Yes, I understand the balancing act here. The reactive Q's (aka phase slopes) are low enough at these freq's that we can even consider doing it - go up another decade or two and forget about it.KSTR said:
Well my winding-on-resistor-body idea won't work, the L is too high, about 3.3uH, so I'd need much finer wire than the 26 ga. and don't have anything on hand. 330nH||10R is no easy feat, either! So I think I'll stay with the simpler configuration for now, which is still way better than anything else to date and simple enough for anyone to build. I'm pretty sure okapi is waiting for the design to stabilize and plans to lay out a board for it, which would be great.
Yes, and if we raise the Rout design goal up to 50mR-ish, we could fall back and use the lowly LM317, which many would like to see happen, if only because they are dirt cheap and almost everyone has some of them lying around not being used...
And the question of "if it is suitable" is one we have yet to look at carefully. But we should be able to do so soon, once we have linear-phase reg systems for each Rout design level to compare.
the half wave cycles that alternately flow in the +ve and -ve feeds are effectively step functions.jbau said:
These half waves are reputed to contain frequencies many decades higher than the frequency of the basic sinewave that is split into the two half waves.
AndrewT said:
Perhaps, but that's an issue for input filtering, they are not loads on the regulator.
"Reputed" is the key word here. The spectrum of a step function is well known and there's nothing mysterious about it that makes it impossible to deal with through normal filtering. I've never seen an example of it feeding through the reg from input to output. I'm not saying it doesn't exist, I've just never seen it. Can you measure it and show one?
But the basic point is, it's not a load on the reg.
This all pretty much depends on what our goals are. For improving existing design one would essentially need to find the place/circuit/IC which is most prone to supply effects and design the reg(s) so that at this place performance is optimized (flattest PSU impedance way up to the GBW, not only to 20kHz -- and don't forget the load current rectification into the rails from class-AB circuits, too). I mean, what's the benefit of a flat 20mR Zout at the reg when trace inductances, series R/L/ferrites and local decoupling change it all again at the point of concern?jbau said:Yes, and if we raise the Rout design goal up to 50mR-ish, we could fall back and use the lowly LM317, which many would like to see happen, if only because they are dirt cheap and almost everyone has some of them lying around not being used...
And the question of "if it is suitable" is one we have yet to look at carefully. But we should be able to do so soon, once we have linear-phase reg systems for each Rout design level to compare.
BTW I still think (dare I say I know, because I did) that the lowly LM317 can be succesfully optimized in the same basic way (but with handling of some of its special aspects).
- Klaus
My take is Andrew is talking about those rectified load current drawn from the reg, not the the supply inputs. This plus the notion that the supply impedance shall be flat up to and beyond the circuit's highest inner bandwidth makes the (large-signal) step response pretty relevant, IHMO. Measured at the point of load, not the reg output, again.jbau said:
- Klaus
KSTR said:
In theory, yes. But in reality, there is nothing we can do with the regulator circuitry to create low, linear Z/phase into the MHz region. That's a power distribution matter and is most effectively (i.e. must be) done locally at the devices.
Oh now Klaus, your brilliant and insightful mind is taking things WAY out of context here. I'm not doing anything to these regs that make them more susceptible to this rectification, nor does it change the device-local impedance considerations either. Those considerations exist no matter what regulators you use. So don't make it sound like this design compromises them.
The benefit is markedly better sound, which you would surely hear if you built one and listened to it. The Sony DATs I'm modding are not even close to being ideal layouts power-wise: the PSU is on a separate board, connected to the main circuit board by about 16" of 18 ga wire for rails and ground. Trace length from power entry point to the 8 opamps it feeds is another 6" to 8". Even with that long powerline path, I am not exaggerating when I say the sound and holography improvement is stunning. I expect the same to be true for optimizing the ±5V regs for the A/D converter analog supply. The DAC I think is questionable whether it will benefit or not.
Oh, I don't doubt it at all, but I doubt you can pull off a mean Z value in the 20mR range with a 317. Something in the 40-50-ish range is definitely possible. But the sensitivity to Vin-Vout remains with this device, making a pre-regulator almost a practical necessity.
I've never seen this phenomena in any measurements, is it under high-current draw conditions?
See my reply above. I'm not dealing with Z/phase at the device - yet. I'm dealing with power generation and power distribution as separate issues, and haven't dealt with the second one yet. My approach isn't the only way to skin this cat, but so far, so good.
I hope you'll continue to give your valued input as we continue. But please try to restrain yourself, Klaus - let's deal with one issue at a time!
