Maker Pro
Maker Pro

H bridge unusual behavior

C

Chuck Simmons

Jan 1, 1970
0
I'm using a combination motor driver from Mitsubishi. It has all H
drivers in it. One three phase brushless DC motor driver (12 volt PWM,
transconductance), One two phase stepper motor driver intended for
microstepping (12 volt PWM, transconductance), brush type motor driver
(gain of 8, no PWM, 5 volt) and two VCM drivers (gain of 12, no PWM, 5
volt). This is a highly popular motor driver configuration in a single
42 pin surface mount package.

In using the VCM drivers, I have opamps and a current shunt to allow
current feedback or I can set voltage gain and let the VCM coil
resistance set the transconductance. The two modes are configurable by
either shorting a capacitor and deleting one resistor or the other way.

In the computed open loop response of the VCM servos using an SRS785 (I
think it is) analyzer, I see a gain and phase glitch at about 1KHz. By
opening the loop, the response of the Mitsubishi VCM drivers can be
measured directly. The VCM drivers show a gain/phase glitch at about
1KHz. The VCM specifications indicate no dynamics other than rigid body
from 50Hz to at least 50KHz. Nevertheless, I tested the VCM drivers into
resistors. The gain/phase glitch moved up to about 1.5KHz. The
gain/phase glitch is seen in both voltage gain mode and current feedback
mode for both VCM channels

The mystery is that in VCM drivers I have designed for the same purpose,
I have not had the gain/phase glitch. I have only seen it with the
Mitsubishi part. The frequency is too low to be an internal pole/zero
especially considering the part is intended for 5KHz servos and better.
Anybody have an idea where gain/phase glitch comes from?

Chuck
 
R

Robert Baer

Jan 1, 1970
0
Chuck said:
I'm using a combination motor driver from Mitsubishi. It has all H
drivers in it. One three phase brushless DC motor driver (12 volt PWM,
transconductance), One two phase stepper motor driver intended for
microstepping (12 volt PWM, transconductance), brush type motor driver
(gain of 8, no PWM, 5 volt) and two VCM drivers (gain of 12, no PWM, 5
volt). This is a highly popular motor driver configuration in a single
42 pin surface mount package.

In using the VCM drivers, I have opamps and a current shunt to allow
current feedback or I can set voltage gain and let the VCM coil
resistance set the transconductance. The two modes are configurable by
either shorting a capacitor and deleting one resistor or the other way.

In the computed open loop response of the VCM servos using an SRS785 (I
think it is) analyzer, I see a gain and phase glitch at about 1KHz. By
opening the loop, the response of the Mitsubishi VCM drivers can be
measured directly. The VCM drivers show a gain/phase glitch at about
1KHz. The VCM specifications indicate no dynamics other than rigid body
from 50Hz to at least 50KHz. Nevertheless, I tested the VCM drivers into
resistors. The gain/phase glitch moved up to about 1.5KHz. The
gain/phase glitch is seen in both voltage gain mode and current feedback
mode for both VCM channels

The mystery is that in VCM drivers I have designed for the same purpose,
I have not had the gain/phase glitch. I have only seen it with the
Mitsubishi part. The frequency is too low to be an internal pole/zero
especially considering the part is intended for 5KHz servos and better.
Anybody have an idea where gain/phase glitch comes from?

Chuck
--
... The times have been,
That, when the brains were out,
the man would die. ... Macbeth
Chuck Simmons [email protected]

I would say that if an open loop test with resistive loads show this
"glitch" (as you seem to say), then believe what you see and correctly
infer the problem is in the system.
Remove external capacitors and re-test; if you still see it then some
part of that part has an internal pole/zero.
If you can look at the signal "inside" the unit at various stages,
then you can partly isolate the source.
If the problem shows only with feedback, try to find a way to open the
loop at AC at minimum for testing.
Try the test at different loads; maybe it happens at heavy loads but
not light or no load.
If it happens when loaded and the frequency shifts some as loading
gets heavier, the problem may be in the driver IC using PNPs in a
process optimized for NPNs (integrated PNPs tend to have a phase shift
under load).
 
C

Chuck Simmons

Jan 1, 1970
0
Robert said:
I would say that if an open loop test with resistive loads show this
"glitch" (as you seem to say), then believe what you see and correctly
infer the problem is in the system.
Remove external capacitors and re-test; if you still see it then some
part of that part has an internal pole/zero.

There are no external capacitors for the Mitsubishi driver except on the
snubbers. These are invisible in the response because the load
resistance is about 1/10 of the snubber resistors (using recomended
values).
If you can look at the signal "inside" the unit at various stages,
then you can partly isolate the source.
If the problem shows only with feedback, try to find a way to open the
loop at AC at minimum for testing.
Try the test at different loads; maybe it happens at heavy loads but
not light or no load.

The Mitsubishi is tested independently by driving at my driving point,
channel A to the Mitsubishi input terminal and channel B to the current
measuring opamp output. No capacitors in that circuit and the opamp is
an LT1215. The same current measuring circuit I've used in previous
versions. I can't test no load because there must be current for the
circuit to make sense and only current is interesting. Voltage gain is
incidental to the function. The design load for the part is gives about
1 amp peak on both VCM channels. Closed loop typical currents run from 0
to 100ma peak.
If it happens when loaded and the frequency shifts some as loading
gets heavier, the problem may be in the driver IC using PNPs in a
process optimized for NPNs (integrated PNPs tend to have a phase shift
under load).

20 degrees at 1KHz? That's what it is. It is about 200Hz wide.

Chuck
 
Top