BRAVO this is a well thought out method that moves from a real world problem and move toward a solution that like the text book claims works. By using a measured reduction approach of real problem it move to a great solution . Far too often the dismissive approach of ignoring real world problems outside the area of limited interest has been taken by " knowledgeable engineers" .
So then... are there any threads here where best practices of PSU layout have been noted and compiled, whether for PCB, perfboard, or veroboard? Has what you learned been distillable to some extent into some basic patterns and guidelines for aspiring (not expert, no scope) DIYers?
Does that imply that one might get better results with fewer problems by outputting from one rail and then using a voltage divider circuit afterwards? N.Pass has such on a simple design like the B1 Rev2 buffer.
Then one has to get into the mode of testing JFETs (or paying a premium to buy matched sets), right? ...JFET-based PSU's do seem to be the darling of the DIYAudio community these days, probably with good reason.
John Walton's article in Linear Audio magazine shows test equipment measured data, AND listening test results, when connecting 13 different voltage regulators to the same piece of audio equipment. Here's a link. The regulator that won the listening test (highest overall score), contained no JFETs at all. The second place winner used a JFET-input opamp IC, but no discrete JFETs.
The SuperTeddy regulator and the Belleson regulator do use a (discrete) JFET source follower driving a BJT emitter follower (what some people call a "FETlington"), but those regulators finished poorly in the listening tests.
The SuperTeddy regulator and the Belleson regulator do use a (discrete) JFET source follower driving a BJT emitter follower (what some people call a "FETlington"), but those regulators finished poorly in the listening tests.
Last edited:
Not sure what you're referring to here, but apparently it wasn't the same article that I purchased and read.
The Jung/Didden regulator easily won the listening test and used either a JFET input AD825 op-amp or the bipolar AD797.
Second place in the listening tests was the Burson regulators, which do NOT use any ICs at all.
Not sure that I would put much emphasis on that "panel of audio enthusiasts from the New Jersey Audio Society" when it comes to the listening tests though, since everyone "hears" things differently.
Some ignorant questions as I read back through this for the tenth time:
- I’m trying to get my head round the use of current sensing resistors and how that is configured here. Is the 20mR rBuffer a simple 2-terminal job in line with the +Vout leg as the schematic at #248 seems to indicate? The primer I found at Mouser shows application where the resistor is placed across Vout and Ground, not inline with Vout. Also there is a bit of discussion about 4-terminal packages, though if in use here, I don’t see how connected.
- When John Bau spoke about finding a small voltage drop for the LM337 gave best results, say around 3v, I assume he was talking about choosing a transformer with secondaries of low enough voltage so that after the diode bridge drop, the voltage is only 3v above the desired Vout from the reg? For the LT1085 he preferred a 6v drop, which seems a bit awkward if the pos leg was diff from the neg leg. Or was he meaning to control the voltage drop in some other way?
- The schematic at #248 shows a 560R cap, 100nF cap, & 10uF film cap. The 560R is obviously an electrolytic, but I can’t find him saying what the 100nF cap was: lytic, film, tantalum, ceramic, other. Earlier in the experiments he tried several different values and types, but never says what type worked best the time by the time he got here.
Bruce,
a rectified and smoothed supply has an average voltage that when measured by an average reading DMM might indicate 17.3Vdc
When the current draw is very low the ripple on that supply will also be very low, I'll guess for the sake of this example that Vripple = 20mVpp.
If you look at the waveform of that rectified and smoothed DC supply the high points of the ripple are at 17.3Vdc+(50% of 20mVpp ripple) = 17.31Vpk
The low points of the waveform are at 17.3-(50% of 20mVpp ripple) = 17.29Vpk
This range of voltage is presented to the regulator outputting 13Vdc.
The average voltage drop across the regulator is 17.3-13 = 4.3Vdrop This is well above the datasheet recommendation for regulation of most (maybe all) 3pin regulators.
