Please read my posts 🙂 in the he schematic to the left, the connection between the Elvee/Diego network and the 220uF has been cut, and it still sims exactly the same way as when it isn't cut. This leads me to believe that simming with this LM317 model is not useful for our purposes. If I made a mistake, please tell me what.
Yes, I used the LT1085 model as well, sorry for the confusion. But it doesn't seem to work either.
Would a MOSFET capacitance multiplier followed by an LM317 not be orders of magnitude better than a De-Noisifier.
Why has nobody performed a transient response by switching a fast load change. I bet the results would look awful.
The RH117 model has no internal noise programmed, meaning it is impossible to reduce....
Yes, it does, but it cheats because it only operates in AC, and does nothing for DC accuracy, which is totally unimportant for audio anyway.
With better transistors (ZTX), you will gain even more
Say 60dB for a typical cap-mult, + 75dB for a SOTA 317 =135dB.
That's better than a plain vanilla denoiser, but comparable to a Dienoiser or a Nonoiser, without an additional power component in series wasting several volts
I did, and it remains generally well behaved thanks to its low impedance, but if the peak to peak amplitude of the transient exceeds the average DC output current, the correction circuit is overwhelmed, altering the DC output, etc.
For a regulator intended for accurate, low noise application, such a limitation is not really a problem.
Mind you a step function in the load does quite okay except that one should run it off a real input, not a dc source or inject noise onto the input. Yeah, well done whomever thought this out to add some ac only feed-back into an LM317.
No. The LM317 has much more eigennoise.
I did do a kind of transient analysis by putting a square wave on top of the DC input voltage. No problems in sight. On the load side, need to do more soldering.
Last edited:
In the asc you posted, the outputs of the regulators with and without the denoiser are labelled with the same name: out.
For LTspice, this means that they constitute a single node, ie. they are shorted.
When you unshort them by using different names, things become normal:
Attachments
With a 100ns resolution, everything looks normal: there is a visible residue for the regular 317, which is normal, and nothing is visible for the Dienoiser, which is normal too: in transient mode, the tolerance, digit number and resolution should probably need to be increased to see something.
There is no sign of instability, but the sim is not fully reliable for that: one needs to make a physical test to confirm it, which I did
There is no sign of instability, but the sim is not fully reliable for that: one needs to make a physical test to confirm it, which I did
Attachments
Thanks for the help and input. It looks much better to my eye now.
Positive and negative version attached.
The boards will go to the fab this weekend.
Attachments
I use them for a power amp frontend within a cramped case and selfmade adapter boards. Used other LM317 implementations for this before (standard and with tracking pre-reg) and would like to test this one. The amp is a NCC220 from avondale.
Very interesting!!!
A difference of more than 55 dB compared to the regulator alone (without bypass capacitor)!!!.
Could you repeat the measurement by short-circuiting the resistor between base and emitter of the BC560C?
That measurement could help to see the typical differences between the denoiser and the dienoiser.
Best regards
Last edited: