I need to clean up the output. What do you suggest?

Thread Starter

gglone56

Joined May 10, 2019
11
1578068402576.png 1578068455606.png
1578069164435.png
Additional details: The circuit, as shown above, is just one of 3 identical stages in parallel. As shown above the output (pink trace) is measured across R which is actually 2-5 Ohm resistors in series. The problem I see with the output are the over shoots. With a 12V DC supply IN, I'm getting spikes to 25V and down to -9V. The input signal has no over shoots until it is plugged into the circuit. Any help would be appreciated.
 

Alec_t

Joined Sep 17, 2013
14,280
Welcome to AAC!
The spikes look like they're due to stray inductance. Not surprising with all those lengthy straggly wires. Try a tidier, compact layout and see if that improves things.

Edit: Crutschow types faster than me :)
 

pmd34

Joined Feb 22, 2014
527
You need to minimize the potential inductance of all leads: So the gate driver to FET gate must be very short. You should have a big capacitor for the 12V supply right next to the FET as well as a smaller decoupling capacitor to help short any high frequency spikes.
 

DickCappels

Joined Aug 21, 2008
10,153
There seems to be a common theme here. You need to use high frequency wiring techniques. My question is, why are the spikes a problem? Is it a matter of aesthetics or is something bad really happening?
 

RPLaJeunesse

Joined Jul 29, 2018
252
There seems to be a common theme here. You need to use high frequency wiring techniques. My question is, why are the spikes a problem? Is it a matter of aesthetics or is something bad really happening?
Good question! Answer: It could be both. If the scope probe ground lead is long and hooked to the wrong point it may not show what is actually happening at the FET or the load resistor, so call that aesthetics. But you can't tell if that's the case, so something really bad may be happening as well. Power FETs can turn on in nanoseconds, so RF layout practices are called for. Minimize path lengths, maximize trace widths, remember the "skin effect" hurts the use of wires, bypass appropriately (both for voltage and peak RF currents), and observe good grounding practices to prevent ground loops and and inadvertant ground current paths.
 

Thread Starter

gglone56

Joined May 10, 2019
11
View attachment 195974 View attachment 195975
View attachment 195977
Additional details: The circuit, as shown above, is just one of 3 identical stages in parallel. As shown above the output (pink trace) is measured across R which is actually 2-5 Ohm resistors in series. The problem I see with the output are the over shoots. With a 12V DC supply IN, I'm getting spikes to 25V and down to -9V. The input signal has no over shoots until it is plugged into the circuit. Any help would be appreciated.
The long wires are from the 12V PS to the FET and from the FET Drain to the resistor. All wires from driver to gate of FET are short.

Welcome to AAC!
The spikes look like they're due to stray inductance. Not surprising with all those lengthy straggly wires. Try a tidier, compact layout and see if that improves things.

Edit: Crutschow types faster than me :)
The long wires are from the 12V PS to the Source of FET and from the Drain to the load. All other wires are short.
 

Thread Starter

gglone56

Joined May 10, 2019
11
You need to minimize the potential inductance of all leads: So the gate driver to FET gate must be very short. You should have a big capacitor for the 12V supply right next to the FET as well as a smaller decoupling capacitor to help short any high frequency spikes.
The 1uf capacitor is not enough for the 12V supply? Where exactly should both caps be placed.

There seems to be a common theme here. You need to use high frequency wiring techniques. My question is, why are the spikes a problem? Is it a matter of aesthetics or is something bad really happening?
I am attempting build a box that can run a test which will supply a varying frequency and duty cycle signal to the battery input of potentially many different automotive parts. The spikes are bad because these parts may not be able to withstand 25V spikes. We don't want the parts to fail the test due to these spikes.

Good question! Answer: It could be both. If the scope probe ground lead is long and hooked to the wrong point it may not show what is actually happening at the FET or the load resistor, so call that aesthetics. But you can't tell if that's the case, so something really bad may be happening as well. Power FETs can turn on in nanoseconds, so RF layout practices are called for. Minimize path lengths, maximize trace widths, remember the "skin effect" hurts the use of wires, bypass appropriately (both for voltage and peak RF currents), and observe good grounding practices to prevent ground loops and and inadvertant ground current paths.
How short? The long wires are at the source and drain sides of the FET. Because the end circuit (6 sections) requires 50 Amps to pass to the output, I need to take the six output currents straight to a hi current bus because the board can't handle it. It sounds like I will need to make a board that can handle the 50 Amps. Or could I just shorten it as much as possible?
 