The worst Vdrop is 17.29-13 = 4.29Vdrop, when the voltage waveform is at lowest point of the ripple. The average reading DMM reads 17.3Vdc and reads 4.3Vdrop.
The worst case is so close to the easily measured average that you and everyone else is wondering why I am wasting my time explaining this.
Now consider what happens when a high current passes through the regulator.
The average level drops and the Vripple increases.
The maximum probably only drops a few dozen millivolts, lets say down to 17.22Vpk
The Vripple is now 340mVpp. That leaves the worst case smoothed voltage at 16.88Vpk and the average as measured by the DMM will be ~17.05
The regulator Vdrop is now 16.88-13 = 3.88Vdrop when the waveform is at it's lowest.
Note that the difference between the low current and high current Vdrops is 4.29-3.88 = 0.41V
This example confirms that a regulator requiring at least 1.5Volts is well inside tolerances.
But it does show that when getting close to minimum recommended Vdrop that one should consider the worst case.
You will often find that the AC voltage into the rectifier should be approximately equal to the required regulated output DC voltage, i.e for a 13Vdc supply one should use a 13Vac transformer. But we would buy a 12Vac transformer. Now you really must look at the Vdrop to see how close the regulator is getting to drop out.
When Vripple becomes high that's when a DMM reading becomes misleading.
Another factor that is rarely explained in 3pin reg datasheets: As Vdrop changes from a high value to a lower value the performance of the regulator deteriorates. This can affect the performance of the load.
Last edited:
the 100nF is usually a supply rail decoupling capacitor. One usually uses a very low inductance X7R ceramic for this duty. But do check the worst case voltage, many X7R are rated @ 50V
So what you are saying is that yes, essentially the Vdrop is the amount of voltage the regulator has to reduce (Bridge rectifier Vout -> Regulator circuit Vout). Ripple can affect this by a few millivolts but shouldn't be a real factor unless working with too little breathing room. Right?
I'm not sure why VACin often roughly equals VDCout from the rectifier, when Schottky rectifiers might have 0.6vx2 Vfwd drop and Hexfreds 1.2-1.4x2 Vfwd frop in a bridge rectifier setup. That would imply that if VACin to the bridge is 17Vac, VDCout from the bridge will be 14.2-15.8 Vdc for a basic bridge rectifier from a quad of the diodes mentioned and operating at typical temperatures. Or is my ignorance showing?
Curiously, though the LM337 datasheet wants at least a 2.5v diff Vin->Vout, JBau's tests seemed to show it worked best staying fairly close to that minimum margin... along with his other configs.
You could always connect an emitter follower series pass transistor upstream of the LM317, and use a potentiometer to adjust its base voltage. This makes its output voltage (at the emitter) adjustable too. You can adjust the input voltage to the LM317! Connect a big whomper electrolytic capacitor to series pass transistor's base and voila, you've got a "capacitance multiplier" circuit, which filters out some of the ripple, presenting a cleaner DC signal to the LM317 input. Best of all, you can dial the potentiometer to whatever setting you like, and thereby, dial the LM317's (Vin to Vout) voltage difference to whatever setting you like. Maybe YOU will prefer the sound when the LM317's (Vin - Vout) = 5 volts. Or maybe YOU will prefer (Vin - Vout) = 1.5 volts? Try it and find out!
If you select the KSB596Y / KSD526Y transistors for your PNP/NPN emitter followers, their guaranteed minimum value of beta is 120, so base current will be comfortably small. Circuit design will not be terrifyingly difficult.
If you select the KSB596Y / KSD526Y transistors for your PNP/NPN emitter followers, their guaranteed minimum value of beta is 120, so base current will be comfortably small. Circuit design will not be terrifyingly difficult.
This is an old thread, and what I'm thinking of won't apply to the OP because he was wanting 3-terminal replacement regulators in an already-made device with a plus-and-minus supply, but...(if the following was mentioned in the thread, or anywhere else on diyaudio, I missed it).
