Hi folks!
This is not really an audio question. I've built linear current led regulator (1A max, 2.5v-6v). It's working very fine and the voltage range is what I was looking for. The loss is really low (60mV).
But the problem is temp. Between 25°C and 70°C, I've an 100mA drift. I'm looking for the best way to ovoid or reduce this thermal drift at minimum. As well with minimum component addition.
The problem is coming from the bipolar transistor. If I cold it, the drift go away. But I can't in the real life.
May be the transistor is not the best to use, may be I should add a temp compensation circuit. No sure, no sure how to to this?
Any advice?
This is not really an audio question. I've built linear current led regulator (1A max, 2.5v-6v). It's working very fine and the voltage range is what I was looking for. The loss is really low (60mV).
But the problem is temp. Between 25°C and 70°C, I've an 100mA drift. I'm looking for the best way to ovoid or reduce this thermal drift at minimum. As well with minimum component addition.
The problem is coming from the bipolar transistor. If I cold it, the drift go away. But I can't in the real life.
May be the transistor is not the best to use, may be I should add a temp compensation circuit. No sure, no sure how to to this?
Any advice?
Attachments
Early morning armchair guess- you don't have to pull much current out of the base of Q1, and if the Hfe is high, it could be below the range for the 334, or the 334 gain could just be too low. I'd add a resistor from the base of Q1 to the supply (emitter) so the 334 runs over a narrower range and so Q1 can better turn off. I'd also look at the 10uF cap. If it's an electrolytic, the leakage current might be getting involved. Use a low leakage film type. Finally, make sure your current sensing resistors are something stable like metal film, not thick film or MOX. Hopefully one of those things will help, but IMO a far better current regulator can be done with an op-amp and MOSFET as originally published by Siliconix many years ago.
Resistor are all low ppm metal smd. Caps is XR5.
Curently, I've solved the problem using lower Rsense resistors as calculated to take care of the real internal temp in the box. I still have a 100mA difference as well at start.
Curently, I've solved the problem using lower Rsense resistors as calculated to take care of the real internal temp in the box. I still have a 100mA difference as well at start.
I think the issue is that Ic and Vbe of the transistor are temperature dependent. Thus simply forcing a base current you will be affected by both these temperature-sensitive variables.
Make the current source part of a potential divider or drive current through a voltage generator such as a zener, and feed the base from that.
Make the current source part of a potential divider or drive current through a voltage generator such as a zener, and feed the base from that.
From the Jan 1983 Siliconix MOSPOWER design catalog, the couple extra components make this circuit far more robust than just directly wiring the MOSFET like many people do. Apply whatever voltage to the input you need- their divider is just an example. Ditto the value of the sense resistor- use whatever is reasonable for your application.
Attachments
you seem to be stuck with a major conceptual error - yes collector current is a multiple of the base current (beta or Hfe), But the ratio is affected by everything, including temperature as you have demonstrated - so virtually no discrete transistor circuit relies on a Q having a specific value of beta (Hfe)
the most robust way to get a costant current from a bipolar tansistor is to use a resistor to 'measure' the emitter current and controlling the Q base voltage to regulate the emitter current sense resistor V, then the collector current is 1-1/beta times the measured and regulated emitter current - which only relies on beta being large for accuracy
http://www.diyaudio.com/forums/showthread.php?s=&threadid=37662&perpage=10&pagenumber=1
points to some ccs circuits
in
http://www.diyaudio.com/forums/showthread.php?postid=1007597#post1007597
sijosae' collection 1-11 are fine, 12-16,18 probably don't work, I would automatically use his fig 6,11 "feedback ccs" when it meets the application requirements
the most robust way to get a costant current from a bipolar tansistor is to use a resistor to 'measure' the emitter current and controlling the Q base voltage to regulate the emitter current sense resistor V, then the collector current is 1-1/beta times the measured and regulated emitter current - which only relies on beta being large for accuracy
http://www.diyaudio.com/forums/showthread.php?s=&threadid=37662&perpage=10&pagenumber=1
points to some ccs circuits
in
http://www.diyaudio.com/forums/showthread.php?postid=1007597#post1007597
sijosae' collection 1-11 are fine, 12-16,18 probably don't work, I would automatically use his fig 6,11 "feedback ccs" when it meets the application requirements
I think the idea of the original circuit is the 334 will sense and provide enough gain to swamp out the TC of the transistor, but that probably isn't reasonable. Q1 will certainly have a lot of temperature sensitive gain the way it's being used, and the 334 isn't a good choice for the "error amplifier". BTW, what's this circuit being used for?
Conrad Hoffman said:
As I wrote in the first post, this circuit is used to drive a power led. As I need the smaless possible size, minimum component are used. Linear current regulation is used to have something working with a large voltage and current range, minimum dropout voltage, easy to find component, fiability and DX use. Many dedicated step down converter or booster circuits do not have these qualities.
Add 1 or 2 components to the current design, or find a better transistor to reduce thermal drift can be done, not design a completly a new circuit with tens of components added...
No it can't. IMHO, the circuit is inherently flawed because it doesn't address the TC problem in the right place. This has to be done where the current is sensed and converted to a control signal. You could buy special high TC resistors, or add components to try and compensate it, but those schemes are inherently troublesome. All transistors have basically the same temperature dependency, so changing that isn't the answer. I've used the simple op-amp circuit for exactly what you describe (stabilized LEDs) in scientific instruments, and it works well with almost zero TC. The trick is to use a modern op-amp that operates near ground on both input and output, then you have a wide range of adjustment without using another supply. Or, use one of the discrete circuits described earlier that are typically used as current sources in audio amps. Their bandwidth will be better than the op-amp circuit, but their accuracy, though ok, won't be quite as good.
Thanks for the advice. In fact, the circuit is done and used. I can live with it. I should only take care of the temp drift, aka take care of the temp in the box before selecting the Rsense.
For next boards, I'll use the AMC7135 cheaper and efficient (but with higher dropout).
Do you have some op-amp ref which will be good to use for an opamp/mosfet version of the linear regulator? I've some small mosfet in stock and can try to build a prototype.
For next boards, I'll use the AMC7135 cheaper and efficient (but with higher dropout).
Do you have some op-amp ref which will be good to use for an opamp/mosfet version of the linear regulator? I've some small mosfet in stock and can try to build a prototype.
It depends on how much current you need, the value of the sense resistor, the adjustment range, the available supply, and the bandwidth. I usually use LF411/412, TL07x, OP-27. It should be something unity gain stable for least trouble. You only need special op-amps if you want to use a very low supply voltage, or if the MOSFET turns on with a very low gate voltage. Also, if you want to use a low value of sense resistor and a low current setting. Just look at the voltage on the sense resistor at the desired current setting, and make sure the op-amp inputs are ok at that voltage. No numbers off the top of my head, but any amp that advertises that the input CMR includes ground is a good bet. If you use dual supplies, none of that matters at all, but I assume most people want to go single.
If you are using an LM334 as a CCS on the base of the transistor you have chosen a temperature sensor. The current varies with temp.
Use a TL431 CCS, rock solid.
datasheetarchive.com will have the datasheet.
Use a TL431 CCS, rock solid.
datasheetarchive.com will have the datasheet.
- Status
- Not open for further replies.