Constant Current Source With 1 MHz Frequency Problem

found this....

http://www.sciencedirect.com/science/article/pii/S2212670813001127

Assuming the two links provided describe your goal...

...that's kind of scary.

You want to make a biomedical device to presumably to plug into something "bio"...yet you don't have a starting point (not provided, anyway) and have resorted to a 1-post-account on a proclaimed engineering forum (dubious from time to time) to get your answer...?

Maybe you should talk to our friend Napoleon...

To quote Old Salt: "good luck"

If you are attempting an experiment on live subjects or if it will be in contact with people, you need to follow various safety standards regarding insulation, leakage currents, and other protections.

Your current source must have a limited voltage compliance to stay safe. In general, you don't want more than ~48V. Hire an expert if you are in this range or above.

Assuming that the tissues are not on a live subject, you may use a simple commercial function generator set near its max amplitude (usually 20-30V) with a suitable resistor in series to limit the current. Assuming that the impedance of the tissues is low, you would need a resistor of 20V/1mA=20kohms. 1/8 to 1/4Watt is big enough. This simple setup would keep the current almost constant even with a low voltage drop on the tissues. If the exact current is important, you can always regulate with the generator's amplitude. Use two channels from an oscilloscope in differential mode to measure the voltage across the resistor and keep it at 20V to maintain your 1mA test current.

Play safe and have fun.

What is the: output voltage?/bio impedance range?

Can you add a block diagram

SolarEagle's first reference suggests replacing a Howland current source with an alternate op-amp-output-based current source. Both circuits are disasters at high frequencies, because they attempt to use a feedback scheme to turn an op-amp output, with its intrinsically-low-impedance, into a high-impedance current-source output. A true current source should continue delivering the desired current no matter what happens to the load; if the load's voltage drop changes the current should continue unchanged.

Op-amp outputs present a special difficulty at high frequencies: the loop gain is falling, degrading its operational capabilities, and slew-rate limitations prevent its output from rapidly changing. (An aside, in Howland-style op-amp current sources these effects can be modeled by assuming there's a large capacitor from the current-source output to ground. Such a capacitor is not a good thing!)

SolarEagle's second reference is also a disaster. What you need is a design based on true transconductance devices (i.e., voltage-to-current), operating in their native high-output-impedance mode, to do the trick. The collector outputs of ordinary bipolar junction transistors (BJTs) have such a property, and they also have low capacitance. What we want to do is use the transistor in what's called the common-base, or cascode configuration, with the input signal going to the emitter and the output taken from the collector.

The simple design at the right delivers a current proportional to the input voltage, Iout = - Vin / 5 kΩ. With the input at zero volts, the transistors each have a little over 1mA flowing; these currents cancel, for zero output current. The push-pull circuit delivers up to ±2 mA, with a compliance range of over ±10 volts, and it can work to well over 10 MHz. It'll have an gain accuracy of about 1%, an output impedance of over 5 MΩ and an output capacitance of under 10 pF. Increase the current capability by scaling the resistor values.

This simple design suffers from a few deficiencies. It prefers regulated, matched supply voltages, and it's also slightly non-linear. That's because each transistor's V_BE is not fixed, but changes with current, by about 60mV per decade. This flaw can be fixed by replacing each transistor with a two-transistor Sziklai configuration (see H&H AoE page 95).

I prefer a design using BJTs with op-amps, in a precision configuration to deliver a programmed current, but these are considerably-more complex.

You want a 1 Mhz signal riding on a 1 mA constant (averaged) DC current or an AC 1mA (RMS or average) current "source"? Or a square wave absolute 1mA current sourse? Not too clear, and not feel like educating myself on "bioimpedance" specifics right now. S.M.