Syntrillium Backup Forums :: View topic - electronic experience

AndyH · Posted - Tue May 20, 2003 6:55 pm

People have so far been quite patient with my circuit questions, even though they are a step removed from direct audio equipment discussions, so I am encouraged to try again. I realize an electronics education would probably allow me to answer most of my own questions, and perhaps I am slowly learning a bit here and there, but I mostly look upon this as a desire to put up a bookshelf rather than to become a master carpenter. I just want to get various things done now rather than wait for my EE degree to materialize.

A basic voltage regulator with a three terminal regulator (e.g. LM317T(+) and LM337T(-)) is a pretty simple circuit which can be added to in various ways, such as with capacitors to improve ripple rejection and transient response. A National Semiconductor data sheet on the LM117/LM3317A/LM317 that I downloaded some time ago shows, among other applications, a slow turn-on circuit.

To possibly relieve people from a necessity to find and download the document, I will attempt to describe the basic circuit. The fundamental regulator circuit is simply two resistors in series from the output terminal to ground, with the adjustment terminal tied to the junction of the two resistors. This is all that is necessary for it to work. Adding an electrolytic or tantalum capacitor from the adjustment terminal/resistor tie point to ground in order to improve ripple rejection is one step up in complexity.

The upper resistor, R1 is a fixed value (e.g. 240 ohm) and the lower resistor, R2, is chosen to set the output voltage, which can vary from 1.2V to 37V using the LM317. Larger values of R2 give a higher output voltage.

The slow turn-on regulator adds two more parts. One is a higher value resistor (50k) in series with that ripple rejection electrolytic's plus side. I presume this will slow the capacitor's charging by the RC time constant rule.

The second additional part is a small signal transistor (2N2905) with emitter to that central tie point, collector to ground, base to the connection of the 50k resistor and the electrolytic. The data sheet circuit diagram also shows a diode between the top of the capacitor and the output terminal, probably only to safely discharge the capacitor if such is ever necessary.

My take on this, perhaps from the fantasy that I know anything at all about electronics, is that the transistor must conduct heavily when power is first applied, and reduce conduction as the capacitor is charged through the 50k resistor. This effectively reduces the value of R2 at startup, therefore the output voltage starts out low. As the electrolytic charges and the transistor reduces conduction, R2's value becomes more dominate, and the final voltage is approached.

By my calculations on the data sheet sample circuit, the 50k resistor and the transistor have little or no effect on the final circuit voltage -- the value of R2 shown calculates to the claimed output voltage.

When I nearly blew out my speakers (based on the complaints they made) by turning on the preamp after the power amp, I decided I needed some help. I found the above circuit and added it to the preamp power supply regulators. The regulator uses an LM317T for +24V and an LM337T for -24V. R2 for both sides is a 5k trim pot so I could easily balance the +/- voltages. I also have capacitors to ground on both the input and output terminals.

I chose a 2N3643 NPN for the + side and a 2N3645 PNP for the minus. There is no example for a minus supply, but this seems logical. Incase it is relevant, my electrolytics are 33uf, the "charging" resistors are 47k, and the diodes are 1N4005. These components are simply what was readily available. I did not believe the substitutions would have any major impact.

Perhaps I was wrong. The output voltages went down to about 8 volts max. I did not write anything down, but memory says the + was under 8V and the - a bit above (e.g. -8.7V). I took the transistor (but not the 47k resistor) out of the + side and the voltage is now back up where it should be (adjustable up to ~29V).

My simple multi meter tests on the removed transistor say it is still good. The power supply minus voltage (still with transistor) now goes up to (down to) -10V, which is a bit more than I remember, but maybe I wasn't paying close attention. I did not leave things running in that condition for long, to avoid any possibility of overheating by dropping the 34V input to 8V.

I figure that either the transistors are unsuited or the example circuit's 50k/25uf values are more mysterious than I believed. If different transistors are necessary, are there any suggestions for the minus side?

My tests of the transistors were simply the multi meter's diode good/bad selection on each junction, so there is the possibility that the transistors are somehow defective without being totally blown. Any easy way to tell? Alternately, the circuit is far more critical when slow turn-on components are added and this is a "don't try this at home, folks" situation for reasons beyond my face-in-the-mud understanding. Any suggestions are appreciated.

