In last Bloomberg magazine is an article claiming that American manufacturing is almost competitive, just a bit of fuel price increase and we are here! But how to calculate cost of loss of schools?
Sad news, Scott.
Here is a finessed, nuanced, engineering approach to the problem.
http://www.wenzel.com/documents/finesse.html
http://www.wenzel.com/documents/finesse.html
Last edited:
I am afraid the gap between 'almost' and 'is' is huge if there is any difference in price for like quality.
Mr Wenzel has been around mostly outside the audio regime, a smart guy there is more to the world than audio.
Scott, I read just about every engineering publication that I can subscribe to. I am still a life member of the IEEE, and of course, I look at audio publications too. Please consider this, I remind you that this is an AUDIO website, AND I was trying to make the point that the LM317 is problematic when used without 'finessing'.
Shunt regulator, in form of filter only, single-ended and asymmetrical. It can serve as a voltage stabilizer as well. In such case pass regulator can be easily transformed into a current source, and we are back at the beginning of the discussion.
I think Wentzel uses the word 'finesse' in his app note for a reason Wavebourn.
That means doing exactly what needs to be done to remove the noise and HF garbage from the PSU line. It turns out that you only need a handful of parts, and its mostly fast small signal stuff you have to deal with. Looking at it another way, he turns an LM317+finnessing componenets into a high performance wideband, low noiseeries regulator. You get all the benefits of the 317 (s/c proof, thermal s/down, 3 pin package, good current capability, reliability, fairly decent out the box regulation etc), and then you take it up a notch or two.
I think this is a great alternative to a current source/shunt or a milli farad cap and a 10H choke approach and deserves serious consideration.
That means doing exactly what needs to be done to remove the noise and HF garbage from the PSU line. It turns out that you only need a handful of parts, and its mostly fast small signal stuff you have to deal with. Looking at it another way, he turns an LM317+finnessing componenets into a high performance wideband, low noiseeries regulator. You get all the benefits of the 317 (s/c proof, thermal s/down, 3 pin package, good current capability, reliability, fairly decent out the box regulation etc), and then you take it up a notch or two.
I think this is a great alternative to a current source/shunt or a milli farad cap and a 10H choke approach and deserves serious consideration.
John what publications would that be? They're getting awfully scarce from where I stand.
Always appreciate a good tip.
jan
I did some sims about 2 years ago to look at the noise reduction part (the 'finesse' - so not with the 317 or similar in the model. I had to fiddle a bit with the values as a result, but the concept did work. I think this is something you have do in a practical design with a scope.
I had sim'd the single transistor version sometime ago and found that it wasn't as effective at removing incoming noise as a sim'd ripple-regulator (aka, cap. multiplier). Output impedance tends to be higher for the clean-up shunt than the ripple-regulator as well, at least, in their respective single transistor versions.
The clean-up shunt appeared best for applications where the primary D.C. regulators are pre-existing and already fixed at the load operating voltages. If, however, there is the opportunity to set the primary D.C. regulators several volts above that actually required by the load, then the ripple-regulator appreared to perform much better, as far as noise rejection.
Hey, great link! I had submitted my last comment, which echoes those findings, before seeing it.
In the 3rd figure in the article, it shows a minimum of 36dB attenuation between about 100Hz and 500Khz. The attenuation null can be much lower, dependent upon the exact value of R4. So, my interpretation of this circuits performance is not that the null can vary from -36dB to -70dB, but that for a handful of cheap components, you can get 36dB of noise attenuation from 100Hz to 500kHz. That's a circa 50x reduction. Not bad.
I cannot comment on the distortion point raised in the document - I would need to investigate. For the op-amp version of Charles Wentzel's circuit, I think the op-amp will have a significant effect on performance - speed, loop gain etc. I have played with shunt regs in LTSpice using op-amps and you can get quite some variation in bandwidth and speed (pulsed loads of c. 200-400mA). For this application (i.e. a 'finesser') I would tend to try to stick with a discrete approach.
Just my two cents worth.
I cannot comment on the distortion point raised in the document - I would need to investigate. For the op-amp version of Charles Wentzel's circuit, I think the op-amp will have a significant effect on performance - speed, loop gain etc. I have played with shunt regs in LTSpice using op-amps and you can get quite some variation in bandwidth and speed (pulsed loads of c. 200-400mA). For this application (i.e. a 'finesser') I would tend to try to stick with a discrete approach.
Just my two cents worth.
Last edited:
Strange. A cap multiplier in my book is something fundamentally different.
In a cap multiplier, you feed the circuit from the emitter, the stage before is connected with the collector. The base is connected to the midpoint of a divider, formed by a resistor and a cap. You do not want ripple to be fed to the base.
So something more "passive".
A ripple eater is insofar more "active" as it feeds the ripple to a transistor which shunts it away.....you are feeding the base with ripple on purpose.
Last edited:
Thanks. Looks like a good idea in principle but not usefull in a practical implementation.
You see that often with this type of feedforward circuits. They depend too much on matching and/or exact measured/adjusted values.
There's not much that can beat feedback ;-)
jan
Last edited:
This circuit is used and works well for what it is intended to do. It removes noise of interest for it's design application.
It is not a regulator or capacitor multiplier (AKA a capacitance gyrator or synthetic inductor).
(This is a bit of a hot button as after I published my tutorial on shunt regulators aka "The Forgotten Circuit" there was an issue about who stole it first and that fellow was using a ripple eater and did not know the difference.)
There is often confusion over just what a circuit does.
Shunt regulators and series regulators are intended to provide a fixed voltage, hopefully with low output impedance and noise. Of course how you model the impedance depends on the shunt or series issue.
A capacitor multiplier use a small voltage drop and current gain in a series active device to reduce ripple and some noise. It does not have a particularly low output impedance or at most minimal regulation. This is actually useful in battery powered circuits as it continues to work as the battery voltage falls. (In a battery circuit the primary purpose is to reduce voltage rail noise from other circuits or loads on the battery.) It really behaves like an inductor and an analysis of the circuit puts it into the gyrator class.
The type of circuit shown is used only to remove specific noise of interest. That is because as T has mentioned the gain at the frequency of interest should be close to one. This ripple eater does not provide a stable voltage or replace an inductor (or simulation of one.)
Last edited:
- Status
- Not open for further replies.