If they are correctly decoupled locally no. What the reg sees is another matter though. Why not use force/sense wiring for that second reg? If you wire it with 4 like on the schematic on 1st page, you are done.
I am planning on no local decoupling. Local coupling introduces a whole lot of issues with resonance from LCR circuits. I have been there, done that, finally found it better without decoupling. But perhaps my previous result only applies to the treble? Perhaps I can use local decoupling on bass without affecting bass?
Remote sensing is out of question because I have too many ICs.
Remote sensing is out of question because I have too many ICs.
Its first time I hear that someone had it better without any decoupling and ICs but if that is your experience in your application, its fine by me.
As for trying local decoupling for the lower frequency section ICs only, sounds like a plan.
As for trying local decoupling for the lower frequency section ICs only, sounds like a plan.
In the active XO/EQ circuit, other than the first IC, there is no frequency higher than 2kHz fed into the IC's inputs. So I wonder why the ICs still want high frequency decoupling? RF cripping into the IC causing instability?
You don't really need to do that. Instead, save yourself some trouble and use this wiring layout http://www.diyaudio.com/forums/solid-state/139239-best-low-noise-regulator-3.html#post1759292
In the "POWER SUPPLY" rectangle, put whatever serial or parallel regulator you are working on. Please note how the load is connected and the use of a shielded cable.
This type of wiring is called "force/sense". It will conserve as much as possible the intrinsic regulator performance right up to the load port, by including the wiring in the regulator feedback loop.
From a certain level up, wiring is 100% controlling the overall regulator performance.
Syn08,
Thanks for your tip. But as mentioned before, too many ICs (18 in total?) would require 3 (+, -, gnd) x 18 = 54 wires going to the circuit board. It is not practical in my case.
Is one IC one load? So 18 loads require 54 wires.
Regards,
Bill
Thanks for your tip. But as mentioned before, too many ICs (18 in total?) would require 3 (+, -, gnd) x 18 = 54 wires going to the circuit board. It is not practical in my case.
Is one IC one load? So 18 loads require 54 wires.
Regards,
Bill
Well, then you don't need to bother much about the intrinsic regulator performance. In fact, whatever regulator you are using, the performance may not exceed those of a LM317 or equivalent.
As far as I know this is the original patent
Highly stable constant-voltage power source device
US Patent 4366432 Noro, Masao Stax Ind Ltd
Regards
James
Highly stable constant-voltage power source device
US Patent 4366432 Noro, Masao Stax Ind Ltd
Regards
James
Attachments
Very nice indeed, thanks a bunch James! I've looked quite hard for the originator of the ccs->shunt regulator but didn't come across this one. Not that anyone wouldn't have done it with tubes before, but still, I can see now how we differ from the original, for instance, he uses an explicit differential amp.
I have now come up with a new idea.
Remote sensing to the local star ground. In this case, the inductance between the PSU and load is eliminated, leaving only the local inductance and resistance on the load board only.
Actually, I now think this may be what Salas in post 61 refers to. Salas, please confirm. Thanks again.
Remote sensing to the local star ground. In this case, the inductance between the PSU and load is eliminated, leaving only the local inductance and resistance on the load board only.
Actually, I now think this may be what Salas in post 61 refers to. Salas, please confirm. Thanks again.
Last edited:
If you have grounded your ICs to a star on your filter boards plus they receive +/-V symmetrically in layout, I think that a main force/sense scheme is going to be beneficial enough compared to just 3 thick wires.
Just a thought, as I have not given it much thought, no doubt that force/sense eliminate the error of the resistance of wires. Does it eliminate the error of the inductance of wires as well???
why are the heatsinks too hot to touch @ 130mA and yet only warm @ 100mA?
There is something you have not told us!!
as far as I have seen, all opamp manufacturers specify local decoupling to ensure their opamps meet specification.
I have been down that way many times and have not found an answer yet. I have had 0.1uF, 0.01uF and other value local bypass capacitors on my active EQ/XO board for the NaO speakers designed by John K. I compared many times adding and removing bypass capacitors.
With local bypass, the higher frequency range within the audioband seem to perform better. However, it also adds some strong colourations, exaggeration or brightness to the sound and make the sound really dirty. I suspect that it is due to oscillation but have not had a scope to measure it.
Without local bypass, the sound may be less liquid at higher frequencies, but a lot more "right", "together" and even more detailed.
I tried to use LTSpice to model it. All regulators are inductive at certain high frequency range. For example, around 5uH for the LM317, so I read. Here is a relatively large L. The wire adds L and R. The local bypass is the C. The LCR circuit has to resonant at some point. With the typical output inductance of a 3-terminal reg, and the typical bypass value of 0.01-0.1uF, resonance occurs anywhere from 800kHz to 3MHz, depending on the values. I have also tried adding R in series with the PSU, making it a RC low pass with the R provides damping to the LCR. Yes the resonance peak may be damped but the sound is boring to listen to! It never sounded right to me. I came to a belief that no R should be in series to the PSU (unless it is before the regulator, of course)! When current flows through the R, the voltage developed is an ERROR. Local bypass capacitor is intended to provide a low impedance path. From the Vishay Blue box' datasheets, a 0.1uF has an ESR of 0.03R, a 0.01uF has an ESR of 0.1R. This is still higher than a regulator like the Jung Supereg, Salas v1 and v2.
I am sure I am not the first one to do it without bypass capacitors. In Jung's Supereg 4 part articles, I think in the last article he said that with the Supereg, no local bypass should be used. He even provided a MOD to his circuit so that for those who stubornly insisted on using local bypass could use it without having the resonance.
I guess that if RF does not get into the signal, and I have 1k-68p-1k-68p in front of the input of the circuit, it may have less of a problem.
I understand that most opamp manufacturer datasheets recommend using local bypass capacitors. I read them a lot very carefully and did what they said but have had no luck with audio due to the LCR resonsance. Then I thought. Perhaps most opamps are designed for a wide range of applications, not just audio. In most other high frequency applications, local bypass may be a "MUST". But when the engineers recommend using local bypass, do they refer to audio in their mind? I don't know.
So a regulator of super low impedance of high bandwidth, like the v2, goes a long way to achieve better sonic. I think this is the design philosophy behind the Jung Supereg.
The problem of without bypass, of course, is that the PSU may need to be located right at the circuit (or remote sensing). And this only ensures that there is a low impedance path to the circuit up to, say, 200kHz. Beyond that, the impedance of a regulator rises. If RF gets into the circuit, the high impedance of the PSU may be problematic to the opamp circuit.
No, I have not an answer yet. I am afraid that I have to experiment, which is a very tedious process.
I am humble to listen to what the experts say.
Last edited:
After that event I wondered if I had the current set to 1A instead of 100mA! But it was highly unlikely. It was a mystery for me and I can't think of the reason why. After the second time turn-on there was no problem.
- Status
- Not open for further replies.