SteveG · Location: United Kingdom

I wish it was that easy to make a dual slow-ramping power supply! The clue to the problem with your power supply is in this paragraph:

AndyH · Posted - Wed May 21, 2003 12:36 pm

I think I could manage a timer, but just what gets switched by the relay, some part of the power to somewhere, or some part of the signal path? I know the power amp has a relay, but I don't have any circuit diagram of it. I might be able to figure out how some of it works, but the circuit isn't really visible. Diodes, or something, are soldered from circuit boards to output heat sinks, folding things up into a box shape, and the inside is hidden without tearing it apart.

My preamp power supply is for +/- 24V. Perhaps it is irrelevant, but the preamp is composed of discrete parts, no opamps. Possibly the circuit could be considered some kind of simple opamp, in principle. Since there are largish filter capacitors in the power supply (6800uf after the rectifiers, 470uf after the regulators), aren't they going to charge at an uneven rate even without the slow turn-on transistors, and won't this effect the audio portion of the circuit similarly?

I understand the idea that the two halves of the power (+/-) are unlikely to come up at the same rate, but what does that mean operationally? That is, what is the consequence, of not coming up at the same rate? If there were opamp chips in the preamp, would they be damaged by this momentary unbalance, or would they just make some other big noise that would obviate any possible benefit of slow turn on?

This modified power supply was not attached to the preamp for my tests. The only load was a 4.6k resistor across the output of each side to provide the regulator's minimum load requirement, and power-on LEDs . Therefore, whatever it is that limits its output to around 8V (with the slow turn on transistors in place) has nothing to do with the audio circuit being unbalanced by unequally raising voltages.

The only logical idea I can come up with is that the transistor(s) continue to conduct significantly, thus effectively maintaining R2 at a much lower value final than intended. If that is the problem, why does it happen? Is it some result of there being both a + and - supply, rather than just the + supply of the data sheet example? Even if this kind of slow turn-on circuit isn't suitable for my preamp, I would like to be able to make it work, or understand why it does not.

I have a regulator circuit from an old magazine, intended for an audio preamp (both + and - supply). Power flows through a power transistor in an emitter follower setup. A 24V zener on its base is the actual voltage regulator . The zener controls the voltage and the transistor allows the circuit to deliver adequate current to the preamp.

There is a 2000uf capacitor in parallel with the zener, and a 1.8k resistor to charge the capacitor (and supply the bias current). This seems to be similar to the LM317 data sheet example. The article says that, in addition to its low noise and good regulation, the circuit "eliminates all turn-on thump."

I never tried this, suspecting the three terminal regulator version would actually be better, at least from the viewpoint of delivering solid, quiet power. The article might lie, or be sadly mistaken, but it seems to indicate a belief that the turn-on problem can be solved without relays.

SteveG · Location: United Kingdom

In a sense, it doesn't matter whether there are opamps in it or not - any abrupt change of voltage level one side of the output capacitor will make it straight through to the next stage. And essentially, this is what's happening. Yes, there are all sorts of other power supply components that will also prevent an identical rise on both the + & - sides.

The purpose of the relay is to switch the signal lines, as in 'signal relay'. There are plenty of relays around that are fine with small signals - my favourite method is to use reed relays, because they are sealed, and easy to actuate. If you want to get really posh, I suppose that you could use mercury-wetted contact relays, but I think that most engineers would regard this as completely OTT for this application!

The slow start circuit should work fine in both +ve and -ve versions, but it really is unsuitable for this application, because the preamp output state will be indeterminate when the supply voltage is below the specified levels - it could be anywhere, as you are operating outside the spec for the device! If your output is capacitively coupled, the only solution I can think of is to get the supply voltage to rise so slowly that the inevitable output swings don't make it through the coupling capacitor at any amplitude that will cause subsequent damage. To do this, you'd have to ramp both supply voltages over at least a second. And on the face of it, this circuit does this, but I think that it probably goes faster because the effective charging resistance value is much lower to start with, as the transistor starts the cycle fully on, which implies base current (which will decrease through the charging cycle).

So to try this approach, I'd up the capacitor value to 47uF or even 100uF, and see how fast the output voltage shifts on switch-on. But if I had the problem, I'd use the timer method, and not bother with ramping the supplies.

LM317s are fairly quiet regulators - but the quietest ones are always made of discrete components as a rule.

AndyH · Posted - Wed May 21, 2003 4:18 pm

Thanks, but no suggestions about the original question? -- why is the slow turn on set-up only going up to about 8V? Take those transistors out and the power supply will adjust up to 29V (or down to -29V). The circuit I used seems to be essentially identical to the National Semiconductor datasheet application example (as specified in 5/20 post).

I do appreciate the information about how and why this kind of circuit may work poorly in this particular application, but that is all about controling the signal to the power amp, not, as far as anything I have understood, about why it just doesn't seem to work at all for its primary purpose -- slow turn-on.

SteveG · Location: United Kingdom

AndyH · Posted - Wed May 21, 2003 6:20 pm

Below is what I wrote out to post before signing on. The problem was exactly what SteveG pointed out. I could offer the excuse that I have not done anything with transistors for years and eaisly get confused by the symbols. True enough, but meassure twice, cut once, is always a good practice.

************************
I have to apologize for haranguing people when the problem, as usual, was my own stupidly. Eventually my slow slow mind worked around to the cause and, once corrected, the power supply does rise to the proper voltage -- s l o w l y.

This is independent of the problem SteveG pointed out. I understand it still might not be very satisfactory at solving the particular audio problem, but it certainly is a lot more satisfactory in other ways then when the circuit doesn't work at all.

Actually, the circuit may be functional in the preamp. SteveG suggested that turn-on over a second might do the job adequately. By my watch it takes over 40 seconds to reach 24 volts. The multi meter does not seem to react for a half second or so after I switch the mains power on, then it displays 2.0V and starts climbing. The rate slows somewhat noticeably as it approaches maximum voltage, but it starts out at going up at about one volt per second. This could sometimes be a little annoying, and perhaps it is so slow that it will cause some other problem, but maybe it is worth the effort to test it by putting the power supply back together (into the case, etc.) and seeing if the speakers complain much when I turn it on.

SteveG · Location: United Kingdom

I think that the reason it's taking 40 seconds is because the whole arrangement is a charging circuit that has to feed itself! If you took the non-transistor end of the 47k resistor to the unregulated side of the regulator, it would come up in roughly a second, because the charging current would always be there! Try it and see.

N.B.
--postscript--
I have thought some more about NatSemi's original slow-start circuit, and I am convinced, without even trying it, that it's one of their sillier ideas. However you look at it, one good transistor with a high Hfe is going to make that poor circuit hang around for ages, just as Andy has discovered. The above fix has to be a more sensible idea, but if the pre-reg voltage is anywhere near the 35v that it could be, you'd want to increase the resisitor value to 100k to stop the ramp time shortening too much. The diode is now a good idea, because it will clamp the voltage on the resistor to the output voltage + 0.7v diode drop. This shouldn't affect the regualtion at all, and will stop the cap from being over-run.

Havoc · Posted - Thu May 22, 2003 12:42 pm

alofoz · Location: Australia

Andy,

If you do as Steve rightly suggests, i.e. take the 47k resistor to the unregulated supply, the 33 uF capacitor would take roughly 1.5 sec* to charge to 22.1V for 24V at the output**. Once full output is reached, the diode turns on and the transistor turns off, thus having no further effect on the regulation.

As Steve pointed out the diode is important. Another reason is that a reverse biased base-emitter junction in a silicon transistor behaves like a zener diode with a breakdown voltage of about 8.5V (although I haven't looked up this particular transistor). This would certainly affect the regulation.

Now, the 8 or so volts you got with the transistors reversed looks suspiciously like the reverse breakdown voltage of the transistor junction and I suspect you'll get a voltage across the transistor equal to the B-E breakdown voltage plus the now forward biased B-C junction of about 0.6V. Did you measure the capacitor voltage at this stage?

* The RC time constant for 33 uF and 47 k is about 1.5 sec. On their own charging from a 35V supply it will take about 1 time constant to charge to 22.1V.

** 22.1V across C + 0.6V base-emitter voltage + 1.2V across R1

SteveG · Location: United Kingdom

alofoz · Location: Australia

No, I don't think so either. As you've shown, there seems to be adequate safety margin in this case, a Shottky diode will only improve that by about 1/2 a volt, so I think a general purpose diode is fine.

AndyH · Posted - Fri May 23, 2003 1:30 am

To clarify, the only semiconductors in the preamp proper are transistors. Output to the power amplifier is through a capacitor.

I could try putting the 'charging' resistor to the unregulated input, and I probably will do so to test it, if for no other reason, but I have more questions. Potential problems due to the two sides not ramping at the same rate have been pointed out, and this may not be alleviated by the change.

