Author Topic: Howland Current Pump  (Read 4133 times)

0 Members and 1 Guest are viewing this topic.

Offline JoelKTopic starter

  • Newbie
  • Posts: 7
  • Country: de
Howland Current Pump
« on: December 09, 2019, 09:31:37 pm »
Hello everyone,

my master thesis has the advanced howland current pump as topic. I work since quite a long time on the topic because it started as a small project.

I don't just use a single advanced howland current pump, but two identical pumps in a bridge configuration so that one drives and one sinks the AC current at a time. There is also a common mode feedback, which does not really matter in this case. But what matters is an addtional opamp which I use to buffer the positive feedback path. This basically allows me trimming which I need due to the bridge configuration. The problem this addtional opamp brings, is its input impedance which is directly connected to the output of each current pump. This limits the output impedance of my current pump. Most FET input opamps do good at low frequencies but what I look for is a really low input capacity opamp. At the moment I use the LTC6268 which has 0.45pF common mode and 0.1pF differential input capacity. High GBW and high open loop gain is also important but at the moment the input impedance of the LTC6268 dominates all my results. Does anyone know an opamp with less input capacity? Analog Devices shows all the required data one their website, TI is not so nice.

Best regards and nice monday :/
Joel
 

Online Zero999

  • Super Contributor
  • ***
  • Posts: 20357
  • Country: gb
  • 0999
Re: Howland Current Pump
« Reply #1 on: December 09, 2019, 10:29:21 pm »
Schematic?

Are you talking about something similar to the circuit linked below, but without the booster transistors?
https://www.eevblog.com/forum/projects/need-help-with-bi-directional-constant-current-source-(100ma)/msg1972685/#msg1972685
 

Offline JoelKTopic starter

  • Newbie
  • Posts: 7
  • Country: de
Re: Howland Current Pump
« Reply #2 on: December 09, 2019, 11:19:49 pm »
 

Offline JoelKTopic starter

  • Newbie
  • Posts: 7
  • Country: de
Re: Howland Current Pump
« Reply #3 on: December 09, 2019, 11:35:13 pm »
Sorry, for clarity
 

Offline jbb

  • Super Contributor
  • ***
  • Posts: 1265
  • Country: nz
Re: Howland Current Pump
« Reply #4 on: December 09, 2019, 11:48:55 pm »
How much DC precision and bandwidth do you need? A bootstrapped JFET buffer might be an option...
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: Howland Current Pump
« Reply #5 on: December 09, 2019, 11:54:02 pm »
The usual techniques for increasing AC input impedance of a transistor follower apply however the operational amplifier is already doing these internally so I do not know how far that will take you.  A discrete cascode voltage follower with a DC servo loop to control offset might be better but will sure be more complicated.  Construction would be critical as these types of circuits tend to become VHF oscillators.

The first thing I would try is bootstrapping the voltage follower's supply connections from its output but this will only be effective at medium frequencies.  Still, maybe that will be enough for your application.

I was under the impression that when pushing DC and AC output impedance, circuit typologies other than the Howland Current Pump are used.
 

Offline JoelKTopic starter

  • Newbie
  • Posts: 7
  • Country: de
Re: Howland Current Pump
« Reply #6 on: December 10, 2019, 02:47:31 pm »
How much DC precision and bandwidth do you need? A bootstrapped JFET buffer might be an option...

I need about 80dB open loop gain at DC and a GBW above 300MHz would be good. Even lower specs would be possible. The open loop gain only has an effect at lower frequencies. At 1MHz the output impedance of the current pump is dominated by the GBW and input impedance. There is a certain point where either the GBW or the input impedance has more effect on the output impedance. So lowering the GBW to much does not work.
Do you have an example for such a bootstrapped JFET buffer?

The usual techniques for increasing AC input impedance of a transistor follower apply however the operational amplifier is already doing these internally so I do not know how far that will take you.  A discrete cascode voltage follower with a DC servo loop to control offset might be better but will sure be more complicated.  Construction would be critical as these types of circuits tend to become VHF oscillators.

The first thing I would try is bootstrapping the voltage follower's supply connections from its output but this will only be effective at medium frequencies.  Still, maybe that will be enough for your application.

I was under the impression that when pushing DC and AC output impedance, circuit typologies other than the Howland Current Pump are used.