RPLaJeunesse

Joined Jul 29, 2018
252
The 1uf capacitor is not enough for the 12V supply? Where exactly should both caps be placed.
Definitely not enough. Figure 1 of the MIC4451 datasheet shows a particular 1uF (Wima MK22) as well as two each 0.1uF ceramic bypass caps. That's just for the driver. Now consider your load and 12V power supply response. How long does the 12V supply take to recover from that 50A step? You need enough capacitance to supply 50A for that many micro- (milli-?) seconds while drooping not much more than maybe say, 1% (120mV). Since C=I*dT/dV that calculates out to 41600uF for a 100us holdup. Yes over 40,000uF!
 

crutschow

Joined Mar 14, 2008
34,281
Additionally, if you must run long wires (over a couple inches) that carry pulses, used tightly twisted pair wires with one wire the signal and the other the signal ground return.
That will minimize the wire inductance and the consequent ringing.
 

Thread Starter

gglone56

Joined May 10, 2019
11
The long wires are from the 12V PS to the FET and from the FET Drain to the resistor. All wires from driver to gate of FET are short.


The long wires are from the 12V PS to the Source of FET and from the Drain to the load. All other wires are short.
I cut all wires to the minimum (about 2 inches from the board) without any change to the output measured across the load resistor.
 

Thread Starter

gglone56

Joined May 10, 2019
11
Definitely not enough. Figure 1 of the MIC4451 datasheet shows a particular 1uF (Wima MK22) as well as two each 0.1uF ceramic bypass caps. That's just for the driver. Now consider your load and 12V power supply response. How long does the 12V supply take to recover from that 50A step? You need enough capacitance to supply 50A for that many micro- (milli-?) seconds while drooping not much more than maybe say, 1% (120mV). Since C=I*dT/dV that calculates out to 41600uF for a 100us holdup. Yes over 40,000uF!
I do see where I missed the 2 0.1 uf caps so thank you for that. I will have to order them. For the other point, your saying I will need a 42000uf electrolytic cap in parallel with the load resistor to clean up the output signal?
 

RPLaJeunesse

Joined Jul 29, 2018
252
I do see where I missed the 2 0.1 uf caps so thank you for that. I will have to order them. For the other point, your saying I will need a 42000uf electrolytic cap in parallel with the load resistor to clean up the output signal?
I'm saying you need 42000uF in parallel with the supply, preferably placed near the switching FET.
 

RPLaJeunesse

Joined Jul 29, 2018
252
This product might have sufficed if it was faster (only 60K Hz). I need 100K. Is there a faster version?
If you are attempting to switch at 100KHz, at the 50A current level, you are not switching power but rather designing a switching power supply, which is a field unto itself. For this task I suggest you seek out an uber-experienced switching power supply designer. Anything less will likely result in a lot of burned up parts. I've designed automotive switching supplies that are in production, and I wont work on any design more than 20A. You also need to familiarize yourself with SAE J1113 and similar standards that define how automotive 12V powered devices are to be tested.
 

MrAl

Joined Jun 17, 2014
11,389
Are you saying you need to switch 12 volts at 50 amps at 100kHz with a 12v power supply into a device to be tested?
What does the rise and fall times of the 100kHz wave have to be?

When a current is flowing into a load and then the load resistance is suddenly increased by a large factor you are bound to get some spikes showing up because there is energy stored in the inductance of the wires in every wire that carries current.
The defining expression is:
v=L*di/dt
where di is the proposed change in current over the time dt and L is the inductance and v is the voltage. Since dt is in the denominator the faster this current changes the higher the voltage developed. This also means that if you have any oscillations then this occurs over and over again until the oscillations stop.
MOSFETs are known to have gate/source problems with oscillations because the source terminal voltage (outside the package or even inside the package) gets modulated by the current and that n turn means that the MOSFET gate voltage is not constant but can modulate up and down (after all it's the gate to source voltage not the gate voltage to ground). This is caused by even small amounts of inductance in the source circuit.
The gate driver and source of the MOSFET must be very tight with as little wiring as possible. Of course this means that those wireless plugboard breadboards are not sufficient (as well as for other reasons). So the circuit should be hard wired at the very least for one thing, with the driver and MOSFET very tightly wired.

