Will my opamp 100ns 100ma Current source work?

[AQUAMAN]

I want to make a current pulse of 100ns and 100mA to a capacitor of about 20-50nF. I have chosen an LT1363 opamp and simulated it in LTspice (http://cds.linear.com/docs/en/datasheet/1363fa.pdf)

I chose this because 1000V/us slew rate, high output current, and designed for capacitive loads. I want to control it with a pulse generator.

Most of the voltage controlled current sources Ive seen on the internet use two opamps, one in the feed back path too - but in these simulations it seems unneseccary and more unstable?

Also my op amp load is just the 47ohm resistor and the capacitor which has an ESR of about 1ohm, in the datasheet they specify specs with loads of about 500ohms (this seems to be how i get 100mA out, as in the data sheet it only says 60mA)

What do you think of my circuit?

My simulation and waveforms are attached

[Andy Watson]

Do you want to limit the current pulse, or do you want to control it? If it's the latter, the changing output voltage is going affect the current - so it's not really a defined current source. If you really do want a current source, look up Howland current source - this configuration converts a voltage to a current with very good compliance. Either way, 100nS is going to be tricky to attain.

[AQUAMAN]

Do you want to limit the current pulse, or do you want to control it? If it's the latter, the changing output voltage is going affect the current - so it's not really a defined current source. If you really do want a current source, look up Howland current source - this configuration converts a voltage to a current with very good compliance. Either way, 100nS is going to be tricky to attain.

I just want a repeatable 100mA current pulse, I dont really mind the shape as long as its repeatable

Basically I am using it to estimate the ESR of small capacitors to mohm accuracy. I will take a voltage reading at the end of this 100ns pulse.

[c4757p]

Just for fun, I tried it a different way. How's this look? Both R6 and R1 will need to be 1/2W resistors.

Note that it's not very accurate. Another one coming shortly!

Edit: Actually, how accurate do you need it to be? I think you could do a pretty decent job of compensating the temperature coefficients of D1 and Q1 to make it actually pretty good. (You'd need to calibrate for the actual current in software, unless you really want to use a half-watt trimmer for R2...) My second idea isn't looking so good.

[AQUAMAN]

Just for fun, I tried it a different way. How's this look? Both R6 and R1 will need to be 1/2W resistors.

Note that it's not very accurate. Another one coming shortly!

That looks great

I have tried to simulate this with transistors, albeit just an opamp controlling an npn transistor

But when I put a temperature sweep it seems to go wrong. So I put a 2m0hm per C sensitivity to temperature on the capacitor ESR, and during the temperature sweep I should then see the voltage across this capacitor increase (ESR increasing), but with these transistors it seems that the temperature effects their gain and it is then the current pulse is not repeatable throughout the temperature rang

[c4757p]

The current depends on the Zener voltage of D1 and the V_BE drop of Q1. Hold on, I'll have a go at compensating it.

[AQUAMAN]

Just for fun, I tried it a different way. How's this look? Both R6 and R1 will need to be 1/2W resistors.

Note that it's not very accurate. Another one coming shortly!

Edit: Actually, how accurate do you need it to be? I think you could do a pretty decent job of compensating the temperature coefficients of D1 and Q1 to make it actually pretty good. (You'd need to calibrate for the actual current in software, unless you really want to use a half-watt trimmer for R2...) My second idea isn't looking so good.

I have just seen your edit

Yes I think with transistors I will have to do some temperature compensation? Although this could be difficult, it could provide opportunity for more current which would make my ESR measurment easier.

I dont need it to be accurate, but I need it to be repeatable. I dont mind if the current is 105mA every time instead of 100mA as long as there is very little variation with temp etc

I am measuring hte ESR to mohm accuracy, so if sensitivity is 1mohm per degree, at 100mA that is 0.1mV per C I have to measure... it needs to be quite repeatable

Do you think transistors are the better option over this opamp?

[c4757p]

OK, here it is with temperature compensation. Note that this is absolutely something you will need to test in real life! We're in "SPICE ain't that good" territory right now. And these pulses are fast enough to make testing on a breadboard damn near impossible, though you could test the temperature compensation that way.

One thing to keep in mind is that Q1 dissipates 500mW when the current is switched on, so if you are running this at a high duty cycle, you may have additional temperature dependency issues. Run a short series of fast pulses and then let it cool.

I'd prefer the transistors over the op amp easily. This sort of thing requires quite a lot from the op amp, you'll need a fast, good one and a very well-designed circuit.

The second circuit I was trying out was an op amp-controlled current source that is always sourcing, and a switch to divert the current off of the capacitor. As seen in the VCO in my function generator project. But as in that project, minimizing the current spikes when the switches activate is quite difficult.

An "ideal" choice for D1 would be a 6.2V Zener diode. Their temperature coefficient is close to the inverse of a standard silicon diode, so the two in series are nicely compensated. However, 5.6V is already quite large, increasing the power consumption and decreasing the maximum output voltage and the linearity (the closer the output voltage comes to the maximum output voltage, the more the current starts to deviate). Tradeoffs...

[AQUAMAN]

I will try to study your circuit. I am amazed that you came up with this in minutes. I dont undestand all of it.

Yes the duty cycle will be about 5%, but I do think there could be some issues with the temperature dependency of this transistor due to the environment I want to operate this in.

What sort of problems do you envisage with this opamp configuration? Is my design reasonable however?

I think there must be some advantages in opamp - most notably component count! (and maybe need for temp compensation??)

Up till now I have used a pulse generator with a 10V pulse to and 47ohm resistor before the capacitor in order to get the 100mA to measure the ESR, one of the problems I have encountered is indeed the current spike due to the inductance and then oscillations in my voltage reading on the oscilloscope

[c4757p]

What sort of problems do you envisage with this opamp configuration? Is my design reasonable however?

