PWM Back Emf

This is slightly off topic, but it is motor control...

I am building a small dc motor controller to drive toy brushed dc
motors. This is all low voltage (<5v) and low current (150mA max).
I want to measure the back emf between PWM pulses to provide a
fairly constant motor speed. To measure the back emf, I have two
resistors acting as a divider (10k each) between the motor
terminals. The PWM only pulses the High side FETs, so when they
turn off, the motor terminal runs through the divider and to
ground. The divider gives me 1/2 the back emf.

This picture shows the voltage divider trace -
http://home.people.net.au/~frankt/pwm_bemf.png

After the PWM pulse is switched off, the voltage divider shows a
negative voltage, about -0.4V for about half the off pulse. After
that, it pops back up and shows the back emf which I can measure.
This unfortunately is causing me grief, because I have to wait until
just before the next pulse to measure the back emf. What's even
worse, is that period of negative voltage is fixed - if I double the
PWM frequency I will not get any back emf. The sample shown in the
capture is at a noise 4kHz - I'd like to get well above that.

So, the questions, what is causing that small negative dip? Is it
normal? Is there a way I can remove it? Or, with an increased
frequency, do I have to stop PWMing until the back emf signal
stabilises to read it?

Thanks,
Frank