Bringing the voltage up faster by charging from the unregulated supply will obviously present a smaller time window for disaster, but will the difference between 1 second and 40 seconds means much at the electronics level? Will potential problems be reduced, increased, or unchanged by shifting the resistor? While forty seconds is quite a while in certain respects, it would probably almost never be significant in real terms. By the time I get an LP out of the jacket, or a CD into the player, the preamp would generally be ready to go.

I made a few measurements before I realized the transistors were switched, but I did not record them, so I don't remember anything except the approximate final output voltage. Now, in its proper configuration, I measure only 0.1V across the charging resistor: 22.7 at the emitter end and 22.8 at the base end. I do measure 1.248V across R1 (2V scale), but my reading of the data sheet is that the LM317T maintains ~1.25V drop across R1 regardless of circuit configuration, no? (This voltage actually fluctuates, as described three paragraphs below).

The data sheet shows how to add a capacitor and diode to the adjustment terminal to make a significant improvement in ripple rejection (fig.3, page 7). This seems desirable, especially for the phono section. The circuit change for slow turn-on immediately bothered me a little because the capacitor and diode are no longer tied directly to the adjustment terminal. Further thinking suggests to my ignorance that perhaps this addition actually eliminates the ripple rejection feature.

To get maximum benefit, the circuit may need a capacitor and protection diode on both the adjustment terminal and the transistor's base, each serving a different, independent function. This could be especially true if the charging resistor is tied to the unregulated input rather than to the adjustment terminal. Does that make any sense, or does the measured 0.1V between emitter and base mean that the capacitor is coupled closely enough?

In addition to the previously described +/- supply, the PCB has an independent + regulator circuit, intended to supply a headphone amplifier that will live in the preamp case. All three of these regulated outputs exhibit a peculiar behavior: At irregular intervals ranging from about 10 seconds down to 1/2 second, the voltage jumps up to 24.1V (or down to -24.1V). It stays there only very briefly, so that even when it happens twice per second, it seems to be at the "correct" voltage most of the time.

0.1V probably means nothing, as far as how well audio circuit works, but I would think that random changes of 0.1V would be noticeable as clicks, static, or some such unpleasantness. No?

Disconnecting the regulator from the rectifier/main filter section, I see that the unregulated DC does exactly the same thing. Since it seems to be at random intervals, I don't see how it could be an oscillation or feedback problem. I would also think that such variations on the unregulated side should not produce any symptoms on the regulated side Also, I don't understand how this could be happening at all under static conditions, using 6800uf filters. Any ideas?

I don't think it is my meter because I have a commercially made power supply that also provides +/- 24V using the LM317 and LM377 regulators. It's output does not vary this way, nor any way that I have observed. Additional mystery: I left the meter on the adjustment terminal side of the charging resistor (my home made supply), reading its 22.7V, for an extended time and I saw no variation. As soon as I moved the probe to the output terminal, the meter started recording the 0.1V jumps.

SteveG · Location: United Kingdom

The purpose of the second diode is to discharge the second capacitor if the output gets shorted - if you don't do this, it is possible to damage the regulator. It's probably fine to add the cap, as it will be ramped up from 0v slowly anyway by virtue of the existing ramping circuit. The only problem is that with the values of C you've got around this circuit anyway, and the fact that it only has a very low current draw, it isn't going to make a scrap of difference to the already very low ripple!

Your 0.1v jumps are a bit of a mystery - although since you are saying that this happens with all of the supplies, it must be a common problem. OTOH, are you reading this jump as the final digit on the meter? If you are, then just ignore the whole thing, because all it will mean is that you've reached a 0.1v threshold in the meter response, and the real jumps are nowhere near this large at all.

The only other thing I've seen that causes this is instability. And strangely enough, having large capacitors on the output side of active regulators can cause this! You will note that in the data sheet, there are no large value caps shown in this position - and there is a good reason for this. Basically, it stops the internal regulator working correctly, because the output doesn't change at the rate that the regulation expected it to (it's got to alter the charge level on a sodding great 470uF cap). So it supplies a little more 'oomph' until it does. And then the voltage jumps 0.1v... the cure is pretty obvious, I would have thought. This is one of the perils of active regulation - to work effectively, you have to trust it somewhat! Small variations in the pre-regulated supply are enough to trigger this off - it doesn't take much.

And by the time you've fixed all the potential problems with this arrangement, you might just as well have built a discrete regulated PSU from scratch, for all the grief this version is causing...

AndyH · Posted - Sat May 24, 2003 1:20 am

I'm not quite sure of the sense you intended with "you might just as well have built a discrete regulated PSU from scratch". First, while I have seen "PSU" used before, I don't know if it simply means power supply or if the U on the end has an additional meaning, as it does at the front end in UPS. I don't know if you are saying I should have made some other kind of circuit.

As far as "discrete" goes, I got the strong impression from earlier threads that a three terminal regulator should be superior to my original implementation involving two transistors and a zener. As far as "from scratch" goes, I laid out, etched, and drilled the circuit board, purchased the parts at the surplus electronics store, and soldered it all together. But perhaps I missed your point completely.

While I appreciate that this might seem like a lot of trouble to someone experienced enough to just make what he wants and expect it to work, from my perspective this project has gone very well so far. Most of this is quite new to me, and the only definite problem to date was my exchanging the two slow turn-on transistors, something easily remedied once I realized the problem.

I realized that the 0.1V jumps might actually be less; I have to use the 200V scale to measure 24V. It isn't the magnitude that necessarily concerns me, but that fact that it appears to happen very suddenly. The meter jumps .1V, then returns almost before it can be viewed. I'm concerned that this will put an audible pulse through the audio circuitry, but if you think that unlikely, I guess I can, more or less, leave it unless, and until, I find unwanted clicks and such in my music.

I don't know, but I suspect it is not the regulator that is responsible, only that the regulator is unable to cope with the transients produced elsewhere. The fluctuations exist before the regulator, as well as at its output. In fact, with the regulator PCB not connected, I get the same 0.1V jumps on the output of the rectified/filter PCB and on the transformer leads at the input of that PCB (AC here, of course). The commercially built supply that I mentioned is very similar to mine, except for the slow turn-on option, and it's output does not fluctuate.

I guess the real question is probably "why are there fluctuations on the first PCB?" Are they indications of bad capacitors, bad diodes, bad power-on LEDs? Each side has a 10Kohm bleed resistor to discharge the filters after power is turned off, and a 'power-on' LED fed via a 2.4k resistor. The only other parts are the two 6800uf filters and four 1N4005 diodes.

I understood your statement that the circuit probably won't have much ripple, but if I read you correctly, the 33uf capacitor on the transistor's base is in fact separate from and different than the adjustment terminal ripple rejection capacitor in other data sheet examples.

As I read the data sheet, there are three possible uses for capacitors. At the input they are for decoupling. Since the main filters will be in one box with the transformer and the regulator will be in another box with the audio PCBs, I used both a .01uf and a 33uf on the input terminal to cover the frequency range. There is at least one example circuit with 1000uf on the input terminal.

At the output terminal, capacitors are for improving transient rejection and reducing output impedance. I used 470uf because I was modeling my efforts after the previously mentioned commercial product. The NS data sheet says values from 1uf to 1000uf are commonly used. My guess, based on this reading (and zero experience), would have been to go to up from 470uf, since these fluctuations seem to be getting through from the input side, but perhaps that is not what the data sheet actually means by "rejection of transients."

Finally, a capacitor on the adjustment terminal is to improve ripple rejection from 65dB without to 80dB with. I guess I can experiment, once I get enough of the rest of the parts together, and see if adding a capacitor and diode make any measurable difference.

I'm not repeating all this to give lessons, but rather so make my interpretation clear, so that someone can point out my mistakes where ever they exist; thus I learn.

I am still confused as to whether it is believed that making the slow turn-on faster by feeding that sub circuit from before the regulator would
** reduce the possibility of audio circuit damage from the unbalanced ramp up,
** increase the probability of damage,
** make no difference to that probability.
Is the possibility of damage likely enough that I should not even experiment with the audio circuits attached?

SteveG · Location: United Kingdom

AndyH · Posted - Sun May 25, 2003 3:10 am

The original question of "why is the output voltage only 8 - 10V" was answered. The slow turn-on circuit was made to work by installing the transistors correctly; it is exceptionally slow. However, various considerations have been put forth to suggest that it may not be safe to use this circuit, either as is or with the suggested modification to speed the ramp up.

Unbalance voltages as ramp up occurs may (or may not) damage some audio components. Unfortunately, I see no way to test this without endangering the circuits, and since it is rather random, if I understood the comments correctly, I could never be sure of 'next time' no matter how well it behaved every previous time. Therefore, I probably should just take it out and forget it. A pity.

It is interesting, however, that the simpler but very similar circuit I described from an older magazine -- a series past transistor with a 24V zener on its base and a 2000uf cap across the zener, charged from the input side via a 1.8K resistor -- was claimed to work perfectly to eliminate turn on thump. Since it also was for a +/- preamp supply, its potential problems would seem to be identical.