(My ex used to harp on me for doing this same thing... now that we've been divorced for 18 years, I guess it's time to say, she was right.) 😀
even though the existence of these non sinusoidal currents demands are made by the load, it seems we have a Thomas among us.jbau said:
I have never seen them, I have never measured them, I have never knowingly heard them.
I believe the many reputable sources that tell me they exist.
Any ClassAB stage when moving from the ClassA output current limit into the ClassB region changes the current demand in the supply rails from sinusoidal to half wave. The further beyond the ClassA limit the worse the step function becomes due the the nature of the curve which inherently becomes steeper the closer one approaches the zero current crossover point.
In the ClassA region the loading is sinusoidal if the reproduced signal is sinusoidal. In the ClassB region the rails must supply step function current.
For a designer to say "my circuit does not need to be tested for susceptability for this type of step function load" because "
is complete balderdash.
Having had one's attention drawn to the possibility that some form of step function testing may be useful in determining your own regulator circuit's behaviour to real world current loading is doing Jbau's designer reputation no good whatsoever.
AndrewT said:
So that should motivate you to verify it in your own measurements.
The only rubbish here is your swift and harsh judgement. I didn't say anything of the sort. I said two things:
1. I've never measured this phenomena, which is true.
2. I doubt anything I've done makes this situation worse.
So stop posturing and post the results of your work that show us otherwise. Better yet, since you understand it so well, show us how this design can be improved. I'm open to being proven wrong - are you?
Who cares? I'm not here to promote my reputation. I'm here to share a learning process and get results. If you have something positive to contribute, I welcome it. But this sort of personal attack is juvenile and completely unnecessary behavior.
Gawd, the egos here are intense.
Is that reply another way of saying "I don't intend testing my regulator for stability using some form of step function", because Jbau does not believe us?
Andrew, I sent you a private email, but apparently you would prefer to hash this out on the forum. I don't think it's appropriate to do so.
If you would use your eyes instead of testosterone, you would have seen that Okapi has already volunteered to do the step test on it. Patience.
About the "Believe us."
Sounds like faith-based engineering to me. Believe what? I don't have to believe anything. I asked you to describe the phenomena, which you did, thanks. I asked, have YOU measured it, and how did you do it. But you haven't. If you want to condemn me for something that you haven't done, then rail on to your heart's content. Dig a hole and then crawl into it.
And who is "us" ? The gang? And you're the spokesman for them? Give me a break. I'll take the other side of that trade any day.
"Men go crazy in congregations. We only get better one by one."
You may not like the way I go about things, but the fact is, this project has largely achieved what it set out to. If you want to condemn it, go ahead. It will only drive the more civil-minded people away from this forum.
Frankly, I think your attitude stinks. It's so easy in these public forums to sit and throw darts at people. And with lax moderation, there are no consequences for doing it, either. I'm largely left to moderate my own thread. Apparently you're not mature enough yet to realize, your action says more about you than it does about whoever you're attacking. Since you're so concerned about reputation, I suggest there are better ways to build yours than to do it at the expense of others'.
You're treating this like a competition, and it's not. So I'll say again. If you want to make a positive contribution, I welcome it. Otherwise, bugger off.
AndrewT said:
If you would use your eyes instead of testosterone, you would have seen that Okapi has already volunteered to do the step test on it. Patience.
About the "Believe us."
Sounds like faith-based engineering to me. Believe what? I don't have to believe anything. I asked you to describe the phenomena, which you did, thanks. I asked, have YOU measured it, and how did you do it. But you haven't. If you want to condemn me for something that you haven't done, then rail on to your heart's content. Dig a hole and then crawl into it.
And who is "us" ? The gang? And you're the spokesman for them? Give me a break. I'll take the other side of that trade any day.
"Men go crazy in congregations. We only get better one by one."
You may not like the way I go about things, but the fact is, this project has largely achieved what it set out to. If you want to condemn it, go ahead. It will only drive the more civil-minded people away from this forum.
Frankly, I think your attitude stinks. It's so easy in these public forums to sit and throw darts at people. And with lax moderation, there are no consequences for doing it, either. I'm largely left to moderate my own thread. Apparently you're not mature enough yet to realize, your action says more about you than it does about whoever you're attacking. Since you're so concerned about reputation, I suggest there are better ways to build yours than to do it at the expense of others'.
You're treating this like a competition, and it's not. So I'll say again. If you want to make a positive contribution, I welcome it. Otherwise, bugger off.
Andrew,
If we have nothing to contribute we better keep our mouth shut.
I haven’t seen a thread of yours in which you share your findings with test results of something but have seen you criticizing others work many times. Please lighten yourself up and support others work with positive inputs.
If we have nothing to contribute we better keep our mouth shut.