Bootstrapping the supply rails is something I already thought about. But my problem is the limitation of a single 5V supply rail. With a signal swing of only +/-1V at the output and a rail-to-rail opamp for bootstrapping it would be possible. So that the feedback opamp is normally supplied by the rails 1V and 4V which rise to 2V and 5V in case of a possitive 1V swing at the current pump output.
But does this make sense or do I even decrease my specs when I use bootstrapping under this limitation?
The current pump is not intended to be used with DC if you mean 0Hz DC. I think the most used frequency will be about 1kHz but the goal is to design a wideband current pump to be used at many different frequencies.

Thank you for your replies!
 

Offline fitzgerald

  • Contributor
  • Posts: 18
Re: Howland Current Pump
« Reply #7 on: December 10, 2019, 04:39:33 pm »
The LTC6268 has a very low input capacitance of only 450 fF. If you look at the simplified schematic diagram on page 11 of the datasheet you can see that there is a special input buffer for the ESD input diodes to get such low input capacitance.

Are you sure the input capacitance of the LTC6268 is your problem? The impedance of 450 fF is >1 MOhm for frequencies below 350 kHz.
 

Offline JoelKTopic starter

  • Newbie
  • Posts: 7
  • Country: de
Re: Howland Current Pump
« Reply #8 on: December 10, 2019, 05:35:38 pm »
The LTC6268 has a very low input capacitance of only 450 fF. If you look at the simplified schematic diagram on page 11 of the datasheet you can see that there is a special input buffer for the ESD input diodes to get such low input capacitance.

Are you sure the input capacitance of the LTC6268 is your problem? The impedance of 450 fF is >1 MOhm for frequencies below 350 kHz.

Yes it is a problem cause the target is 1MOhm at 1MHz. But you are right that the LTC6268 is already pretty optimized. I know the target is also pretty high but I want to get a close as possible.
 

Offline jbb

  • Super Contributor
  • ***
  • Posts: 1265
  • Country: nz
Re: Howland Current Pump
« Reply #9 on: December 10, 2019, 06:03:09 pm »
....
Do you have an example for such a bootstrapped JFET buffer?
...
But my problem is the limitation of a single 5V supply rail.
...

Sorry, I don't really have one on hand.  But a JFET buffer might not be practical here, because you likely need >1V headroom for gate threshold on the drive side and again on the load side (assuminging matched dual JFET used).
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Howland Current Pump
« Reply #10 on: December 10, 2019, 06:58:27 pm »
Why do you need such a high impedance at high frequencies?

It can be possible to make such impedances, but it is very often possible to either compensate (e.g., instead of using a current source to measure an impedance, use an impedance divider) or null (say in a narrowband condition).

For example, you're using an op-amp in the first place, can't you tweak it to give a negative capacitance?  (Be careful it doesn't oscillate ;D )

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: Howland Current Pump
« Reply #11 on: December 10, 2019, 08:57:50 pm »
Do you have an example for such a bootstrapped JFET buffer?

Some oscilloscope high impedance input stages used bootstrapping to raise their input voltage range.  More often it is used with bipolar input stages to raise low frequency input impedance but the circuits for bipolar transistors still apply to a JFET design to control input capacitance which is done in operational amplifiers.  The Tektronix 7A12 and 7A13 are examples but I doubt they are very applicable to what you are trying to do.

I have also seen it done with active probes where it improves the precision of the FET buffer.

Below is an example from National Semicondcutor application note 32.  The DC precision could be considerably improved by adding a matched JFET current sink with matched series bipolar base-emitter junction.  Alternatively a DC servo loop could be used and I think without adding to the input capacitance.

Quote
Bootstrapping the supply rails is something I already thought about. But my problem is the limitation of a single 5V supply rail. With a signal swing of only +/-1V at the output and a rail-to-rail opamp for bootstrapping it would be possible. So that the feedback opamp is normally supplied by the rails 1V and 4V which rise to 2V and 5V in case of a possitive 1V swing at the current pump output.
But does this make sense or do I even decrease my specs when I use bootstrapping under this limitation?

A low supply voltage is going to make things more difficult.

Quote
The current pump is not intended to be used with DC if you mean 0Hz DC. I think the most used frequency will be about 1kHz but the goal is to design a wideband current pump to be used at many different frequencies.

What about adding a series output inductor to raise the output impedance at higher frequencies?
« Last Edit: December 10, 2019, 09:14:36 pm by David Hess »
 

Offline JoelKTopic starter

  • Newbie
  • Posts: 7
  • Country: de
Re: Howland Current Pump
« Reply #12 on: December 10, 2019, 09:08:33 pm »
Why do you need such a high impedance at high frequencies?