When you get into fast rising/falling currents another thing that comes up is lead inductance from the main power source to the MOSFET. These wires should be short and as noted in previous posts some local storage capacitance is in order. However, when it comes to fast rising currents flowing from capacitors another issue that becomes important is the cap ESR.
The cap ESR has to be small enough so that there is not a large voltage drop across that resistance or else it defeats the purpose of having a cap in the first place.
So decent size cap with low ESR.

Now there's always going to be some inductance in the gate/source circuit as well as in other places, so most of the time a snubber comes into the picture. A snubber acts to eat up the energy of the spike so as to minimize its amplitude while the energy in the inductances dissipates. The snubber is usually placed across the transistor drain source but you may want one across the load also. The one across the transistor protects the transistor. The one across the load protest the load. The design of the snubber varies considerably because the wiring in a new design is somewhat random. It's mostly based on the energy from the inductances.

Another idea is to look into the design of a 100kHz switching power converter.
 
Last edited:

Thread Starter

gglone56

Joined May 10, 2019
11
If you are attempting to switch at 100KHz, at the 50A current level, you are not switching power but rather designing a switching power supply, which is a field unto itself. For this task I suggest you seek out an uber-experienced switching power supply designer. Anything less will likely result in a lot of burned up parts. I've designed automotive switching supplies that are in production, and I wont work on any design more than 20A. You also need to familiarize yourself with SAE J1113 and similar standards that define how automotive 12V powered devices are to be tested.
I do need it to be a hi side switching power supply. This is required for a General Motors Test called, "Switched Battery Line". We currently use an amplifier but are technically switching the ground or Low Side switching which is technically not according to the test requirements. We have looked into purchasing a tester for this purpose, but management deems it too costly. If you could point me to a good designer or possibly someone that has already designed and built this, it would be greatly appreciated.
 

Thread Starter

gglone56

Joined May 10, 2019
11
Are you saying you need to switch 12 volts at 50 amps at 100kHz with a 12v power supply into a device to be tested?
What does the rise and fall times of the 100kHz wave have to be?

When a current is flowing into a load and then the load resistance is suddenly increased by a large factor you are bound to get some spikes showing up because there is energy stored in the inductance of the wires in every wire that carries current.
The defining expression is:
v=L*di/dt
where di is the proposed change in current over the time dt and L is the inductance and v is the voltage. Since dt is in the denominator the faster this current changes the higher the voltage developed. This also means that if you have any oscillations then this occurs over and over again until the oscillations stop.
MOSFETs are known to have gate/source problems with oscillations because the source terminal voltage (outside the package or even inside the package) gets modulated by the current and that n turn means that the MOSFET gate voltage is not constant but can modulate up and down (after all it's the gate to source voltage not the gate voltage to ground). This is caused by even small amounts of inductance in the source circuit.
The gate driver and source of the MOSFET must be very tight with as little wiring as possible. Of course this means that those wireless plugboard breadboards are not sufficient (as well as for other reasons). So the circuit should be hard wired at the very least for one thing, with the driver and MOSFET very tightly wired.

When you get into fast rising/falling currents another thing that comes up is lead inductance from the main power source to the MOSFET. These wires should be short and as noted in previous posts some local storage capacitance is in order. However, when it comes to fast rising currents flowing from capacitors another issue that becomes important is the cap ESR.
The cap ESR has to be small enough so that there is not a large voltage drop across that resistance or else it defeats the purpose of having a cap in the first place.
So decent size cap with low ESR.

Now there's always going to be some inductance in the gate/source circuit as well as in other places, so most of the time a snubber comes into the picture. A snubber acts to eat up the energy of the spike so as to minimize its amplitude while the energy in the inductances dissipates. The snubber is usually placed across the transistor drain source but you may want one across the load also. The one across the transistor protects the transistor. The one across the load protest the load. The design of the snubber varies considerably because the wiring in a new design is somewhat random. It's mostly based on the energy from the inductances.

Another idea is to look into the design of a 100kHz switching power converter.
Thank you very much for your thoughts. Between yours and rplajeunesse comments, it makes me quite sure this task is beyond my knowledge and skills. Do you know where I might look to get a hi side power supply capable of switching up to at least 100K Hz and capable of passing a maximum of 50A? Of course it doesn't have to be a power supply just a switching device because we have power supplies rated at 50A. And by the way, this is required for a General Motors test called, "Switched Battery Line".
 
Top