Sim is not bad. Think that reality will fight you over the last 10 dB. What I like about sims is it tells you how far away you are ( usually ). Remember what you saw these last few days were real and not made impossible to copy. I doubt if anything in hi fi can be more important than the power supply. A pair of Mission 760 speakers good enough to say that. A friend sells an amp called LAB 47. It is a work of art based on a simple power op amp. It has a consideable reputation. It shouldn't be impossible to do a similar PSU + chip.
Well, that's very nice, any time you want to power a resistor.
If you use it to power anything with an audio modulated current draw, the load will modulate the rail voltage. So you get cross talk and intermodulation between different circuit areas. So the output R and C needs to be replicated everywhere the rail is used.
E.g. say your load draws 10mA at a frequency in the low audio range.Your 0R33 turns that into 3.3mV, -50dBV or 40dB worse than the problem you were trying to cure. At 1kHz, the C is filtering the load, but your circuit is still doing as much harm as good.
Be very carefull if using cascaded regulators. At best you might get less than a single one in terms of ripple ( nodal noise mixing ). That's even before one or more gets very hot. Look out tweeters if so.
At 30C up to 20 kHz REF 1V Resistor Noise. 30C seems best we might do and not critical here. One could say 0 to 60C more or less.
0R33 circa -160dB
27R Circa -140 dB
10K Circa -115dB.
100K Circa -105 dB.
At 30C up to 20 kHz REF 1V Resistor Noise. 30C seems best we might do and not critical here. One could say 0 to 60C more or less.
0R33 circa -160dB
27R Circa -140 dB
10K Circa -115dB.
100K Circa -105 dB.
Yes, designing any system where you want anything like 60dB of anything, let alone 100+dB needs great care.
I think it also pay to look at the whole unit and not get too focused about (in this case) the regulator.
Inside a typical 'black box' there are often several circuit areas which are sensitive to their local supply voltage, and other areas which are putting noise/signals on their local supplies. So long as you can isolate the opposing forces, you can make the whole box work OK.
e.g there are a few nodes in my amp which are effectively reference voltages, the rest of the DC voltages have much less influence on the output.
The best audio isn't about having the best reg for the preamp, it's about having everything work together.
I suspect you're optimising your reg for a particular situation and I'm commenting on it in a general way which may be out of context?
Yes cascading regulators can be messy, one downstream may have a response to a load impulse which puts a worse impulse on the one upstream for example.
They're not just DC they are a control loop with bandwidth, delay and damping.
They have dynamic response to both load and supply.
A LM317 used as indicated in the ON Semi datasheet with 10uF Cadj gives a 80dB typical ripple rejection. 66dB min
Cascade three of them to get 240dB typical ripple rejection. 200dB min: Non sense.
This to supply an op amp like the old good NE5532 that has 100dB typical supply voltage rejection. 80dB min
So you get 340dB typical ripple rejection. 280dB min. ROLF.
Cascade three of them to get 240dB typical ripple rejection. 200dB min: Non sense.
This to supply an op amp like the old good NE5532 that has 100dB typical supply voltage rejection. 80dB min
So you get 340dB typical ripple rejection. 280dB min. ROLF.
As ever it is a matter of knowing what you need. In the context of DIY audio the needs are very specific and the degrees of freedom to implement are many. There are no laws about how you should use a semiconductor regulator; only how well the circuit meets your needs. Ripple rejection is just one of those needs and is easily done with passives too.
Semiconductor regulators are obviously very cheap and very small compared with passive solutions. That's great. But they are not perfect replacements and care is needed.
Semiconductor regulators are obviously very cheap and very small compared with passive solutions. That's great. But they are not perfect replacements and care is needed.
haha they don't. I kinda like them myself.
They are but one tool in the tool bag of solutions...seeing the same answer repeated over and over gets old.
I tried to build one. I think I should have tried harder as it never bettered the ripple rejection of LM317. That's most likely a condemnation of me rather than the idea. What seemed quite good was a complimentary feedback pair using BD139/2SA1962. As usual what I had rather than a grand plan. One zener and filter cap and a 2 pole filtered zener supply. A CCS also that had slightly more noise. BC337-40 comes to mind as what could have been the next step. What I didn't try was a larger input cap. Whereas the 44000 uF and LM7812 seems to hit the buffers it might have worked OK with the simpler idea.
As to cascades. Many here seem not to use test gear. My warning was to them.
May be out topic, but I cannot help about.
Discrete is better.
This is a statement coming as an evidence regularly among DYIers.
Unless proven on a specific case, generally speaking that is wrong.
Integrated circuits are able of many features beyond discrete implementation. Matching of temperature and BJT characteristics for instance.
In the cases they are below discrete implementation, that is not because of some questionable commercial reasons, that is simply because:
Either the wanted performance is useless.
Or the wanted performance can be done with simple added componants and then application notes tell how to do so.
Discrete is better.
This is a statement coming as an evidence regularly among DYIers.
Unless proven on a specific case, generally speaking that is wrong.
Integrated circuits are able of many features beyond discrete implementation. Matching of temperature and BJT characteristics for instance.
In the cases they are below discrete implementation, that is not because of some questionable commercial reasons, that is simply because:
Either the wanted performance is useless.
Or the wanted performance can be done with simple added componants and then application notes tell how to do so.
Last edited:
The advantage of discrete is control of performance by means of choosing components. ICs have many advantages but realise they are constrained to silicon and virtually no internal node access and the circuit details are unknown. These are important considerations in high performance audio design. Make an informed choice. This is engineering.