I want to get a steady rising voltage, which is simple as constant current. Say, like a saw tooth but only first rising part and really slow. (I do not want a timer, but a rising voltage)
I was thinking of something like this using the Vbe of a transistor
formatting link
Also a similar thing can be done with an op-amp
formatting link
But I was just wondering, are there other ways of doing this?
That works but won't be very temperature stable. A zener or bandgap instead of the two diodes would improve it.
An LED can be the voltage reference instead of two diodes. That can temperature compensate Vbe. And it lights up! I think I did that a while back and it worked pretty well.
That may oscillate. There are fixes for that. Use a bandgap as the reference. And a mosfet to eliminate base current error.
A depletion fet, LND150 type, can be a simple constant-current source. LND150 is about 1.6 mA but a source resistor can scale that down.
A 3-terminal voltage regulator, LM317 type, and a resistor make a constant-current source.
An opamp follower and a bootstrap current source is nice. That can make nanosecond linear ramps. Bootstraps are fun.
An opamp integrator makes a nice ramp.
Just a voltage, a resistor, and a cap make an exponential rise, which is sorta linear. That's easy and can go very fast. Just live with or compensate for the curvature. Add an inductor to make that more linear.
If someone suggests a current mirror, it's probably a bad idea.
The ultimate horror: a PNP operating in beta-limited constant-current mode. One transistor and one resistor. Beta-bracketed transistors like BCX71 make that slightly less awful.
I did a linear ramp recently with a PV optocoupler driving a cap and an n-channel mosfet follower, with amps of drive capability. It ramps all the way up to the 48-volt power rail. It's a soft-start circuit for a power inverter.
You want to use a dual transistor so both transistor junctions are at the same temperature. By cascoding one or both of the transistors you can do better.
formatting link
It all depends on what kind of performance you need.
John Larkin thinks that he can design electronic circuits. People who actually can are more fond of current mirrors, but they tend to recommend the three transistor Wilson current mirror or it's four transistor elaboration. Integrated circuits are full of current mirrors, but their designers have a lot more things that they can play with
It is the sort of gross over-simplification that only gets mentioned for completeness.
MOSFETs can need a gate-stopper resistor - 10R/22R mostly - in series with the gate to discourage them from oscillating.
Relying on the collector impedance of a BJT over a wide voltage range will get you roughly 1-10% nonlinearity, depending on the device and conditions, on account of the Early effect.
A big resistor in series with the emitter helps a lot, but doesn’t fix the next problem, which is that the beta also varies with Vce.
You can fix that problem by using a FET instead. Their nonlinearity is much worse than a BJT’s, so you absolutely need an op amp wrapped round it, with close attention to stability as JL says.
Another approach to slowish ramps is a bootstrap current source. You hang a voltage reference on the output of an op amp follower, and connect a resistor from the reference to the follower’s input. In principle, that makes a perfectly constant current.
You can make it pretty nearly perfect in real life too, by choosing a FET op amp with really good common mode rejection, such as the OPA140.
The two remaining problems are the reset circuit and dielectric absorption in the capacitor.
I do that for my fastest linear ramps, nanosecond stuff. It avoids making a really good fast low-capacitance constant-current source, which can be a nuisance.
Some fraction of the world's fine-pitch ICs are fabbed by a controller that includes this:
formatting link
Apologies for the trimpot; we had to do this in a rush.
You can concoct a current source out of a three-terminal voltage regulator. Connect the regulator's Vout to the top of a reference resistor and its Gnd/Adj/Set to the bottom. Then Iout = Vout / Rref.
Yes. AIUI they're more something that gets built in to integrated circuits since they don't work well when built from discrete components because they're not exactly matched, and not thermally linked together.
Non-EE academics seem fixated on current mirrors, as seen in the scientific literature. They work fairly well inside ICs but not when built with discretes. Even the mirrors inside ICs don't usually need to be very good.
Most dual transistors are really two die, not thermally coupled.
Zetex sells a sort of current mirror thing. Somebody should sell a good packaged IC mirror but I don't know of one.
For a slow ramp, you want a polypropylene or (better) polystyrene cap. I recently built a slowish, very linear, adjustable ramp generator for characterizing a 12-bit ADC on a chip we've been collaborating on.
I started with wound mylar (polyster) caps, and found that the after a
100-us ramp reset, the output waveform exhibited absorption tails about
2 ms long. (That is, the voltage just sort of sat there for a couple of milliseconds before starting to rise again.)
It wasn't the current source--that was clean as a whistle--but rather dielectric absorption (aka soakage) in the capacitor.
Replacing it with a 630V polyprop Y cap (0.22 uF iirc) reduced but didn't eliminate the tails, so I just made the reset pulse 5 ms long. That reduced the maximum rep rate, but cleaned up the ramp amazingly.
(BTW Newark has some nice-looking 10 nF polyprop caps on super deal just now.)
Yeah, there a number of ways to do it. This one was a low voltage gizmo where the main worry was DNL. Using an integrator ramp made it easy to sort the DNL out from the noise.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.