Well Ken and Bonsai, this is essentially what I do with the Blowtorch power supply (each channel)
First, a 317/337 pair for hum reduction and voltage setting.
Second, an active single device (amazingly similar to the 2N4401 example) shunt regulator to remove transients, and to set a constant overall current load, I use 10 ohms rather than 15.
Third, a capacitor multiplier to reduce noise to as low a value as necessary (near the audio circuit itself). Works for me!
First, a 317/337 pair for hum reduction and voltage setting.
Second, an active single device (amazingly similar to the 2N4401 example) shunt regulator to remove transients, and to set a constant overall current load, I use 10 ohms rather than 15.
Third, a capacitor multiplier to reduce noise to as low a value as necessary (near the audio circuit itself). Works for me!
Last edited:
Hi Ed,
It does destroy the low Zout of the regulator before it, and for noise reduction it depends on the equality of that 15 ohm in series and the Rbe+Re. We both know how accurate/stable that Rbe is. Its unpredictable and unstable.
In my book that's just bad engineering - there are other ways to do something like that that IS predictable and stable.
My 2 eurocents worth.
jan
To my knowledge, capacitance multipliers are called that becuase the capacitor attached to the transistor base appears to be multiplied by the transistor's beta (hfe). Said another way, the A.C. current supplied from the emitter is determined by the capacitor connected to the base, except multiplied by the base-emitter current gain, making that capacitor appear to be larger in value by a factor equal to the beta. In real transistors, internal base-emitter resistance limits the minimum available output impedance to above what one would get from an actual large capacitor.
It sounds to me like you are describing some sort of grounded-base circuit, while a ripple-regulator/capacitance multiplier is essentially an emitter-follower. Ripple is fed to the circuit, yes, but it is shunted away from the base, creating a high A.C. impedance at the regulator's input and a much lower A.C. impedance at the output, filtering away the ripple. The point of this circuit is that is provides decent incoming noise rejection with just a few parts and doesn't contribute very much self noise.
Last edited:
John, what exactly were the performance reasons and implementation circumstances in which you choose to use a clean-up shunt instead of a capacitance multiplier, and vice-a-versa?
Thanks.
Hi,
Yes, this article is very good and shows interesting approaches.
Based on your own experience with these circuits, would you say that are particularly suited to be implemented the average DIY enthusiast, who has fairly minimal instrumentation, in preference of (say) using a simple choke/capacitor with (say) > 60dB ripple/noise filtering at 100Hz and approaching 100dB at 1KHz?
Ciao T
Yes, this article is very good and shows interesting approaches.
Based on your own experience with these circuits, would you say that are particularly suited to be implemented the average DIY enthusiast, who has fairly minimal instrumentation, in preference of (say) using a simple choke/capacitor with (say) > 60dB ripple/noise filtering at 100Hz and approaching 100dB at 1KHz?
Ciao T
Magic of words...
Think of capacitor multiplier as a resistor divider.
It is actually an emitter (or source) follower feed from R-C network. Output resistance is almost R/Beta, so you can use smaller cap with larger R value for the same output resistance and the same time constant. If you shunt capacitor by Zener it turns into a classic pass-regulator.
Similarly, if you shunt feedback cap in that "magic" ripple rejector by Zener, it turns into an "amplified Zener", i.e. an element of pass regulator.
Think of capacitor multiplier as a resistor divider.
It is actually an emitter (or source) follower feed from R-C network. Output resistance is almost R/Beta, so you can use smaller cap with larger R value for the same output resistance and the same time constant. If you shunt capacitor by Zener it turns into a classic pass-regulator.
Similarly, if you shunt feedback cap in that "magic" ripple rejector by Zener, it turns into an "amplified Zener", i.e. an element of pass regulator.
IIRC I got some 20 dB of rejection from the simple version. Anything more
required trimming and was not really stable over temp.
IMHO a capacitance multiplier delivers more.
The circuit has a problem: if the input voltage dies too fast, it may
drive the BE junction into Zener breakdown for large coupling capacitors.
That impairs the noise characteristics of the transistor for good.
regards, Gerhard
ed. on the other hand, one might cascade several of the Finesses
because of their low drop. A normal follower looses one diode drop
(maybe two for Darlingtons) and has 2 mV/°C drift per drop.
Last edited:
Jan,
We call that armchair quarter backing! The design is for use with low noise crystal oscillators. It is very close to a minimal design. The output impedance is not an issue in the design application.
That is why when you did your piece on balanced lines I didn't give you a hard time!
ES
Come on Ed, the large Vout will give a lot more load-induced noise from an oscillator that anything you can remove!
Let me put it this way: if you would apply for a job with me as audio designer, and you would propose this for a low noise supply, you would not get the job ;-)
Which may be my loss, but still.
jan
Jan, I read 'Stereophile' , 'TAS', 'AudioXpress' and 'The Audiophile Voice'. I am comp'd on three and pay $10/yr for 'Stereophile'. On occasion I read 'HFN' when I can find it. I keep an extensive collection of previous issues, and 'WW' and 'JAES' for historical purposes.
However, almost daily, I get input from other engineering magazines like:
'EDN'
'Electronic Design'
'Nasa Tech Briefs'
'Electronic Products'
'Test and Measurement'
'Evaluation Engineering'
'Microwave Journal'
'IEEE Spectrum'
'Microwaves &RF'
etc, etc
I also read 'Positive Feedback, online' where my good friend Teresa contributes.
However, almost daily, I get input from other engineering magazines like:
'EDN'
'Electronic Design'
'Nasa Tech Briefs'
'Electronic Products'
'Test and Measurement'
'Evaluation Engineering'
'Microwave Journal'
'IEEE Spectrum'
'Microwaves &RF'
etc, etc
I also read 'Positive Feedback, online' where my good friend Teresa contributes.
I assume you meant R not V, and this is not a design for an audio power supply!
Remember rule #2. "When the boss is wrong refer to rule#1!"
Hi,
I found that in practice 20dB are achievable with reasonable ease, however the circuit is quite sensitive to differences in the transistors (manufacturers, batches etc.) so without adjusting the real circuit with the real parts it is not a given that even the 20dB will attained.
Higher attenuation into the 30dB+ region is possible but now you need to explicitly tune for a specific batch of transistors and even then you may have quite a number of rejects.
Trimming these Finesse Noise Shunts in practice is a challenge, as there is very little ready instrumentation that cab be used. The low noise pre-amplifier also shown on the Wenzel site is an excellent tool for low noise measurements, however it too is not "ejit-proof" to build (we now use building one of these as test for new engineers to see if they are competent - most fail miserably) and may need trimming.
Another issue is that the output impedance of the "Finesse Noise Shunt" becomes quite high. This is of little concern in the intended application (cleaning up noise from the supply for a clock oscillator), but may not be acceptable for some other circuits.
The large dependence of the noise rejection on the precise active parts led to me designing the dual cascaded Finesse Noise Shunt I originally used out and replacing it with a circuit of my own design that has almost exactly the same component count as the dual Finesse noise Shunt, similar noise limits (self noise of a 2N4401 transistor) but around 70dB to 80dB noise rejection without any trimming and minimal component dependence.
This design is a simple capacitor follower (around 35dB noise rejection) combined with an AC current source and an AC Shunt (combined over 46dB noise rejection). The output impedance is also around 1/10 of that of the Finesse Noise Shunt.
Of course, it is again just another dumb, stupid and obvious application of the CCS & Shunt Reg principle, no finess, no innovation, just brute force and reliable noise killing.
And credit is due to Mr Wenzel, without his article and his circuits quirks I might have never tried and instead used just another dumb, stupid and obvious application of the op-amp & pass transistor circuitry instead.
Ciao T
I found that in practice 20dB are achievable with reasonable ease, however the circuit is quite sensitive to differences in the transistors (manufacturers, batches etc.) so without adjusting the real circuit with the real parts it is not a given that even the 20dB will attained.
Higher attenuation into the 30dB+ region is possible but now you need to explicitly tune for a specific batch of transistors and even then you may have quite a number of rejects.
Trimming these Finesse Noise Shunts in practice is a challenge, as there is very little ready instrumentation that cab be used. The low noise pre-amplifier also shown on the Wenzel site is an excellent tool for low noise measurements, however it too is not "ejit-proof" to build (we now use building one of these as test for new engineers to see if they are competent - most fail miserably) and may need trimming.
Another issue is that the output impedance of the "Finesse Noise Shunt" becomes quite high. This is of little concern in the intended application (cleaning up noise from the supply for a clock oscillator), but may not be acceptable for some other circuits.
The large dependence of the noise rejection on the precise active parts led to me designing the dual cascaded Finesse Noise Shunt I originally used out and replacing it with a circuit of my own design that has almost exactly the same component count as the dual Finesse noise Shunt, similar noise limits (self noise of a 2N4401 transistor) but around 70dB to 80dB noise rejection without any trimming and minimal component dependence.
This design is a simple capacitor follower (around 35dB noise rejection) combined with an AC current source and an AC Shunt (combined over 46dB noise rejection). The output impedance is also around 1/10 of that of the Finesse Noise Shunt.
Of course, it is again just another dumb, stupid and obvious application of the CCS & Shunt Reg principle, no finess, no innovation, just brute force and reliable noise killing.
And credit is due to Mr Wenzel, without his article and his circuits quirks I might have never tried and instead used just another dumb, stupid and obvious application of the op-amp & pass transistor circuitry instead.
Ciao T
Yes Rout, or Zout.
No matter whether it is for audio or not, any design that works, sort of, depending on temp, and with carefull adjustment unit-to-unit, is bad engineering.
My rule # 1.
jan
Sorry John, you know I look only at content, not titles.
I live by the adage from a person you also respect: the late Dick Feyman.
I quote: Never believe anything that is based on authority, tradition or religion.
My rule # 1, & 2.
BTW Thanks for the list of publications. I agree that some of those 'general design' pubs often publish articles pertinent to audio design.
jan
Audio design is part of the general electronic engineering world. I get new ideas, and understanding of the fundamental concepts of what is going on from ALL the publications that I scan. It helps me keep an open mind to new developments, like those in wire, quantum devices, etc. It makes more sense of new audio tweaks when I see that the world outside audio is also using strange concepts (at least strange from what I learned in college) and is going forward with them.
Feynman was a great physicist, I have most of his books, and Jack Bybee tells me about working with him, and HOW he would always try something before condemning it outright. You take Richard Feynman out of context, in my opinion.
Feynman was a great physicist, I have most of his books, and Jack Bybee tells me about working with him, and HOW he would always try something before condemning it outright. You take Richard Feynman out of context, in my opinion.
Since I seem to be misunderstood.
A "ripple eater" is NOT a cap multiplier.
See what I understand as a cap multiplier:
Using a low noise transistor or JFet, you get very low noise at the output, of course at the expense of a not so low output impedance...
Attachments
Speculation because you haven't read it yourself.
Let me fill you in. It is a literal quote from his letter to his daughter on her birthday, don't remember the exact age I believe it was 13th.
The letter was to prepare her for her adult life where she would be herself responsible for separating myth from reasoned fact and take firmly ground-based decisions, in the light of assaults by lots of quacks, snake-oil salesmen and religious fanatics. Just hope she steers clear from audio
.jan
- Status
- Not open for further replies.