On that scale the error is difficult to see (as Steve pointed out while I was typing this).  Yes, there is a little up-tick at the left side of the graph.

WARNING: the plots I suggest below PRESUME IDEAL RESPONSE OF THE VFD TO INPUT VOLTAGE!!!
That may or may not be the case.  They also presume that your volt meter is correctly averaging the PWM signal to give you the DC voltage reading.  As Steve politely corrected me above, that is not always the case.  Some of the error may be due to the volt meter, specially at the lower RPM settings (and this lower PWM duty cycle).

That said, here are some hints for other ways to look at the data:

(1) calculate voltage error for both the 10V and 5V measurements:
Voltage error = measured - ideal

(2) Then calculate and graph voltage error as a percentage of ideal voltage:
percent voltage error = ((voltage error) / (ideal voltage)) * 100

(3) Calculate what the actual RPM should be for the measured voltages (again PRESUMING IDEAL RESPONSE IN THE VFD):
calculated actual RPM = ((measured voltage) / (10)) * 3600
or divided by "5" for the 5V scale

(4) Calculate and graph the RPM error:
rpm error = (calculated actual RPM) - (commanded RPM)

As a curiosity, how does the VFD reported RPM match up with the calculated RPM in step (3) above?

Bob

Rodney,
Thanks for pointing me to this thread.  Here is a curve showing the linearity we get out of the C32:
 
As you can see the curve is a little higher than it is supposed to on the lower end (usually when accuracy counts more).  I think this circuit may be improved. I will take a second look at it.
In your case, you think you may be just fine by moving the curve just a little bit down.  You can do that by lowering the max speed or y playing with the on-board potentiometer.
Other solutions could be:
-   Closed Loop Spindle:  This used to work in Mach3, but the latest I know it is not working on the current versions of Mach3.  It is not currently supported on the PoKeys or ESS, but I am sure it will get implemented soon.  You can also implement it on the pokeys using brain, macro or LUA script by writing a script that would do the PID to correct the PWM to reach the commanded spindle speed.
-   Another solution would be to implement spindle speed control via Modbus.
Arturo Duncan
http://cnc4pc.com

Ok, i have a new problem. I got the Pokeys Motion controller and Pokeys motherboard from Arturo, Now when i set it up i can't jog any axis or start the spindle without getting error message "Mach4 motion steps overflow (-*********x) - check motor speed." Can anyone who has mach4 tell me what this means?

I figured out why i'm getting the step overflow error. It has to do with the resolution of steps on my vistacnc PA1-S pendant. It appears i can add this to the list of things that works spotty in Mach4. The vendor said that everything should work but i can't access the configuration screen in mach4 as that is not programmed in this version of the plugin. It seems to me that these hardware and software manufacturers need to stop selling stuff that works half assed and spend more time debugging before coming out with a product that only works partially, but hey we will have an update maybe for you someday. Sorry for the rant, just a little frustratedc with how my luck with my current hardware and software vendors has been going. I wish i knew then what i know now. I would have went with someone who cares about customer support like the guys at PMDX. These guys don't have a horse in this race but have been more than willing to help me get this sorted out as much as they can. For that fellas you have earned my future business.