R8 sets Vce for denoiser BJT but higher makes for higher startup swing. I had to tame it with few diodes in series across the denoiser output cap.
I also tried playing with C6 value but didn't get much, not worth the extra part hassle.
R9 at 47R does dominate the noise, it's higher than BC8x7 BJTs can do. In all my testing of also poking at that BJT base while powered on I managed to damage a single BJT, and it was my fault. I usually don't install that resistor to keep noise as low as possible but I do usually add that footprint to my designs, just in case.
Without the startup swing limiting diodes the startup swing can go really high with Vce=Vout-1.25V or so.
If you don't mind the poorer tempco you can completely drop the 220uF capacitor, and your R3/R4/R5 chain, and set Vout from your R8.
Here's a sim with the startup swing, it's nonexistent with the DC-coupled Denoisator.
edit: you need IR LED for CCS if going with the DC-coupled Denoisator.
@tombo56
Noise depends on a few factors, BJT and R9 in his case. With BC817 and Denoisator with CCS I found some older measurements
I usually add some 0.2R for LM317 output cap ESR but ESL is also important, at least in the higher gain NoNoiser version.
Capacitance also seems to play a role (at least from simulation) with current spike in denoiser BJT base. In some configurations it can exceed its rating (over 100mA pulse if I remember right), without the 47R protection resistor. Higher output capacitor value seems to tame this startup current spike.
edit: at least from simulation. and BC8x7 seem rated for a max of 100mA base current short pulse.
A general reminder: thus far, I have found no 317/337 models including realistic noise values, which is why I added an auxiliary voltage source.
Sims based on 317 models are always hugely overoptimistic.
The improvement for a regular denoiser is 25~30dB, 50 to 60dB for a nonoiser or a dienoiser.
If you find better figures, they are most likely an illusion
Sims based on 317 models are always hugely overoptimistic.
The improvement for a regular denoiser is 25~30dB, 50 to 60dB for a nonoiser or a dienoiser.
If you find better figures, they are most likely an illusion
It is likely that with the addition of a follower stage (implemented by Q and R in a basic reference scheme), many of the current parameters will improve significantly.
I would suggest testing it.
I would suggest testing it.
This circuit with 3 transistors will not do better than the original Elvee denoiseur with 1 transistor.
The CCS is non sense in this design.
What is the improvement by the CCS instead of a resistor as in the original denoiser ?
PSRR please ! With meaningfull asumptions about component values.
What is the improvement adding an emitter follower.
Why have you removed a resistor and a capacitor from the design of Elvee ? ( the RC at the transistor collector )
Sorry, I say it bluntly: Mods on a circuit you do not understand is plain stupidity.
The CCS is non sense in this design.
What is the improvement by the CCS instead of a resistor as in the original denoiser ?
PSRR please ! With meaningfull asumptions about component values.
What is the improvement adding an emitter follower.
Why have you removed a resistor and a capacitor from the design of Elvee ? ( the RC at the transistor collector )
Sorry, I say it bluntly: Mods on a circuit you do not understand is plain stupidity.
You are being more than blunt: you are being rude and inaccurate. You better look back at first Diego's design, which was called "dienoiser", that Elvee considers on a par with his Nonoiser design, and his Denoisator version for fixed voltage regulators. They prove that Diego does understand Elvee's design quite well and should deserve your respect.
Awaiting for answers to my questions near above ( post #3,027 ). With facts, figures, technical reasons.
Difficult to tell out of the blue, but CCS loading pushes a pole to infinity, which might be good (or bad?). It certainly has an effect(s), but I cannnot fathom all the consequences without simming or testing it extensively.
Many participants to this thread have good competences and abilities and can be relied upon if they provide good enough evidences. For the moment , I sit in the middle as I have no definitive information
Only for example:
The C8 capacitor is not shown in the suggested schematic, for graphical clarity. Presumably, it includes it and for which values would surely have to be adjusted in practice (as for the values of the other components).
For the variant with CCS (similar to post N° 3026), there are also obvious improvements (probably about 30 dB or more). In any case, the real test is mandatory.
It should be noted that the follower stage can be implemented in two possible ways, with PNP or NPN transistor. The results do not seem to change substantially. However, there may be special polarity requirements for some existing variants.
The C8 capacitor is not shown in the suggested schematic, for graphical clarity. Presumably, it includes it and for which values would surely have to be adjusted in practice (as for the values of the other components).
For the variant with CCS (similar to post N° 3026), there are also obvious improvements (probably about 30 dB or more). In any case, the real test is mandatory.
It should be noted that the follower stage can be implemented in two possible ways, with PNP or NPN transistor. The results do not seem to change substantially. However, there may be special polarity requirements for some existing variants.
Last edited:
CCS change I did not look in its effect on stability, I doubt there is any significant stability effect.
The main effect I see is changing the transitor collector load.
CCS instead of 1.8k resistor, changes the visible load from 1.8k to MegOhms.
But the actual AC load is way lower because of the dynamic collector resistance that is here some 10k and the divider at the LM317 that is 220 ohm //1.9k seen above.
( CCS or 1.8k ) // 10k // 1.9k // 220 boils to about 200 ohms.
This proves that the CCS does about nothing.
The main effect I see is changing the transitor collector load.
CCS instead of 1.8k resistor, changes the visible load from 1.8k to MegOhms.
But the actual AC load is way lower because of the dynamic collector resistance that is here some 10k and the divider at the LM317 that is 220 ohm //1.9k seen above.
( CCS or 1.8k ) // 10k // 1.9k // 220 boils to about 200 ohms.
This proves that the CCS does about nothing.
Last edited:
I see a new twist.
Discussing adding an emitter follower to the original 1 transistor design by Elvee.
That is a new discussion, previously that what adding the EF over the adding of the CCS.
OK in this new discussion the EF improves the denoising. May be the claimed 20dB above.
BUT
There are better ways to add a BJT for better PSRR, just to mention the original 2 transistors denoising by Elvee.
Discussing adding an emitter follower to the original 1 transistor design by Elvee.
That is a new discussion, previously that what adding the EF over the adding of the CCS.
OK in this new discussion the EF improves the denoising. May be the claimed 20dB above.
BUT
There are better ways to add a BJT for better PSRR, just to mention the original 2 transistors denoising by Elvee.
Tell me more about the obvious improvements
There is no worthwhile improvement, as proved in my post #3,032 above.
#3026 is your stuff, test it.
Last edited:
I was not the one who proposed the variant with CCS instead of the collector load resistor of the original circuit.
I have proposed variants around CFPs (named Dienoiser), but loaded with resistors on their collector. I don't remember raising anything with CCS.
What I did recently propose is to incorporate a voltage follower to the existing variants, and, among the latest schemes that were discussed (as I saw in a trileru post) I cited one that he presented.
I think that the follower could be incorporated into several of the options that have been presented (I don't know if all of them). That will have to be verified on a test bench.
I have proposed variants around CFPs (named Dienoiser), but loaded with resistors on their collector. I don't remember raising anything with CCS.
What I did recently propose is to incorporate a voltage follower to the existing variants, and, among the latest schemes that were discussed (as I saw in a trileru post) I cited one that he presented.
I think that the follower could be incorporated into several of the options that have been presented (I don't know if all of them). That will have to be verified on a test bench.
For the original Denoisator it does add some 4dB of PSRR which Denoisator could use at that level. It also seems to help a bit with output impedance.
And also allows for higher Vce working point for denoiser BJT. With just a resistor it seems more unstable at that point, it didn't work so good in my tests.
It also works good with DC-coupling the Denoisator to LM317, allows Vout tweaking by altering Rcb value without changing the current through the BJT.
edit: Forgot to mention that it does help a lot with lowering the noisefloor. It's always higher with just a resistor. This was clear from all of my measurements, and I measured the original Denoisator, the CCS version and Diego's Dienoiser (which also has low noisefloor). Same as the NoNoiser. With CCS Denoisator the noise floor is closer to NoNoiser than Denoisator with resistor. But I didn't rule out if the noisefloor is lower because of running higher Vce for Denoisator or if it's because of CCS, the high Vce resistor-Denoisator wasn't stable to measure.
Last edited:
The original Denoisator was always stable with LM317. There are some issues with LM337. But I am using the CCS Denoisator atm in my headphone amp and works fine both LM317 and LM337.
CCS NoNoiser is a bit harder to keep stable, output capacitor network is more important, a bit of extra inductance is needed to keep LM337 stable.
DC-coupled LM317 was also stable, I can't remember LM337 or even if I ever measured it.
The regular Denoisator or CCS Denoisator should be the biggest improvement in PSRR and noise floor and output impedance while keeping LM3x7 tempco.
CCS NoNoiser is a bit harder to keep stable, output capacitor network is more important, a bit of extra inductance is needed to keep LM337 stable.
DC-coupled LM317 was also stable, I can't remember LM337 or even if I ever measured it.
The regular Denoisator or CCS Denoisator should be the biggest improvement in PSRR and noise floor and output impedance while keeping LM3x7 tempco.
Forgot to mention, if you ever have issues with LM337 stability just remove the output cap's ground leg and add some 3cm of 1mm diameter wire between pin and its PCB pad, should make it stable. Or whatever thickness wire you have, 0.5mm at 2.5cm, something that should give you some 20-25nH.