[chrugel]

This will be my next move:

Replace the 1n4007 diodes in the gate network with HER105.

This beacase the HER105 is a super fast diode compared to the slow 1n4007, trr of 50nS compared to 2000nS

Replace the IRFP260 transistors with IRF640.

The gate capacitance of the 260 is 5200pF while that of the 640 is1300pF. The original UHU was designed for IRF540 which has a capacitance of 1700pF

Create a RLC snubber network on the output with these values:
R 25 Ohm
L 5uH
C 6nF

Since it seems to be nearly impossible to obtain this inductor at the time beeing this will be omitted and there will only be a RC snubber there for preliminary tests.
I dont think this should be any problem since the C is so small, of cause this will yeld higher dv/dt rates at the motor.
(frantically searching the web for diy inductor design formulas . . . )

Just to be clear, my goal is not a HP UHU controller just a high voltage version without the snubber heating problem and cross-conduction.
Ultimate goal: 150V 3 A

regards
chris

[kreutz]

This will be my next move:

Replace the 1n4007 diodes in the gate network with HER105.

This beacase the HER105 is a super fast diode compared to the slow 1n4007, trr of 50nS compared to 2000nS

Replace the IRFP260 transistors with IRF640.

The gate capacitance of the 260 is 5200pF while that of the 640 is1300pF. The original UHU was designed for IRF540 which has a capacitance of 1700pF

Create a RLC snubber network on the output with these values:
R 25 Ohm
L 5uH
C 6nF

Since it seems to be nearly impossible to obtain this inductor at the time beeing this will be omitted and there will only be a RC snubber there for preliminary tests.
I dont think this should be any problem since the C is so small, of cause this will yeld higher dv/dt rates at the motor.
(frantically searching the web for diy inductor design formulas . . . )

Just to be clear, my goal is not a HP UHU controller just a high voltage version without the snubber heating problem and cross-conduction.
Ultimate goal: 150V 3 A

regards
chris

I don't know about the PCB you are using, changing the output Mosfets by IRF640 is a wise decision, also replacing the bootstrap diodes by HER105. You might get a big improvement just by doing that. Add a couple of 100K 1/2 W carbon composition resistors from the high side mosfets' sources to power ground.

I recommend using the same output snubber used with the HP_UHU it was calculated for that voltage range. Current won't be a problem. Reduce the gate resistance to 51 ohms and get rid of the diode resistor combination at once. You can leave the 15 V zeners on the gates. Do another test. Resetting and UHU chip losing eeprom contents depends a lot on the ground noise and that is dependent on the PCB design.

If you were able to completely suppress cross-conduction, ground noise is not going to become a big deal, but all depends on the logic ground and power ground being connected only at one point near the power ground terminal.

Good Luck!

Snubbers act as "dissipative" high dv/dt energy suppressors, in a certain way they act as low pass (dissipative) filters. The calculations are completely different to "normal" low pass filters.

Kreutz.

[chrugel]

I don't know about the PCB you are using, changing the output Mosfets by IRF640 is a wise decision, also replacing the bootstrap diodes by HER105. You might get a big improvement just by doing that. Add a couple of 100K 1/2 W carbon composition resistors from the high side mosfets' sources to power ground.

I recommend using the same output snubber used with the HP_UHU it was calculated for that voltage range. Current won't be a problem. Reduce the gate resistance to 51 ohms and get rid of the diode resistor combination at once. You can leave the 15 V zeners on the gates. Do another test. Resetting and UHU chip losing eeprom contents depends a lot on the ground noise and that is dependent on the PCB design.

If you were able to completely suppress cross-conduction, ground noise is not going to become a big deal, but all depends on the logic ground and power ground being connected only at one point near the power ground terminal.

Good Luck!

Snubbers act as "dissipative" high dv/dt energy suppressors, in a certain way they act as low pass (dissipative) filters. The calculations are completely different to "normal" low pass filters.

Kreutz.

Thank you for the respons kreutz!
What is the purpose of the 100k resistors from the output to Vmotor?

Why do you recomend replacing the whole gatenetwork with one resistor?
Isn't the purpose of this network to create higher turn-on-times and faster turn-off-times?

The following part of the post will be about my understanding of snubbers and the original UHU snubber. It is going to be long and boring, so I completly understand if you don't continue reading beyond this point. ..

