Dear Forum Members,

I am into designing Precision Resistance welding controllers. For past few days I have been struggling with a very strange problem the problem is predicting thrysitor firing angles in order to get the required target welding current from welding transformer. Lets have a brief introduction before jumping to main problem.

- A thyristor is used to feed the primary of resistance welding transformer.
- Thyristor is fired at diffrent diffrent firing angles for getting target/ required welding current.
- A rogowski coil is used for measuring the attained welding current wich acts as a feedback device to controller.
- The controller keeps adjusting the firing angle depending upon the feedback from rogowski coil in order to keep the welding current constant.
- since transformer has got magnetising current and is itself an inductive load thus the thyristor is never fired at zero cross but always fired with a minumum 2500uS delay past zero cross at this firing position maximum current is obtained and as the firing delay increases the current obtained goes on reducing. The maximum firing delay is 7500uS giving minimum welding current.
- For adjusting the firing angle a table in code memory is used. The table is basically a table of theoretical RMS values corresponding various firing angle positions i.e. firing delays for 2500uS to 7500uS. There are 100 values in this table each each representing firing delay wich are 50uS apart. Since I didnt want to deal with floating point I have scaled the table to fixed point 8 bit values. The value corresponding to 2500uS delay is scaled to 255 and likewise other values too have been scaled. I call this table as RMS factor table.
- The next firing delay is calculated as follows...
new RMS factor = Previous cycle RMS factor * target current / measured current in previous cycle.
than from table the next firing delay corresponding to new rms factor is determined.

http://www.8052.com/users/prahlad/rmstable.txt

Now the problem is the values calculated using the RMS factor table do not match the practically measured values. Here is an excell table representing the theoretically calculated and actual measured current values at various firing positions.

http://www.8052.com/users/prahlad/thexp.xls

The two values [measured and calculated] are very close at the middle part of the table [shown blue] but at the top part the calculated values are higher and at the bottom real measured values are higher. Does any body have experience of dealing with such a situation I want to do something that can direcly predict very correct firing angle and no diffrence in predicted and measured current value. It will be a great help if some body suggest me some way by using wich the diffrence in between measured and predicted values can be brought down. If the above information is not sufficient to visualize the problem/situation please do let me know I will furnish required information.

Thanks & Regards,
Prahlad Purohit