high voltage programmable bidirectional current source circuit

[zac]

I'm working on a microcontroller controlled bidirectional constant current source capable of putting out up to +/-100V.  The maximum current requirement is about 10 ma, but it needs to also be controllable down to the tens of microamps.  I would like to have 100 khz (10 microsecond/DAC sample) frequency response, but 50 khz is probably adequate.  This will be for a small battery powered device. 

I'm considering a bootstrapped howland topology using an op-amp with a high common-mode voltage tolerance.  This would take a small bipolar voltage input and produce a bidirectional current output. 

It will need a switching power supply to provide the +/- 100V supply rails so I'm trying to figure out the best design for the power supply which will need to convert a 3-9V (1 or 2 rechargeable lithium cells) to +/- 100V (with current capability of least 10 ma). 

The built-in DAC in the microcontroller (an arm based STM32F407) is unipolar so would need to covert it to bipolar to drive the howland amp.  Any ideas on the best way to do that?  Alternatively, I could use an external DAC capable of bipolar output, but that would need an additional set of +/- low voltage power supplies.

The microcontroller would update the DAC every 10 microseconds (or 20 microseconds depending on output amp bandwidth ) to produce the desired current source waveform. 

Is this the best way to do this? 

This is an arbitrary waveform generator that is able to produce the desired current source output:

http://www.keithley.com/products/dcac/currentsource/highperfor/?mn=6221

Thanks,

Zac

[Gallymimu]

What are you making?  An electrotactile stimulator?

Posted on: August 26, 2013, 02:13:59 02:13 - Automerged
for your unipolar DAC you can just set the reference level of your howland amp to be half the full scale value of your DAC, that will give you bipolar output

[zac]

What are you making?  An electrotactile stimulator?[

It's a electrical stimulation device that is used like a TENS device.  Different waveforms and frequencies are supposed to be therapeutic.  

PS.  I reposted this topic on the projects forum.  Could someone move this over there?  Thanks.

Posted on: August 29, 2013, 02:41:48 02:41 - Automerged
I posted this in the electronics-general forum, but this is probably a better place so I'm reposting it here:

I'm working on a microcontroller controlled bidirectional constant current source capable of putting out up to +/-100V.  The maximum current requirement is about 10 ma, but it needs to also be controllable down to the tens of microamps.  I would like to have 100 khz (10 microsecond/DAC sample) frequency response, but 50 khz is probably adequate.  This will be for a small battery powered device.

I'm considering a bootstrapped howland topology using an op-amp with a high common-mode voltage tolerance.  This would take a small bipolar voltage input and produce a bidirectional current output.

It will need a switching power supply to provide the +/- 100V supply rails so I'm trying to figure out the best design for the power supply which will need to convert a 3-9V (1 or 2 rechargeable lithium cells) to +/- 100V (with current capability of least 10 ma).

The built-in DAC in the microcontroller (an arm based STM32F407) is unipolar so would need to covert it to bipolar to drive the howland amp.  Any ideas on the best way to do that?  Alternatively, I could use an external DAC capable of bipolar output, but that would need an additional set of +/- low voltage power supplies.

The microcontroller would update the DAC every 10 microseconds (or 20 microseconds depending on output amp bandwidth ) to produce the desired current source waveform.

Is this the best way to do this?

This is an arbitrary waveform generator that is able to produce the desired current source output:

http://www.keithley.com/products/dcac/currentsource/highperfor/?mn=6221

Thanks,

Zac

[koseyel]

The built-in DAC in the microcontroller (an arm based STM32F407) is unipolar so would need to covert it to bipolar to drive the howland amp.  Any ideas on the best way to do that?  Alternatively, I could use an external DAC capable of bipolar output, but that would need an additional set of +/- low voltage power supplies.

As Gallymimu said, you could bias the opam with Vcc/2 and you'll get bipolar ouput.
I'm working on a similar project but with lower voltage compliance. Still gathering infos to figure out the best way to design.

[bobcat1]

Hi Zac

There was application note (from Linear tech and article on EDN) regarding high voltage high current source design

May bee one or two years ago - As far as I remember not an easy stuff. 

All the best

Bobi

[Gallymimu]

We built one of these for skin stimulation.  Was pretty easy using an APEX high voltage amplifier and turning it into a horowitz Howland amp.

don't forget to make it fault tolerant and safe, you can kill someone.

do you have a particular application/customer or are you just playing around?

[zac]

We built one of these for skin stimulation.  Was pretty easy using an APEX high voltage amplifier and turning it into a horowitz amp.

don't forget to make it fault tolerant and safe, you can kill someone.

do you have a particular application/customer or are you just playing around?

The apex parts are too expensive for most applications though quite convenient to test a concept.  What is a horowitz amp? 

Safety should be assured by current limiting (to perhaps 10 ma) in the power supply for the howland pump in addition to limits/checking in the firmware.  Is there some other mechanism you would suggest? 

Yes, this is a for a specific application that is used like TENS (electrodes stuck on skin), but aren't supposed to go into the details.   

[Gallymimu]

The apex parts are too expensive for most applications though quite convenient to test a concept.  What is a horowitz amp?  

Safety should be assured by current limiting (to perhaps 10 ma) in the power supply for the howland pump in addition to limits/checking in the firmware.  Is there some other mechanism you would suggest?  

Yes, this is a for a specific application that is used like TENS (electrodes stuck on skin), but aren't supposed to go into the details.  

At a minimum you should be single fault tolerant.  Meaning, any single component or subsystem failure must not be able to create an unsafe condition i.e. not deliver more than 10mA (if that is the spec or safety limit).

A DC blocking cap sized properly can limit the current pulses as an extra layer of protection.

Sorry I meant Howland amp, Horowitz is an electronics book author, my bad.

Also leakage current under fault conditions is going to be a big one too for safety.

[LithiumOverdosE]

Safety should be assured by current limiting (to perhaps 10 ma) in the power supply for the howland pump in addition to limits/checking in the firmware.  Is there some other mechanism you would suggest?  

Yes, this is a for a specific application that is used like TENS (electrodes stuck on skin), but aren't supposed to go into the details.  

I'm sure that you're aware that 10 mA impulses through or near heart are quite capable of causing arrhythmia or worse. Also placing it near oesophagus can cause muscular constriction and choking. I'm sure that there are other places on body where muscular spasm could cause harm as well, perhaps cellular damage as well. 

I hope you know what you are doing.

Perhaps it might be a good idea to check following thread if you want to proceed further with this line of development.
http://www.sonsivri.to/forum/index.php?PHPSESSID=25f0l6l9jhou9tctp9oiu75642&topic=50642.0

[zac]

At a minimum you should be single fault tolerant.  Meaning, any single component or subsystem failure must not be able to create an unsafe condition i.e. not deliver more than 10mA (if that is the spec or safety limit).

A DC blocking cap sized properly can limit the current pulses as an extra layer of protection.

Sorry I meant Howland amp, Horowitz is an electronics book author, my bad.

Also leakage current under fault conditions is going to be a big one too for safety.

It would be single fault tolerant within the definition UL uses.  A reasonably sized DC blocking cap probably wouldn't work as some of the desired waveforms are very low frequency (almost DC) of only a few hertz.  

What other safety mechanisms would you suggest?

I have limits/checks in firmware, hardware current limiting in the high voltage (+/-100V) op-amp power supplies, and the inherent current limits in the design(gain) of the howland pump.  I'm considering adding programmable current limiters in the HV supply rails where the microcontroller would set up the current limits to be the maximum desired for a particular waveform.  Another idea is an analog integrator on the output that will shut off the HV rails if the delivered energy exceeds a threshold which the microcontroller could set.  This would be something in the the neighborhood of a few millijoules/second.  



Posted on: September 01, 2013, 08:40:01 20:40 - Automerged

I'm sure that you're aware that 10 mA impulses through or near heart are quite capable of causing arrhythmia or worse. Also placing it near oesophagus can cause muscular constriction and choking. I'm sure that there are other places on body where muscular spasm could cause harm as well, perhaps cellular damage as well.  

I hope you know what you are doing.

Perhaps it might be a good idea to check following thread if you want to proceed further with this line of development.
http://www.sonsivri.to/forum/index.php?PHPSESSID=25f0l6l9jhou9tctp9oiu75642&topic=50642.0

I was a principal engineer on 2 medical device projects and been involved with a few others to a lesser degree.  I found the safety implementation to be pretty inconsistent though perhaps better than with some other non-medical projects.  The people working on it at least thought about the safety aspects more because of the need to document everything.  