Today I've spent quite some time trying to understand the concept of "snubbers". Found some good pdf files online the explained them quite well.
Basically a snubber absorbs energy, either in the form of controlling voltage or current. Instead of looking at the frequency respons of a filter we are here looking at the impulse respons of the same type of circuts.
What i dont understand is the snubber-netvork on the original UHU design with the values 1k8 and 100n.
These values don't fit innto any of the formulas I've found so fare.
The rule of thumb i found for choosing the C value was 4 times the output capacitance of the Transistor, in this case about 6nF. The R value on the other hand is harder since this is best placed equal to the impedance of the resonant circut, in this case the output capacitance and stray inductance circut. This way the resonant circut will be critically damped. This is impossible to achive without trial and error in the specific controller - motor configuration and a digital storidge. But a rule of thumb for the mximum size here is to make Tau, the timeconstant for the RC snubber, nothing less than 10 times bigger than 1/f. And a rule of thumb on the lower limit is to not allow discharge currents thru the transistor at higher rates than the peak current capabilities of the transistors.

In the original UHU case with IRFP260 transistor this yealds C = 6nF and R between 2 an 800ohm probably in the 20 to 100ohm range.
The timeconstant in the original UHU is 4 times lower than the period time. For all practical purposes this will give a powerloss in the snubber-resistor that is half the supplyvoltage^2 / R. At 40V this is 0,4W. and 60V it is 2 W!
I have to get my hand on a storage oscilloscope so i can make som serious testing here!
Anyway thats all for now.

chris

[kreutz]

Thank you for the respons kreutz!
What is the purpose of the 100k resistors from the output to Vmotor?

Why do you recomend replacing the whole gatenetwork with one resistor?
Isn't the purpose of this network to create higher turn-on-times and faster turn-off-times?

The following part of the post will be about my understanding of snubbers and the original UHU snubber. It is going to be long and boring, so I completly understand if you don't continue reading beyond this point. ..

Today I've spent quite some time trying to understand the concept of "snubbers". Found some good pdf files online the explained them quite well.
Basically a snubber absorbs energy, either in the form of controlling voltage or current. Instead of looking at the frequency respons of a filter we are here looking at the impulse respons of the same type of circuts.
What i dont understand is the snubber-netvork on the original UHU design with the values 1k8 and 100n.
These values don't fit innto any of the formulas I've found so fare.
The rule of thumb i found for choosing the C value was 4 times the output capacitance of the Transistor, in this case about 6nF. The R value on the other hand is harder since this is best placed equal to the impedance of the resonant circut, in this case the output capacitance and stray inductance circut. This way the resonant circut will be critically damped. This is impossible to achive without trial and error in the specific controller - motor configuration and a digital storidge. But a rule of thumb for the mximum size here is to make Tau, the timeconstant for the RC snubber, nothing less than 10 times bigger than 1/f. And a rule of thumb on the lower limit is to not allow discharge currents thru the transistor at higher rates than the peak current capabilities of the transistors.

In the original UHU case with IRFP260 transistor this yealds C = 6nF and R between 2 an 800ohm probably in the 20 to 100ohm range.
The timeconstant in the original UHU is 4 times lower than the period time. For all practical purposes this will give a powerloss in the snubber-resistor that is half the supplyvoltage^2 / R. At 40V this is 0,4W. and 60V it is 2 W!
I have to get my hand on a storage oscilloscope so i can make som serious testing here!
Anyway thats all for now.

chris

The 100K resistors give a reference ground for the upper Mosfet capacitors to bootstrap circuit while the low side Mosfet is OFF( high impedance). It solves a sudden death Mosfet driver problem at power up as per International Rectifier tech notes.

On your realization that the original UHU snubber was arbitrarily calculated, that is my first impression when I saw it. The resistor value is too high for any useful work. Since I don't know the original designer, I will probably never know the reason.

About the original gate circuit; it is designed to give asymmetric turn on-off times, but the Mosfet driver already has asymmetric current handling, so it could be eliminated in favor of faster turn on-off times and lower switching losses because we control dv/dt at the motor's terminals (with the RLC snubber).

Best regards,

kreutz.

[chrugel]

Success, at least a partial one!
I finally got my new UHU controller chips today and got arround to test the thing. It worked without problems at 140V. When testing i realized that this gives way to much power on my Y axis so i decided to go with 80V instead. This gives me with ease 9000mm/min if i want to and that is moore than enough!
I got the strangest error her the other day, suddenly the controller started to just send gibborish back in the terminal window. What was strange about it was that if I send go commands the seems to be carried out correctly.
I'm still working on this one but i think i got to be the cabel or something other than the UHU chip since everything seems to be working as it should.

I have set the "m" parameter in the controller so that the resolution of my system is 0,0254mm pr step, after my initial tests I'm wondering if this might be to big since i get very jerky movment at low speeds.
what do you think?

regards
Chris