I was absolutely not arguing against Thorstens suggestion of an LM317 CCS, but merely pointing to a very good, and equally simple CCS.
Thorstens point about availability in the junk box of the LM317 and TL431 is a good one.
DN2540 CCS cascoded with another DN2540, as SY is stating, will get you an incredibly simple to build CCS with even more incredible performance.
As it could be completed with a TL431 based shunt, in which the TL431 is only used as stable voltage reference, heavily bypassed for low noise.
Thorstens point about availability in the junk box of the LM317 and TL431 is a good one.
DN2540 CCS cascoded with another DN2540, as SY is stating, will get you an incredibly simple to build CCS with even more incredible performance.
As it could be completed with a TL431 based shunt, in which the TL431 is only used as stable voltage reference, heavily bypassed for low noise.
I still don't know what "it" is, John. Can you enlighten me as to what you mean?
I don't have a favorite regulator circuit- a regulator is one part of a whole design and the choice of topology and devices depends on what's needed for a particular design. I can't think of many that I haven't used in one place or another.
I don't have a favorite regulator circuit- a regulator is one part of a whole design and the choice of topology and devices depends on what's needed for a particular design. I can't think of many that I haven't used in one place or another.
John,
The fun part is, we do not usually need such a regulator.
While of course in theory zero noise, zero output z and so on are desirable, understanding which parameter matters for a given circuit and provide with an optimised circuit.
To wit, take an Audio Clock (Crystal) oscillator circuit.
No matter what you use, discrete, canned etc, they all draw essentially constant current, can tolerate some slow DC drift without degrading our performance, but if we want low jitter than very low noise is mandatory, on the other hand DC precision, output impedance and even transient response can be largely (within limits of course) be ignored.
Now the circuit that works exceptionally here would be a disaster if we attached it to the rail of a headphone amp, but who would do such a thing?
So, thankfully, we do not need regulators that do it all, we need to know what the circuit needs and we need to design techniques that allow us to give the circuit what it requires.
In fact, a regulator truly optimised for it's job will, for the critical parameter usually beat any "general purpose" regulator called "Super Regulator" or "Hyper Mega Regulator" or "I hype you one more regulator" and it will do so often with truely ignoble and generic parts and not even a lot of them...
Ciao T
The fun part is, we do not usually need such a regulator.
While of course in theory zero noise, zero output z and so on are desirable, understanding which parameter matters for a given circuit and provide with an optimised circuit.
To wit, take an Audio Clock (Crystal) oscillator circuit.
No matter what you use, discrete, canned etc, they all draw essentially constant current, can tolerate some slow DC drift without degrading our performance, but if we want low jitter than very low noise is mandatory, on the other hand DC precision, output impedance and even transient response can be largely (within limits of course) be ignored.
Now the circuit that works exceptionally here would be a disaster if we attached it to the rail of a headphone amp, but who would do such a thing?
So, thankfully, we do not need regulators that do it all, we need to know what the circuit needs and we need to design techniques that allow us to give the circuit what it requires.
In fact, a regulator truly optimised for it's job will, for the critical parameter usually beat any "general purpose" regulator called "Super Regulator" or "Hyper Mega Regulator" or "I hype you one more regulator" and it will do so often with truely ignoble and generic parts and not even a lot of them...
Ciao T
Well for a power amp, it would be OK. For a preamp, not really, for a pre-preamp (my territory) NO WAY! Yes, I like low noise, that makes me try harder to the best job possible, and why I get rewarded by doing the almost impossible, at times. You should see my latest power supply regulator designs. WAY over the top! (but necessary)
Because it's copyrighted.
Werner Ogiers did some interesting measurements here: Simple Voltage Regulators Part 1: Noise - [English] . Why don't you look at them, read some of his links, and discuss them?
Werner Ogiers did some interesting measurements here: Simple Voltage Regulators Part 1: Noise - [English] . Why don't you look at them, read some of his links, and discuss them?
There is no free lunch, John. Avalanche break-down produces noise. If we need voltage reference based on Zener diode, we'll get it, but buying one stable voltage source we are getting another one, noise source, for free. The only way is to filter them out using reactive elements. Even if we use MOSFET current source loaded on plain resistor, anyway we are getting noise that have to filter out if we don't want it.
However, all ducks can be killed by a single shot, when we use a shunt regulator, where Zener itself is shunted by caps, and we draw power from this Zener.
However, all ducks can be killed by a single shot, when we use a shunt regulator, where Zener itself is shunted by caps, and we draw power from this Zener.
Sounds good, Wavebourn. This is what 'intuition' gives us, over 'brute force'.
What I was concerned with was the using of a Zener and NOT buffering it with an RC filter. That is easy and cheap. Regulation can be maintained well enough, at least, AC regulation which is the most important.
What I was concerned with was the using of a Zener and NOT buffering it with an RC filter. That is easy and cheap. Regulation can be maintained well enough, at least, AC regulation which is the most important.
Right, and even when we use pass regulators we face the same problems, since pass regulator needs reference voltage source as well.
Onvinyl, EVEN IF his measurements were limited by his test equipment, they would not be good enough for most of my designs. Real-time designers, like Wavebourn and I, instinctively know, from experience and previous study, what the APPROXIMATE noise will be. Ogier is far from minimum. But he has improved over the typical.
- Status
- Not open for further replies.