It can be possible to make such impedances, but it is very often possible to either compensate (e.g., instead of using a current source to measure an impedance, use an impedance divider) or null (say in a narrowband condition).

For example, you're using an op-amp in the first place, can't you tweak it to give a negative capacitance?  (Be careful it doesn't oscillate ;D )

Tim

I personally don't need it, but maybe someone else. It's a master thesis so part of it is research and to see whats possible. The output impedance is currently about 300kOhm at 300kHz. So the input impedance is definitely not the only problem. Right now I am trying uncompensated opamps like the LTC6268-10 together with lead-lag compensation. Thus stability is a difficult topic. What do you mean by tweaking the opamp itself? On the current prototyp is a negative impedance converter at the output of each current pump which is not populated (for later).
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: Howland Current Pump
« Reply #13 on: December 10, 2019, 09:29:12 pm »
Right now I am trying uncompensated opamps like the LTC6268-10 together with lead-lag compensation. Thus stability is a difficult topic.

I remember trying that to improve circuit performance before I learned better.  It always comes down to raising the noise gain to provide stability which has the same effect as if the input signal was attenuated and closed loop gain increased; the whole exercise is counterproductive.  There might be something to gain (haha) if you have access to the nodes required for external compensation but modern operational amplifiers do not provide access to them.

Quote
On the current prototyp is a negative impedance converter at the output of each current pump which is not populated (for later).

Doing a transformation to add a negative input capacitance can be done at lower frequencies but I did not propose it because at higher  frequencies it is difficult.
 

Offline JoelKTopic starter

  • Newbie
  • Posts: 7
  • Country: de
Re: Howland Current Pump
« Reply #14 on: December 10, 2019, 09:52:39 pm »
Below is an example from National Semicondcutor application note 32.  The DC precision could be considerably improved by adding a matched JFET current sink with matched series bipolar base-emitter junction.  Alternatively a DC servo loop could be used and I think without adding to the input capacitance.
But dont I get a simulare problem like trying to bootstrap the supply rails of the opamp? The input swing of the discret buffer won't be high neither.

What about adding a series output inductor to raise the output impedance at higher frequencies?
I will simulate this. Thank you!
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: Howland Current Pump
« Reply #15 on: December 10, 2019, 11:47:01 pm »
Below is an example from National Semicondcutor application note 32.  The DC precision could be considerably improved by adding a matched JFET current sink with matched series bipolar base-emitter junction.  Alternatively a DC servo loop could be used and I think without adding to the input capacitance.

But dont I get a simulare problem like trying to bootstrap the supply rails of the opamp? The input swing of the discret buffer won't be high neither.

Not all operational amplifier SPICE models support simulating bootstrapping the supply pins.

In the example shown, the bootstrapping is AC so can extend beyond the supply rails.

Quote
What about adding a series output inductor to raise the output impedance at higher frequencies?

I will simulate this. Thank you!

It is a common way to improve AC impedance in current sources and sinks.  Just to be clear, the inductor goes between the current pump and the load.  Over a narrow frequency range, a parallel LC parallel resonant circuit provides higher impedance.  At lower impedance and higher frequencies, a resistor might be used because it has a higher resonant frequency.
« Last Edit: December 11, 2019, 12:42:56 am by David Hess »
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Howland Current Pump
« Reply #16 on: December 11, 2019, 01:48:07 am »
Note that, for a source that has a Norton equivalent (current || resistance || capacitance), a series inductance increases the impedance but only around the resonant frequency; at best, at higher frequencies, the impedance is dominated by the damping of that inductance (which for critical damping, equals sqrt(L/C), and the impedance simply shelves at or near its peak).

In other words, you can shelve the impedance, but you can't keep it going up (inductive) forever.

Since this is a masters thesis -- perhaps it would be more beneficial to present the process one should follow to synthesize a desired source, or design and analyze a fixture?  These methods of improvement only scale so far, and depend critically on the components used.  There are limits, both among commercial supply, and sheer physics, on those components (especially GBW and stability where op-amps are involved), which simply cannot be as general as the more involved methods.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline coppercone2

  • Super Contributor
  • ***
  • Posts: 11336
  • Country: us
  • $
Re: Howland Current Pump
« Reply #17 on: December 11, 2019, 02:02:29 am »
there is a paper on this device for electromedical purposes if you care to find it (made in the 2000's )

it is called a circlotron (in this case op-amp based). the paper covers high Z at AC
« Last Edit: December 11, 2019, 02:05:59 am by coppercone2 »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf