required amount of coal needed over time. We have 25 weighfeeder in the plant, and all of them were set up with tacho feedback. As the tachs wear out, we have been instructed to remove them and, and set them up for AVF. With the exception of the coal feeders, we calibrate all of them by placing weights on the loadcell and performing a chain test and then calculating the error. We have no way of knowing for sure how accurate this method is, but so far there have been no complaints. When calibrating the coal feeders we call for 1 ton/hr and run the feeder into a dumptruck, and weigh it. Then calculate the error and enter it into the controller. Sample size and density constantly varies, and as far as I know there has never been any reason to question the accuracy. Like I said in my previous post, performing several tests with the tach feedback gave us zero percent error three out of three tries. Today using AVF we we had a 1-2% error out of all 4 tests. I understand that the weighfeeder should be able to adjust for any speed variance caused by the drive, but I do suspect this error to be caused by the switch.

 

[Compositepro]

The load cell will measure the load on the weigh table but if you don't know the belt speed with accuracy then you cannot get an accurate integration of the weight with time. So your observation makes sense for a weigh-belt feeder. A loss-in weight feeder weighs the hopper and would not have this problem. So it seems your tach is important for accuracy.

 

[cswilson]

As DickDV suggests, the voltage drop across the armature is dependent both on the motor velocity and the electrical drops (resistive and perhaps inductive). It seems that you are (implicitly) assuming that it is due to motor velocity alone.

Do you have any way of:

1.) Measuring the armature current in a way that is useful for this purpose?

and

2.) Scaling this measurement into an voltage value equivalent to your armature voltage measurement, then subtracting this from the armature voltage measurement to isolate the velocity component?

Curt Wilson
Delta Tau Data Systems

 

[DickDV]

cswilson, when you place a modern digital DC drive in armature feedback, the drive does the calculations you mention. The result is a value for the internally generated voltage which, with a constant field is a measure of shaft speed. Unfortunately, it is only around 4% accurate so you will not get tight speed regulation under varying load conditions.

 

[roydm]

If you are using Cascade control the encoder on the tail pulley and load cell calculate Speed x Weight to give you mass/hour. Feeding this signal into the drive as a speed setpoint it should automaticly take care of any error due to pulley slip and motor feedback. Could it be that you are overshooting the tonnage setpoint and end up with more in the truck?
It's crucial to get your weigh scale aligned properly, any idlers that are high or low will cause large errors but correctly aligned you can expect +/- 0.5% or better with a chain calibration. Training idlers are the worst in my opinnion.
Regards
Roy

 

[Thedroid]

This is exactly how we are doing it. I'm not sure how we could be overshooting the setpoint though.

 

[roydm]

I assumed you were running all the coal to a truck for a set time then stopping the belt, does it stop straight away or run on a little?
In mining to verify a scale we would run the belt at a known speed, stop, then weigh the ore from several feet of belt on a lab scale. We had a couple of plates shaped to the belt profile fixed a set distance apart.
I can't think why your arangement should be any less accurate than tacho feedback. The weigh scale doesn't care.
Roy

 

[Thedroid]

That's exactly how we're doing it. The belt doesn't run on any it stops right away. We've decided to except the error and start up the belt next week. Your right though the scale should not care, but I did notice a difference.