I had some interesting disagreements with some engineers about this.  Some of them with considerably more experience than myself had many incorrect concepts of electrical safety.  For example, one engineer with 25 years experience thought that he could be electrocuted by a 12V car battery.  To prove him wrong, I brought a car battery and a multimeter into a meeting and I held the terminals in my hands.  I even wet my hands and repeated the experiment.  The current was less than a milliamp and I didn't even feel it.  The inherent resistance of the human body precludes a dangerous current from a 12V power source.  (there could be an exception for those with pacemakers/defibs as a relatively small current could confuse the ecg sensing)

For reference, a cardioversion or defibrillator pulse is typically in the 100-300 joule range.  To deliver that, the voltage needs to be in the 200-1500V range.  The device I am working on is only intended to deliver millijoules.  The power supply (100V @ 10 ma) can deliver only a watt (1 joule/second).  Still, 1 joule/second could potentially be hazardous and probably painful so multiple safety mechanisms will be included.  

[LithiumOverdosE]

You do have an experience and that's great. 

I once saw what inexperienced designer did to himself while trying one of his "TENS" devices. Nerves can quite easily be damaged that's for sure.

[Gallymimu]

Zac,

Too bad that DC blocking might not work for you, but that said, I have to ask why you would be delivering current pulses of such a low frequency.  A lot of research has shown that a few hundred microsecond to about 1 millisecond pulses are optimal for nerve and muscle stimulation, concurrently, going over 1000Hz is also typically overkill as the neurons sodium/potasium pumps don't repolarize fast enough.  Usually in the 100s of Hz is also appropriate.  Whether it be sensory, muscular, or pain management I believe the above should be pretty consistent.

As for everything you mentioned for safety it sounds like you have it reasonably well covered, DC block would be a nice addition of course.  One word of warning.  Be careful of limits set by the microprocessor.  For instance if you have a current set point that get's garbled or calculated wrong, and that propagates to your other processor controlled current limits, that would end up being a single point of failure that negates several of your protections.  I typically cover my ass with a 3rd party review of the design before it gets too far, but also well before I would be seeking true 3rd party sign off.

What are you doing for ground isolation?  Or is this battery powered?

Can you disclose where on the body you are placing the electrodes?  Certainly localized stimulation with thoughtful electrode placement also GREATLY reduces the possibility of causing a spasm in a sensitive muscle (like the throat or heart!)

As you know TENS delivers quite a bit of current but it is done through localized electrode placement reducing the risk to minor burns associated with poor electrode contact.

I think it's implied that you are using bipolar stimulation, are you delivering a charge balanced waveform (probably are but just curious).

Is this for chronic or short term use (again just curious because we've had a bitch of a time with chronic use electrode options)

[zac]

Zac,

Too bad that DC blocking might not work for you, but that said, I have to ask why you would be delivering current pulses of such a low frequency.  A lot of research has shown that a few hundred microsecond to about 1 millisecond pulses are optimal for nerve and muscle stimulation, concurrently, going over 1000Hz is also typically overkill as the neurons sodium/potasium pumps don't repolarize fast enough.  Usually in the 100s of Hz is also appropriate.  Whether it be sensory, muscular, or pain management I believe the above should be pretty consistent.

The intent of this is not only to stimulate nerves.  One of the modalities that this device is intended to perform is called FSM (frequency specific microcurrent) which supposedly stimulates cell metabolism.  There are also many other waveforms that we want to evaluate for efficacy.  That is why I want to design this device as an arbitrary waveform generator which will allow many modalities to be tested.  We also want the ability to measure body resistance/impedance and look for any changes during or after the treatments.  

As for everything you mentioned for safety it sounds like you have it reasonably well covered, DC block would be a nice addition of course.  One word of warning.  Be careful of limits set by the microprocessor.  For instance if you have a current set point that get's garbled or calculated wrong, and that propagates to your other processor controlled current limits, that would end up being a single point of failure that negates several of your protections.  I typically cover my ass with a 3rd party review of the design before it gets too far, but also well before I would be seeking true 3rd party sign off.

