# 4 quadrant power supply for testing a large, low voltage capacitive load

#### RKropp

Joined Jan 5, 2017
6
I have an application where I need to charge a massive capacitive load (high ESR, high self discharge, unknown capacitance but >5 farad) to roughly +1.5V and then discharge it to roughly -1.5V. The switching frequency is roughly once every 10 minutes. Fully characterizing the load requires testing at both constant current and constant voltage. The early prototypes pull 3-5 amps for a short time when connected directly to a 40A 1.2V supply, but the full scale units might be 10X larger. I am trying to measure voltage and current while the load runs for multiple charge/discharge cycles to test different active materials used in constructing the capacitor. Some experiments run for multiple days, so automated data collection is definitely required. I am using two INA219 current sense modules with an Arduino writing data to an SD card.

Currently I am using two high current lab power supplies and switching between them with a DPDT mechanical relay configured as a sort-of H-Bridge. This gives me (manual) control of voltage setpoints and current limits in either direction. This setup works except during the time when the capacitor is providing power to the supply, which causes the supplies to lose control of the load and interferes with the data collected during the switching transient. I know that an active load would dissipate this power but I don't want to purchase additional lab equipment and bodge together some nightmare of relays and control to switch them in and out at the appropriate times. I have seen a few four quadrant power supplies, which seems like a good starting place, but the ones I have seen don't offer automated sequencing with separate current limits and voltage controls.

I am a chemical engineer but I know my way around basic electronics. I have been looking at the LT8714 four quadrant synchronous controller which seems to be pretty close to what I need, but I would have to implement some kind of configurable current limits for some tests. If it is worthwhile I could dive into the data sheet and figure it out but it would take me a long time to brush up on KiCad and SPICE to get something implemented.

My questions are as follows:

-> Is a four quadrant supply the right piece of equipment to build my testing apparatus around or is there a better method?
->Does anyone have knowledge of a vendor that makes an integrated unit that can address all 4 quadrants with optional current limiting and high current (>20A preferred) output?
-> Should I just bite the bullet and start designing something custom?

#### Attachments

• 1.8 MB Views: 7

#### crutschow

Joined Mar 14, 2008
25,976
Seems like it's unlikely you will find an off-the-shelf solution for your 4-quadrant requirement, so you may just have to start designing something yourself.
But perhaps someone else knows of something better.

#### Ian0

Joined Aug 7, 2020
1,096
So, all you have to do is charge a capacitance to +1.5V at a constant current, then discharge it to -1.5V at a constant current, the constant current being 50A maximum.
What is the approximate value of the ESR?

#### Irving

Joined Jan 30, 2016
1,070
Surely its "Charge from 0v to +1.5v then discharge to 0v, then charge to -1.5v then discharge to 0v, repeat..."?

#### djsfantasi

Joined Apr 11, 2010
7,187
Surely its "Charge from 0v to +1.5v then discharge to 0v, then charge to -1.5v then discharge to 0v, repeat..."?
Or charge to 1.5V... then charge to 3V and discharge to 0V

#### Ian0

Joined Aug 7, 2020
1,096
Or charge to 1.5V... then charge to 3V and discharge to 0V
Irving is more correct. If the capacitor has two terminals A and B, then after the first charge A will be positive with respect to B, then it will discharged, then charged with B positive with respect to A, then discharged. If it is simply charged to 3V, then the polarity never reverses, and I suspect that is important to the experiment.
Imagine it were an aluminium electrolytic. In the first case it would explode, and the second it wouldn't (unless it was a 2V capacitor)

#### Irving

Joined Jan 30, 2016
1,070
@RKropp - what's the tolerance on the 50A CC?

#### nsaspook

Joined Aug 27, 2009
7,850
Surely its "Charge from 0v to +1.5v then discharge to 0v, then charge to -1.5v then discharge to 0v, repeat..."?
If the actual load was a electrical capacitor. I suspect (as a WAG) it's a transducer of some sort that might actually be moving something with positive and negative displacement from a neutral position like a pump.

#### Irving

Joined Jan 30, 2016
1,070
The LT8714 might be a basis for your need. As it stands it doesn't do CC so, as you surmised, you need to do a 4-quadrant current -> voltage conversion for the feedback loop. There is some complexity there, plus 20 - 50A CC is way outside its native capabilities.

#### Ian0

Joined Aug 7, 2020
1,096
I like a discrete solution!
Transistors are NPN = BC847B and BCM847B or BC547B
PNP = BC857B and BCM857B or BC557B
MOSFETs are logic level 30V 60A - plenty to choose from
Meanwell make 5V 50A power supplies, or a pair of 6V lead acid batteries is a possibility.

I don‘t know the length of time it has to supply full current. If it can be indefinite, it will need some serious heatsinks.
Input signal is -1.5V to make it charge positive or+1.5V to make it charge negative or discharge positive.

#### Attachments

• 592.2 KB Views: 3
Last edited:

#### KeepItSimpleStupid

Joined Mar 4, 2014
4,369

#### Irving

Joined Jan 30, 2016
1,070
I like a discrete solution!
Transistors are NPN = BC847B and BCM847B or BC547B
PNP = BC857B and BCM857B or BC557B
MOSFETs are logic level 30V 60A - plenty to choose from
Meanwell make 5V 50A power supplies.

I don‘t know the length of time it has to supply full current. If it can be indefinite, it will need some serious heatsinks.
Input signal is -1.5V to make it charge positive or+1.5V to make it charge negative or discharge positive.
Interesting solution, lots of work to make it reality I suspect. You're right about the heatsinks, those MOSFETs at 20A will be pumping 100W, 250W at 50A (+30W in the sense resistors) so substantial heatsinking with forced air cooling will be necessary. (though not as bad as my 2.5kW active load!).

The right MOSFET will be critical, not just a low Vgs(th) but reasonably linear for Id 0 - 50A v Vgs 0 - 3v say, IRL2505, which is rated at 74A continuous with a thermal transfer junction->heatsink of 1.25degC/W and a max junction temp of 175degC. On a really good forced air heatsink of 0.5degC/W that represents a maximum dissipation at 30degC ambient of 82W or 16A @ 5v so will need a minimum of 3 MOSFETs in parallel. Don't be fooled by the very high current spec of some MOSFETs, the maximum current is actually dictated by the connections (package pin or wire bond) which for standard TO-220/TO-247 package is around 25A continuous (75A for the 'new' TO-247PLUS package). Fortunately MOSFETs don't suffer from thermal runaway so can simply be paralleled up pin to pin.

#### Irving

Joined Jan 30, 2016
1,070
Wow! This https://www.kepcopower.com/bop.htm series is what came to mind first,

You can also look at what's called an "Electronic Load"
When they don't give prices (POA) but quote $200 for the service manual and$150 for rack slides you know this is going to be silly expensive!

#### Ian0

Joined Aug 7, 2020
1,096
not just a low Vgs(th) but reasonably linear for Id 0 - 50A v Vgs 0 - 3v say
The feedback from the bipolar transistor monitoring the source resistor should keep it linear. I’d be more bothered about the bipolar transistor’s temperature coefficient as everything around it warms up, but RKropp didn’t say whether it was critical or not. Bob Pease has a neat circuit wih a schottky diode to cancel the tempco.
You're right about the heatsinks, those MOSFETs at 20A will be pumping 100W, 250W at 50A (+30W in the sense resistors) so substantial heatsinking with forced air cooling will be necessary
The original spec sort of implied it only ran briefly at full current, but I’d agree on big heatsink just in case it didn’t.
Fortunately MOSFETs don't suffer from thermal runaway so can simply be paralleled up pin to pin.
Unfortunately, you’re several decades out of date with this.
True when the MOSFET is fully switched on
True for lateral MOSFETs (Hitachi 2SK135/2SJ50 and descendants) and early vertical devices (2SK405, IRF510) but on more recent devices that Vgs zero-tempco point is at higher and higher currents. Below the zero-tempco point, in the linear region, Vgs(th) falls as temperature increases, so current increases and the device runs away. Manufacturers have only just started putting honest SOA curves in their data sheets showing the square-law slope like a bipolar transistor.
If using multiple MOSFETs, better to have a source resistor with each one.

Last edited:

#### Irving

Joined Jan 30, 2016
1,070
The feedback from the bipolar transistor monitoring the source resistor should keep it linear. I’d be more bothered about the bipolar transistor’s temperature coefficient as everything around it warms up, but RKropp didn’t say whether it was critical or not. Bob Pease has a neat circuit wih a schottky diode to cancel the tempco.
Agreed, but the feedback loop can only do so much. A highly non-linear variation, or a Vgs > say 4v for 50A won't deliver. Good point about the Tco. There are various ways to deal with it, including a more active op-amp based feedback.

#### Ian0

Joined Aug 7, 2020
1,096
Interesting solution, lots of work to make it reality I suspect.
But the parts are cheap (especially if the big heatsinks are not required) Depends whether he wants to save his time or his money!

#### Ian0

Joined Aug 7, 2020
1,096
Agreed, but the feedback loop can only do so much. A highly non-linear variation, or a Vgs > say 4v for 50A won't deliver. Good point about the Tco. There are various ways to deal with it, including a more active op-amp based feedback.
If we only have to keep it at one point on the curve, it should cope. Dealing with several different constant current choices may need the op-amp, then we have to consider keeping it stable.

#### Irving

Joined Jan 30, 2016
1,070
Unfortunately, you’re several decades out of date with this.
True when the MOSFET is fully switched on
True for lateral MOSFETs (Hitachi 2SK135/2SJ50 and descendants) and early vertical devices (2SK405, IRF510) but on more recent devices that Vgs zero-tempco point is at higher and higher currents. Below the zero-tempco point, in the linear region, Vgs(th) falls as temperature increases, so current increases and the device runs away. Manufacturers have only just started putting honest SOA curves in their data sheets showing the square-law slope like a bipolar transistor.
If using multiple MOSFETs, better to have a source resistor with each one.
I'm not, but you're correct. However for this device, even at the max junction temp and Vgs <3.5v (the inflexion point) three parallel devices would be OK, especially if the power supply was current limited to 50A. Four devices would be better to reduce the max Tj and alleviate the risk further. A small source resistor of say 0.05ohm might be beneficial but I don't think its essential here. Certainly I'd test fully to assess this before making it final.

#### Ian0

Joined Aug 7, 2020
1,096
I'm not, but you're correct. However for this device, even at the max junction temp and Vgs <3.5v (the inflexion point) three parallel devices would be OK, especially if the power supply was current limited to 50A. Four devices would be better to reduce the max Tj and alleviate the risk further. A small source resistor of say 0.05ohm might be beneficial but I don't think its essential here. Certainly I'd test fully to assess this before making it final.
The current sense resistor is only 0.012Ω for 50A, so replacing that by four 0.05Ω would do the job nicely