I haven’t seen a thread of yours in which you share your findings with test results of something but have seen you criticizing others work many times. Please lighten yourself up and support others work with positive inputs.
Hi,
Probably off topic given the wise persons present here:
Walt Disney - whatever one may think of him - had a very powerful three-step strategy for development. First, he would elaborate his dream. Second, he would make a ‘Bill of Materials’ to construct his dream. Third, he would criticise the outcome of his first two steps. If he was not satisfied, Disney would pass simply (partly) the routine again. The third step of the critic appeared to be very useful as homework before presenting his projects, especially for financial stakeholders. Those who cross from dream directly to critic might never kick-off.
The nice experience with the diyaudio forum is that the Disney strategy almost always works without saying and contributions are made according to competences. Occasionally, an accident happens and wise man proceed professionally.
A summer thought, Arjen.
Probably off topic given the wise persons present here:
Walt Disney - whatever one may think of him - had a very powerful three-step strategy for development. First, he would elaborate his dream. Second, he would make a ‘Bill of Materials’ to construct his dream. Third, he would criticise the outcome of his first two steps. If he was not satisfied, Disney would pass simply (partly) the routine again. The third step of the critic appeared to be very useful as homework before presenting his projects, especially for financial stakeholders. Those who cross from dream directly to critic might never kick-off.
The nice experience with the diyaudio forum is that the Disney strategy almost always works without saying and contributions are made according to competences. Occasionally, an accident happens and wise man proceed professionally.
A summer thought, Arjen.
is there consensus on what circuit i should test
i am setting up to do the transient load test.
based on this comment:
i am assuming i should test this configuration:
http://www.diyaudio.com/forums/attachment.php?s=&postid=1891710&stamp=1248907284
is this correct?
what load would you like me to test?
i don't have a 560uF on hand - would it be a problem if i paralleled a 220uF and 330uF instead?
i am willing to test other configurations but it will have to wait until i return from halifax in about 10 days.
cheers
i am setting up to do the transient load test.
based on this comment:
jbau said:
i am assuming i should test this configuration:
http://www.diyaudio.com/forums/attachment.php?s=&postid=1891710&stamp=1248907284
is this correct?
what load would you like me to test?
i don't have a 560uF on hand - would it be a problem if i paralleled a 220uF and 330uF instead?
i am willing to test other configurations but it will have to wait until i return from halifax in about 10 days.
cheers
Re: is there consensus on what circuit i should test
Hi okapi, that's the one.
Close enough for rock'n'roll... do you have the 20mR resistors?
I'm jealous! Have a great trip!
okapi said:
Hi okapi, that's the one.
330 ohms on the output for a quiescent load. Someone else will have to specify the step load they want to look at.
Close enough for rock'n'roll... do you have the 20mR resistors?
I'm jealous! Have a great trip!
Re: Re: is there consensus on what circuit i should test
thanks jbau. first vacation in 3 years and boy do i need it.
i do have several smt 0.01 ohm resistors - i'll put a couple in series.
anyone want to speak up about a load they would like to see tested - Andrew, Infinia?
jbau said:
thanks jbau. first vacation in 3 years and boy do i need it.
i do have several smt 0.01 ohm resistors - i'll put a couple in series.
anyone want to speak up about a load they would like to see tested - Andrew, Infinia?
I think I said in my other post ... no load and loaded. loaded could be quiescent and/or 1/2 rated I for full load. Duty cycle to match heatsinking.
infinia said:
The 45mA quiescent current is an important part of the design. It is not designed to be operated with no load.
A Circuit Board Submission
I've been following this thread with great interest, since I was debating on using the 3 term regs or going all out with "super regs" for my preamp project. I decided to follow jbau's lead and try the 3 terminal reg approach first, due to its simplicity and cost. I'd been working on a power supply board design when this all started, so I've finished it up using the latest configuration as given in this thread. The layout was done in ExpressPCB, so I don't have gerber files available to share (sorry). The board is configured to allow for either bipolar supplies, dual positive supplies, or a single positive supply. Let me know what you think, but be kind - this is my first attempt.
I've been following this thread with great interest, since I was debating on using the 3 term regs or going all out with "super regs" for my preamp project. I decided to follow jbau's lead and try the 3 terminal reg approach first, due to its simplicity and cost. I'd been working on a power supply board design when this all started, so I've finished it up using the latest configuration as given in this thread. The layout was done in ExpressPCB, so I don't have gerber files available to share (sorry). The board is configured to allow for either bipolar supplies, dual positive supplies, or a single positive supply. Let me know what you think, but be kind - this is my first attempt.
Attachments
- Status
- Not open for further replies.