Someone else is supposed to figure the legalities and whether we need to get 510K equivalence.   Some companies that make related devices do 510K as a TENS while others (mihealth, etc.) skip approval altogether and claim to be an exempt biofeedback device.  

I plan to have hardware current limts in the power supply and the software settable limits will be such that the highest possible selection is still within the safety criteria.  I also plan to have internal diagnostics that will exercise and verify that the safety mechanisms are working.

What are you doing for ground isolation?  Or is this battery powered?

It is battery powered only and we want to keep it small enough to be kept in a pocket or belt pouch.  

Can you disclose where on the body you are placing the electrodes?  Certainly localized stimulation with thoughtful electrode placement also GREATLY reduces the possibility of causing a spasm in a sensitive muscle (like the throat or heart!)

The electrodes are typically put in various position on the back and possibly on the hip and lower abdomen.  I don't know of any reason to place them in the heart area or on the neck.   There is a plan to test placement behind the ears and on the sides of the head.  

As you know TENS delivers quite a bit of current but it is done through localized electrode placement reducing the risk to minor burns associated with poor electrode contact.

In the FSM mode, this will deliver only a few hundred microamps which is typically not even felt.  I've tested several types of stick on electrodes and found there is a wide range of resistance between different brands.  I'm using the larger (2"x3.5") surface area types to reduce the current density for the other modes which use higher currents.  Dura-stick plus is the best I've found so far.  

I think it's implied that you are using bipolar stimulation, are you delivering a charge balanced waveform (probably are but just curious).

The devices will be able to do both though the bipolar stimulation is the typical use.  The waveform is typically balanced/symmetrical, but we plan to try unipolar and unbalanced waveforms as well.

Is this for chronic or short term use (again just curious because we've had a bitch of a time with chronic use electrode options)

The typical use is 20-40 minutes per session and no more than 1 session per day.  

What type of device have you been working on?  The TENS, FSM and related stuff is new to me.  I used it to fix my chronic shoulder pain earlier this year and decided to worked on an improved device in conjunction with a doctor friend.  I ended up using a function generator and an AC ammeter for my own treatment (all battery powered). 


Posted on: September 02, 2013, 07:22:55 07:22 - Automerged
Horowitz sounded familiar.  It's a nice reference:

http://home.ustc.edu.cn/~dtruijun/Pre-Lab/The_Art_of_Electronics.pdf

[LithiumOverdosE]

How do you plant to test the device? After all measurements are done and safety checks in place one will have to test it on a living subject.

I'm just curious in what way do you cover yourself legally if something happens during trial runs?

[zac]

How do you plant to test the device? After all measurements are done and safety checks in place one will have to test it on a living subject.

I'm just curious in what way do you cover yourself legally if something happens during trial runs?

Someone else is responsible for the legal issues, but my understanding is this device will probably be a 510K TENS device.  The initial testing will be on myself, my doctor friend and her family.  I don't think there is any real risk, but we will be covering ourselves legally.  For example, everything is done as a corporation.  I'm not sure about the plans to run clinical trials.

[Gallymimu]

I don't know much about FSM though it sounds interesting and you will surely see an effect.

As for impedance changes, I'm not sure what information you will gain from it, but you will definitely see impedance changes in the skin.  One the the things that is a pain for sensory stimulation is that as the skin is stimulated various pores tend to open up in the skin which change the sensations as current densities change and shoot through currents start sneaking through the opened pores in the skin.

your electrodes and duration of use shouldn't be a problem though with unipolar stimulation you might end up causing some skin irritation.

The 510K approach seems interesting to me in that you would be getting approval as a TENS device but it sounds like it isn't really how it would be used, so any use of it for what you intend would have to be either off label or approved separately?!

Our background in this area has been more on the sensory side and some muscle stim.  We developed some sensory feedback systems that translate pressure sensation in non-existent or non-sensate limbs to stimulus on available patches of skin such as the back or neck.

[zac]

I don't know much about FSM though it sounds interesting and you will surely see an effect.

As for impedance changes, I'm not sure what information you will gain from it, but you will definitely see impedance changes in the skin.  One the the things that is a pain for sensory stimulation is that as the skin is stimulated various pores tend to open up in the skin which change the sensations as current densities change and shoot through currents start sneaking through the opened pores in the skin.

