Multistage up-counter loses count

Device: 6ED1052-1MD00-0BA8

Using LOGO!Soft Comfort

I am using (among other circuits) two up counters connected to I3 with a 200 Hz 0-12Vpp 50% duty cycle rectangular signal. One counter called "A" can count up to 99999 w/o reset, I compared "A" counts with an external event counter, the counts are identical. The second counter "B" resets using on=100 and off=110 ro trigger a third counter using an M flag from output to the Reset input. After approx. 700000 counts counter  "B"  lost 7 counts compared to counter A. It gets worse when increasing the input frequency. Is there any way to remedy this miscounting? Thanks in advance!

Stephan

Hi, thanks for answering my call. The counter "A" is the B001 displayed at the bottom of the display. The counter "B" is B009 displayed vertically with B041 and B047. What I am trying to realize here is a 3 stage counter that turns on a relay at a given setting.

Yes, I left out all the circuitry which is not relevant to the counter parts, different displays and monitoring inputs, some modbus stuff. What should the on/off values be to make the counter go a full 1 million i,e, reset at 1E6 to 0 since the counter hangs at 999999? I loaded and tested  the circuit I sent into the device and it shows the same behavior as in the full circuit. The counting works fine till around 27000, at 30000 the "B" counter is 3 counts behind "A", the input frequency is 1 kHz, and it gets worse... Some counts get obviously lost, not by design but probably internal timing errors. Would you by any chance have a better design for a timing safe multi-stage counter?

Btw, at 2 kHz input freq it's a complete mess, it loses 100;'s of counts before reaching 30000... so there is definitely a problem when resetting a fast counter in the middle of counting...

I added some circuitry (from the manual) to measure cycle time, it's 0,40 ms for this circuit, lower than the full circuit at around 0,57 ms. So somewhere around 1 kHz (1ms period)  and higher counting loss occurs, obviously we are getting close to the cycle time...

The other unfortunate conclusion is that cycle time is not constant, so we may even have random errors at low input frequencies. 

Oh, at 100 Hz input frequency I get no error up to 70000 counts.