I think there must be some advantages in opamp - most notably component count! (and maybe need for temp compensation??)

First of all, that op amp isn't rated to put out 100mA. In fact, they say the typical short circuit current is 105mA and can be as low as 70mA. Though you could use an external buffer transistor.

An op amp solution could certainly work. You'd have to play around with it a bit, though, it can be tricky to get right. (So can this, though!) Keep trying. It's not an unreasonable solution.

Another problem with the op amp solution is that the amplifier may saturate when set to zero output current (particularly with a buffer transistor, which permits it to only source and not sink), making the switching messy and slow. This could possibly be solved with a properly biased class AB output buffer. It needs to be able to sink current as well as source to allow it to balance itself at one particular point, rather than just "anything less than the transistor's threshold will do".

Yep, component count and the lower temperature dependence.

I will try to study your circuit. I am amazed that you came up with this in minutes. I dont undestand all of it.

I came up with it quickly because it's closely related to a couple things I've worked on recently. Hold on a minute, I'll write up an explanation of it. I don't want to just dump a complex circuit on you and say "off you go now!"
Working backwards, because it's less dangerous than walking backwards: Q1 acts as a simple voltage-controlled current source: the base voltage is duplicated at the emitter (minus V_BE), and Ohm's law gives the output current.

D1+D5 acts as the voltage reference, with R3 as the load. A Zener diode can clamp and unclamp very fast. C1 and R4 compensate the capacitance of the K(D5)/B(Q1) node, allowing it to switch quickly. (C1 without R4 injects large current pulses through Q1's base.)

M2 and M1 act as a totem-pole switch to control that voltage reference. Q2 acts as a phase inverter, giving the input signal and its inverse to switch the totem pole transistors - it is operated between the active region and cutoff (never saturating), allowing it to run quickly. Q3/Q4 and Q5/Q6 drive the MOSFET gates.

M2's gate must be pulled about 4V above its source to switch it on, and its source must be able to reach the power supply voltage. Because of this, D4 is used to create a lower "power supply" for the output stage, giving some headroom above that to swing the gate.

Temperature compensation: Simple - the Zener diode is chosen to have the lowest temperature coefficient possible (5.6V and 5.1V are good), and D5 cancels out Q1be. Looking at the datasheet shows that 5.1V is slightly better, so you might try that instead.

[c4757p]

Edit: Remove D3, change R6 to 330. This circuit was just on the verge of not working right - a bit of variation in resistors or in M1 could have caused malfunction.

Also: If Q2 is replaced by a BFS17W, the potential operating speed increases significantly. SPICE gave me reasonably clean 20ns pulses, though it may have been bullshitting me. This cleans up the edges as well, so that might be desirable anyway. It's a cheap RF transistor.

And a note: with pulses this fast and somewhat high current, the PCB layout will be important. Minimize the inductance around Q1e, R1, and between R1 and C2 (add a smaller ceramic decoupling cap, or a few of them, as well). You may in fact find that even if you can get nice, clean 100mA pulses, they may be unusable for your intended application. You sure as hell do not want probe leads. The DUT will need to be connected very near to Q1's collector.

[AQUAMAN]

Thank you very much I will read all of this tomorrow

[AQUAMAN]

One issue I have is that there appears not to be any feedback

So as my load is a capacitor, and the voltage drop across this is actually rising a few hundred mV in those few nanoseconds - this will change the amplitude of the current pulse as the current through that transistor is also dependent on my load.

Perhaps I am worrying too much about this and it doesn't really matter that much. After all this principle works fine with a pulse generator and that has no feedback (that I know of).

I want to measure ESR of capacitors to mOhm accuracy (I have the measurement system set up fine) at different temperatures. Given that this ESR could change about 20mOhm over a 30-130C temperature range, perhaps feedback for the current pulse is not as much of an issue I think it is?

[c4757p]

The current through the transistor is indeed dependent on the load, but barely. An emitter follower makes a pretty decent VCCS for low output voltage swings.

I think you should investigate other methods, though. Like I said, the inductance will be quite unfriendly. Have you considered some sort of impedance bridge?

[AQUAMAN]

What do you mean by impedance bridge? You mean to measure the ESR? I am quite inexperiences in this side of instrumentation/measurement electrical engineering

I remember reading this terminology I think when I was researching ways to measure the ESR of a cap, but I have decided on this constant current method and I have proof of concept of this with my pulse generator for what I want to achieve

With the inductance - indeed there were massive overshoots and ringing/oscillations all along that 'linear' part of the voltage waveform as the cap charges that spice doesnt not come close to simulating - even if I try to include large parasitic components in the simulation.

I dealt with this problem by increasing the rise and fall times of my pulse from 5ns to 20ns - this brought the issues down to an acceptable level. I would hope that there is possibility to control this in my final design. Indeed I thought this was another advantage of that opamp - I could slow down the turn on, at least in simulation.

I am going to build your circuit in LTspice over the weekend and try to get my opamp/transistor based combos to work better

[awallin]

A few time-to-amplitude circuits over here:
http://ko4bb.com/~bruce/TAC.html

or have a look at the PICTIC time-to-amplitude circuit here( page 8/8 in the PDF):
http://ko4bb.com/dokuwiki/lib/exe/fetch.php?id=precision_timing%3Apictic&cache=cache&media=precision_timing:simple_pictic.pdf

The strategy in these circuits is to keep the current source running at all times. When we do not want a pulse a BJT/FET or other switch routes the current to ground.
When we want a (fast) current pulse we use the switch to route the current to the cap instead of ground.