How much current/joules were injected before the pores opened?  Was this observed with a microscope?

I'm not sure what impedance change will mean either.  This part of it is highly experimental.   

your electrodes and duration of use shouldn't be a problem though with unipolar stimulation you might end up causing some skin irritation.

What is the approximate threshold (duration/current or joules delivered) when you noticed irritation?

The 510K approach seems interesting to me in that you would be getting approval as a TENS device but it sounds like it isn't really how it would be used, so any use of it for what you intend would have to be either off label or approved separately?!

As far as I know, most devices of this type are not specifically approved.  They are used off label with a 510K approval as a TENS device or sometimes not at all.  We certainly don't have the funding at this time to conduct the extensive clinical trials to obtain specific approval. 

Our background in this area has been more on the sensory side and some muscle stim.  We developed some sensory feedback systems that translate pressure sensation in non-existent or non-sensate limbs to stimulus on available patches of skin such as the back or neck.

That's an interesting idea.  How much current/joules were needed to create a sensation?   Was the sensation controllable by the current or waveform? 

[Gallymimu]

So I can't give specific information related to the pores opening.  It's been too long.  Most of the knowledge on this was in published neuroscience literature.

Irritation generally occurred with a few mA or more of current delivery and occurred more rapidly with higher currents and more rapidly with unipolar stimulation.  Some reddening of the skin can be observed after a few minutes of stimulation but for true irritation I can't recall.  I do know we stimulated with AgCl electrodes for several hours and the skin was red, not sore, but also slightly numbed from the stimulation.

I can give better information for sensation for sure.  Thresholds for sensation are very dependent upon the person, area of skin used, and how long the skin has been moistened by the electrodes (again using AgCl with an electrode gel).  Thresholds were also frequency, and pulse width dependant, but in general, a few hundred microseconds, at 60-300Hz at around 0.5 to 1.5mA was usually sufficient to elicit vibration sensation (activation of pacinian corpuscle afferents).  This was with 4mm circular electrodes separated by 10mm.

Adjustment of the current level provided the greatest dynamic range of sensation (dynamic range defined as onset of sensation up through the stimulation becoming uncomfortable).  Adjustment of frequency also impacted intensity of sensation as well as modality of sensation.  We didn't try a variety of waveform shapes as the literature shows that bipolar rectangular pulses produces the best results for sensation.  There are also physiological reasons rectangular bipolar pulses produce the best activation of afferents and efferents.

Wow that post made me feel REALLY REALLY nerdy!

[Checksum8]

Here is a non scientific project for burning warts. It uses 21khz unipolar pulses at 25 volts. Current is limited to 3mA.

http://www.zen22142.zen.co.uk/Circuits/Misc/wart_zap/wart_zapper.htm

It seems your +/-100v at 10mA could do localized tissue damage if the electrodes were placed in close proximity?

[zac]

Here is a non scientific project for burning warts. It uses 21khz unipolar pulses at 25 volts. Current is limited to 3mA.

http://www.zen22142.zen.co.uk/Circuits/Misc/wart_zap/wart_zapper.htm

It seems your +/-100v at 10mA could do localized tissue damage if the electrodes were placed in close proximity?

The maximum peak power could be 1 watt. But, the average power is in the milliwatts.  Average current is typically in the hundreds of microamps and average voltage <10V. 

[Checksum8]

Yes, skin resistance and low duty cycles. We are talking low currents.

The author of the zapper, claims the average current is about 100uA and no more than 200uA.

[zac]

Yes, skin resistance and low duty cycles. We are talking low currents.

The author of the zapper, claims the average current is about 100uA and no more than 200uA.

The current density is very high as they're using a pointed electrode.  Presumably, this is intentional.  With the device I'm working on, I will be using large electrode surface area (at least 4 square inches) to reduce current density and minimize skin irritation.  For a wart, I imagine they want the opposite and are trying to irritate it as much as possible. 

[Gallymimu]

I love that wart zapper!  With the pointed electrode and high current density I'd imagine it hurts like hell.  Pain fibers (nociceptors) are the smallest so require the highest current density to activate but I bet that thing tickles them just fine.