Choosing a Mosfet for a Constant Current H-bridge

Thread Starter

danielantonic

Joined Sep 22, 2019
68
Hello all!

I am making a Frequency Specific Microcurrent (FSM) device (similar to a TENS machine) as my first analog electronics project.

I borrowed a CellRegulogie5 from Biobedded Systems GmbH for my wife and found health benefits as she has various medical issues (product info - http://proceptsystem.biobedded.de/pdf/CellRegulogie5_HandbuchVers01.pdf - it is in German, stats on page 9)

However, it is very expensive and I like the challenge of trying to make my own.

The machine works by outputting constant current between 100-600uA in symmetrical biphasic DC pulses of set frequencies (10hz - 10khz). To keep the constant current for human skin resistance, the HV is around 25-30V.

I have uploaded videos I made using my DCScope. Its not the best resolution/quality, but it gives me an idea of what the machine is doing - if anyone is interested the links are:
http://www.dcnsolutions.net/FSM/20191114_105749.mp4
http://www.dcnsolutions.net/FSM/20191114_105827.mp4
http://www.dcnsolutions.net/FSM/20191114_105905.mp4

In order to replicate this I am using an embedded Arduino Mega and custom PCB (see Board.jpg, Board2.jpg) powered with a rechargeable LiPo.

Each channel has a Mosfet H bridge (CH1.png). The electrodes are connected to CH_OUT directly. The drains of Q3 and Q5 are designated CH_CTRL. This is where I had the idea of running CH_CTRL through another IRFZ24N Mosfet (CC Mosfet) to control the constant current. This is done by attenuating the output of a AD9833 Function generator via MCP4131 digipots (Attenuation.png). I thought I could exploit the Mosfet's threshold voltage between between off and on to vary the amount of current.

The output of the AD9833 is attenuated relative to ground, fed through MCP6004 rail-to-rail opamps to tune the signal (via R57) and offset by the 3.2V threshold voltage (see thresh.png) to make a pulsed DC waveform. This then drives the CC Mosfet that controls the H-bridge via CH_CTRL (label CH5 in the picture).

When I designed the PCB I wanted to keep my options open for changing configurations, so I have 4 pairs of test points J40-J43:
Sense (J40) - 120R
FB (J41) - closed
SIG (J42) - closed
Return (J43) - open

The above configuration worked perfectly in my initial tests... until I realized that I was testing with no load resulting in improper grounding.

I tried the next configuration which resulted in no output at all:
Sense (J40) - 120R
FB (J41) - closed
SIG (J42) - 10k
Return (J43) - open

I think this will work if I change the CC Mosfet to a different model than the IRFZ24N but I'm not sure what to look for (other than a max voltage rating of at least 30V, the max current under 1mA, 10 and 10k hz operating frequency etc.)

Thanks in advance - any advice and feedback is greatly appreciated
 

Attachments

Ian0

Joined Aug 7, 2020
9,502
Why do you need a 17Amp MOSFET for a few hundred microamps?
A simple complementary emitter follower would do the job.
Screenshot from 2023-10-07 07-31-10.pngIn fact, if you don't need particularly square squarewaves LM358 would manage up to a few kHz. If you want a constant current output, then perhaps a Howland current pump?
 

Ian0

Joined Aug 7, 2020
9,502
Another possibility would be the obsolete CA3080, but the CA3080 lives on as part of the LM13700.
Ignoring the darlington buffer, you have a transconductance amplifier with a high slew rate and an output current that you can control up to 2mA and a maximum supply voltage of 36V.
 

Thread Starter

danielantonic

Joined Sep 22, 2019
68
Why do you need a 17Amp MOSFET for a few hundred microamps?
A simple complementary emitter follower would do the job.
View attachment 304358In fact, if you don't need particularly square squarewaves LM358 would manage up to a few kHz. If you want a constant current output, then perhaps a Howland current pump?
Thanks for that. That is essentially what i have except overkill mosfets in the actual bridge.
How would you control the current level via the microcontroller?
 

AnalogKid

Joined Aug 1, 2013
10,944
Here is a very basic concept schematic of one approach to the problem. All of the part numbers are subject to change

This uses a feedback loop to control the constant current rather than relying on the stability and repeatability of a specific portion of a device's transconductance curve.

U1 and Q1 form a traditional constant-current sink that is independent of the switching bridge. As shown, a sink current of 600 uA requires an input voltage of 6 V. R1 can be changed for a different input voltage range.

Q2 - Q5 are controlled by full-bridge control signals.

Because the output current is so low, the U1-Q1 circuit could be replaced by a standard Howland current pump. Note that a Howland circuit requires very tight matching of its feedback resistances to achieve a high output impedance.

ak


FSM-Device-1-c.gif
 

Thread Starter

danielantonic

Joined Sep 22, 2019
68
Here is a very basic concept schematic of one approach to the problem. All of the part numbers are subject to change

This uses a feedback loop to control the constant current rather than relying on the stability and repeatability of a specific portion of a device's transconductance curve.

U1 and Q1 form a traditional constant-current sink that is independent of the switching bridge. As shown, a sink current of 600 uA requires an input voltage of 6 V. R1 can be changed for a different input voltage range.

Q2 - Q5 are controlled by full-bridge control signals.

Because the output current is so low, the U1-Q1 circuit could be replaced by a standard Howland current pump. Note that a Howland circuit requires very tight matching of its feedback resistances to achieve a high output impedance.

ak


View attachment 304414
Thanks for that - its similar to my design, I thought that you would need to offset the signal by the mosfet's threshold voltage though...?

I have modified my circuit as you suggested using 2N7000 mosfets, however my oscilloscope just died so I'm waiting for my new one to be shipped out. Will keep you posted
 

AnalogKid

Joined Aug 1, 2013
10,944
The range of Vgs for many MOSFETS is +/-20 V. If the source is a few volts above GND, and you pull the gate to GND to turn it off, that is not a problem.

Big question: What is the min - max resistance going to be between the Pads nodes? This will set the min / max voltage from the Q3-Q4 sources to GND.

ak
 

Thread Starter

danielantonic

Joined Sep 22, 2019
68
The range of Vgs for many MOSFETS is +/-20 V. If the source is a few volts above GND, and you pull the gate to GND to turn it off, that is not a problem.

Big question: What is the min - max resistance going to be between the Pads nodes? This will set the min / max voltage from the Q3-Q4 sources to GND.

ak
Ah good point - human skin resistance can be anywhere from 10k to 600k.
 

AnalogKid

Joined Aug 1, 2013
10,944
The extremes are 100 uA and 600 uA, and 10 K and 600 K. This means that with Ohm's Law, the voltage between the Pads nodes is anywhere from:

E = I x R

a) 100 uA x 10 K = 1 V

b) 600 uA x 10 K = 6 V

c) 100 uA x 600 K = 60 V

d) 600 uA x 600 K = 360 V

At this point, it looks like your 30 V system cannot reach all four corners of your stimulus space. Also, for conditions a and b, The Vgs resulting from the Q2 and Q3 gates being pulled down to GND would exceed 20 V.

On paper, the solution is to move the constant current circuit to the top of the H bridge, but that just shifts the problem to somewhere else. Also, the input stages of most opamps do not operate close enough to the positive rail for this approach. There are FET driver chips that are designed for this, but there still remains the problem of conditions c and d.

ak
 

Thread Starter

danielantonic

Joined Sep 22, 2019
68
Thanks for that. I can adjust the HV side up to about 40V i think (using an LT1072). The machine I based this on only goes up to 25V, which is probably why they want you to moisten the pads before applying. This would bring the resistance down somewhat but I think a max voltage of 40V will cover it.

Would that mean that I'm looking for mosfets with a minimum 40V Vgs?
 

Ian0

Joined Aug 7, 2020
9,502
No obsession - still learning :)
So use BJTs for the H-bridge, but an N Channel Mosfet for the constant current?
You could use either
But if you want a less complicated circuit - look at the LM13700. It can do the whole lot (both sides of the bridge) in one IC.
 

AnalogKid

Joined Aug 1, 2013
10,944
BJTs everywhere are fine. I went with a FET in the current sink because your question was about FETs and there already were two of the same part number in the bridge. 2N4401/4403, 2N3904/3906, etc.

The LM13700 is a great part, but the for ward transconductance can vary over a 2:1 range. I think you would need some form a feedback to stabilize the output currents. For the two sections to drive the load differentially, would you need bipolar, tracking current programming voltages, an additional circuit complexity?

ak
 

Ian0

Joined Aug 7, 2020
9,502
The LM13700 is a great part, but the for ward transconductance can vary over a 2:1 range. I think you would need some form a feedback to stabilize the output currents. For the two sections to drive the load differentially, would you need bipolar, tracking current programming voltages, an additional circuit complexity?

ak
it would depend on what level of absolute accuracy is required.
I assumed (maybe wrongly) that the user adjusts the current to achieve a certain level or response. In which case, a 2:1 variation may be acceptable.
Also, it‘s a pretty good bet that the two devices on one chip would have some transconductance matching.
 

AnalogKid

Joined Aug 1, 2013
10,944
OR -

An H-bridge is a trick to effectively double the apparent differential voltage across a load. But the goal is a controlled bipolar current with a relatively high (for simple opamp circuits) voltage compliance. Given the very low current and power levels, what about a single high voltage opamp and +/-40 V supplies? Apex and Burr-Brown still make them (National's LM143/343 would be perfect if you can find them), or you can put a couple of 60 V transistors around a normal high-precision opamp.

There are jillions of opamps with very low offset and bias errors, so that part is easy. Opamp "output booster" and "current booster" circuits have been around since the 60's. Fun fact, one of the versions of the Crown DC-300 is basically just that, an opamp with a dozen power transistors wrapped around it.

https://how-to.fandom.com/wiki/How_to_boost_the_output_voltage_swing_of_an_operational_amplifier

The smallest 24 V transformer you can buy will give you +/-33 V rails, slightly less with regulation; but given that the circuit is a current regulator, stiff rails are not mandatory.

I'll redo the #6 circuit. Basically, delete all 5 transistors, and put the load between the opamp output and the top of R3.

ak
 
Last edited:

AnalogKid

Joined Aug 1, 2013
10,944
How is the current to be adjusted? A simple pot, computer control, a variable amplitude of the input square wave, other - ?

ak
 
Top