It occurs to me that the idea of the usual positive-with-respect-to-ground and negative-with-respect-to-ground regulators are just old fashioned, and are related solely to the older style power transformers with a single center-tapped winding.
Modern transformers in this type of service are now (almost?) univerally made with two identical but unconnected secondaries (I saw such a transformer decades ago, though I don't know when this replaced center-tapped). If you really want the old circuit, you can connect them in series with proper phasing to get the same old power supply with the bridge AC terminals connected to the outer windings and the "center tap" as the ground.
What I've done is use the dual secondaries to make two separate and identical supplies (this is a bench supply for op-amp based circuits), each with its own bridge, cap(s) and LM317 regulator set for 15V. These are brought out to four banana/universal jacks so they can be run separately, or the + of one tied to the - of the other to make a standard +/- 15V supply (a fifth banana jack is chassis ground, connected to the green ground wire from the IEC connector/power cable - like other bench supplies, it's up to the person using the supply to decide whether to connect the chassis ground to the powered circuit).
The advantages outweigh any disadvantage I can think of (only the need for a second bridge, but they're cheap enough - the circuit I mention below already uses two bridges!). You only have to design one regulator, using whatever parts and design provide lower noise and/or best performance in other parameters (I have the impression positive regulators are less noisy than their equivalent negative counterparts, though I haven't looked).
Ironically, I see a dual-secondary transformer supply in post #301, each winding with its own bridge rectifier and reservoir capacitor, and then the positive of one and the negative of the other are connected together through a ground and then these go into the same old-fashioned separate positive and negative regulator circuits.
It occurs to me that the idea of the usual positive-with-respect-to-ground and negative-with-respect-to-ground regulators are just old fashioned, and are related solely to the older style power transformers with a single center-tapped winding.
Modern transformers in this type of service are now (almost?) univerally made with two identical but unconnected secondaries (I saw such a transformer decades ago, though I don't know when this replaced center-tapped). If you really want the old circuit, you can connect them in series with proper phasing to get the same old power supply with the bridge AC terminals connected to the outer windings and the "center tap" as the ground.
What I've done is use the dual secondaries to make two separate and identical supplies (this is a bench supply for op-amp based circuits), each with its own bridge, cap(s) and LM317 regulator set for 15V. These are brought out to four banana/universal jacks so they can be run separately, or the + of one tied to the - of the other to make a standard +/- 15V supply (a fifth banana jack is chassis ground, connected to the green ground wire from the IEC connector/power cable - like other bench supplies, it's up to the person using the supply to decide whether to connect the chassis ground to the powered circuit).
The advantages outweigh any disadvantage I can think of (only the need for a second bridge, but they're cheap enough - the circuit I mention below already uses two bridges!). You only have to design one regulator, using whatever parts and design provide lower noise and/or best performance in other parameters (I have the impression positive regulators are less noisy than their equivalent negative counterparts, though I haven't looked).
Ironically, I see a dual-secondary transformer supply in post #301, each winding with its own bridge rectifier and reservoir capacitor, and then the positive of one and the negative of the other are connected together through a ground and then these go into the same old-fashioned separate positive and negative regulator circuits.
What you're referring to is exactly what I was trying to figure out how to connect here:
http://www.diyaudio.com/forums/power-supplies/293758-lt3062-based-psu-7.html#post5104091
I've always used CT style transformers for dual supplies.
The whole dual secondaries transformer thing is new to me, but I see a lot of people using them on this site.
Clearly, there must be some advantage(s) to using them over CT transformers.
for a dual polarity regulated supply there can be an advantage to using two separate windings dedicated to their own PSU+regulator.
The advantage is that some positive regulators work better than negative regulators.
For higher currents there are few negative regulators, leaving the dual positives as the sensible option.
However, when it comes to unregulated supplies, that advantage disappears and there is little if any difference between Dual Secondaries and Centre Tapped Secondaries.
There is one area where Dual may have an advantage: simpler solutions to grounding that avoids Hum and Buzz due to incorrect wiring. But take away the incorrect wiring and this apparent advantage disappears.
The advantage is that some positive regulators work better than negative regulators.
For higher currents there are few negative regulators, leaving the dual positives as the sensible option.
However, when it comes to unregulated supplies, that advantage disappears and there is little if any difference between Dual Secondaries and Centre Tapped Secondaries.
There is one area where Dual may have an advantage: simpler solutions to grounding that avoids Hum and Buzz due to incorrect wiring. But take away the incorrect wiring and this apparent advantage disappears.
Those ideas are all well and good but miss the intent and result of this thread, optimising specific three pin regulator pairs (LT1085/LM337) for best results through measurement and experimentation (& listening): balanced impedance, low & flat/linear impedance, use of current sensing resistors, optimal voltage drop, optimal base load, best cap configs, etc.
Sure it's a little old school compared to the popular designs for enthusiasts around here, but there are still a lot of current threads on 3-pin reg design implementations, and the efforts of JBau & okapi & KSTR seem to have provided exceptional results, not to mention some testing guides that might be helpful for other applications.
Sure it's a little old school compared to the popular designs for enthusiasts around here, but there are still a lot of current threads on 3-pin reg design implementations, and the efforts of JBau & okapi & KSTR seem to have provided exceptional results, not to mention some testing guides that might be helpful for other applications.
Nelson Pass also prefers to use a transformer with two isolated secondaries, plus two complete rectifier bridges, to build ±V supplies. Here are his remarks from three years ago: link
And then after you do that, why not use the most modern, highest performance voltage regulator ICs on the market? The LM317 is more than 35 years old; but the TPS7A47 is shiny and new. It has much lower noise and much higher bandwidth and much greater line rejection than the 317. Why stubbornly cling to your grandfather's obsolete components?
And then after you do that, why not use the most modern, highest performance voltage regulator ICs on the market? The LM317 is more than 35 years old; but the TPS7A47 is shiny and new. It has much lower noise and much higher bandwidth and much greater line rejection than the 317. Why stubbornly cling to your grandfather's obsolete components?
Thanks for posting that link Mark. I'd seen it before and had wanted to link to it in other discussions, but couldn't find it again! On the TPS7A47 I just had a quick look at the datasheet, and my answer (to why would you cling to your grandfathers obsolete components) would be simplicity (completely ignoring the quest for best performance).
Tony.
(completely ignoring the quest for best performance).Tony.
May well be. I'm not trying to argue for the superiority of one reg. I was merely pointing out that others have recognised the value of his work in finding other parameters as actually more significant than absolute noise, namely Z-phase & impedance output curves. The work is throughout; one of the later chart results was on post 242. Or this note on post 311:
LM723 is even older but performs better than most of any regulators out there in noise department. Other than that, TPS7A47 and relative 'modern' parts have such a packaging that makes more sense in mass production assembly lines for mobile products rather than DIY. Additionally, if anyone interest on showing behavior of these "modern" parts on audio domain in terms of impedance and phase linearity (which must be our main subject on this thread), I see no reason to stick with the better performer despite unfriendly packaging issues. But situation here is a bit different.
This is not fashion. This is science. For about more than 40 years, semiconductor IC technology and its principals didn't evolve something "better" or more "modern" in nowadays. It is just same thing. If some smart guys designed such a part that serves the needs of current implementations 40 years ago on some degree, we should think twice before despising them as "grandfather's obsolete components". That just doesn't make any sense.
Last edited:
I was curious as to where this led jbau since he fell silent. I did some digging and found the answer https://patents.google.com/patent/US20120001702A1/en
Tony.
Tony.
Agreed.
In the power supply unit of Jeff Rowland's Coherence ONE is also a transformer with two isolated secondary windings in use - go to post #53 (second PDF) under
http://www.diyaudio.com/forums/soli...d-coherence-1-schematic-modules-wanted-6.html
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.