An oscilloscope is no doubt a fine toy, but not in my budget. If CE's frequency analysis display won't show me the ripple, or changes in the ripple due to circuit modifications, for my purposes it does not exist. I will, however, continue to play with power supply variations; this is one of the project's main purposes.

I did see the comments that three terminal regulators might not be the best solution for a audio preamp, and that the other methods of regulation I know about might not be very good either. I appreciate that the data sheets, and any other particular source, may be biased or incorrect, but I can only try what I know and (hopefully) somewhat understand. If there are suggestions, I am interested, but I realize that can be going way beyond a audio equipment forum.

I believe I understood the logic of why larger filter capacitors may not always be best, and may be causing a problem in this setup. I can't question the principles, or anyone's experience, but that does not seem to answer the question here. The commercially produced supply I mentioned is very close to mine from the regulator onward. Its output terminal cap is also 470uf. Its adjustment terminal cap is 47uf. The same number of resistors are used, there is even a power on LED on each rail. The resistors values are a little different, but that is all the unlikeness I see.

This supply has the same 0.1V fluctuations on the input side of the regulator, but none at all on the output side. That says to me that the fluctuation isn't the meter but is a real circuit event (even if its value can't be accurately measured with my meter). Since the output caps are the same in both circuits, they can't reasonably be causing the problem in one while providing a steady voltage in the other. Unfortunately I have little in the way of other ideas about the source, but I won't actually worry about it unless it causes an audio problem.

One more question, if anyone has the energy. The data sheet seems to say that the LM317 has an internal constant voltage reference between the output and adjustment terminals. Why is it possible to get a more stable output by using an external voltage reference, as SteveG suggested? Is there some reason that the on chip reference can not be made as accurate, as stable, or what ever is necessary, as that of a single purpose package?

SteveG · Location: United Kingdom

AndyH · Posted - Sun May 25, 2003 4:26 pm

Anyone who cared to notice over the last eighteen months or so could see that I am easily confused. You have done it again.

I did read the data sheet, at least as much as much as I can make sense of with my limited background. Since so much is new to me, I sometimes notice things only after the sixth or seventh reading. I take note that you've said that data sheets can not always be trusted, but the context of your advice seems to say that this particular bit of information does follow directly from the data sheet.

SteveG · Location: United Kingdom

AndyH · Posted - Wed May 28, 2003 12:06 am

I suppose this could get to be tedious, in which case I probably won't get any more replies. I also suppose this could make me seem erratic, which is probably fitting. The next step in the preamp project, which project is somewhat long term and exploratory, will be to mount the PCBs in a case so I don't have so many clip leads running around, and that awaits a trip to the hardware store. Meanwhile I hope to do another project as quickly and simply as possible.

I have use for a headphone amplifier. It doesn't need to be anything great, it will only be used to prevent the TV viewer from imposing his taste upon everyone else. It will be fed from the line level audio out on the TV set. I choose
http://headwize2.powerpill.org/projects/showproj.php?file=cmoy2_prj.htm
because it claims to be good quality, it is a simple circuit, and it has few parts. Right now I simply have to assume I will be able to obtain the opamps, or some others reasonably suitable. The tie in with this thread is the power supply.

I intend to use a wall wart transformer that I already have. It says 16.8VAC, 1.5amp. Far more power than needed, but they don't come any cheaper. Putting a single diode on it, I measure 8.0VDC. With a 100uf cap it measures 23.0V, and with a 45ma resistive load across that, it measures 22.0V. I figure I can get at least 18V regulated, perhaps 20V.

I intend to use an LM317T again, because I don't know anything else that will do the job. I intend to rectify, regulate, then create a floating ground for a +/- supply in the only way I know how: two resistors in series between the basic + and ground, with two electrolytic in parallel with the resistors. Headphone circuit ground is the join between the two resistors.

Here is where the difficulty comes in. I chose small value capacitors for the regulator, as SteveG recommended. However, the two capacitors involved in creating the +/- supply will essentially be in parallel with the regulator output terminal capacitor, which will increase the output terminal capacitance significantly. Maybe it won't make a big difference in this probably-not-too-critical application, but if there is a better way, why should I not learn now?

I know that a resistor followed by an electrolytic is often used to isolate each stage in a multi stage amplifier. Would putting a resistor on each leg, say 10 to 100 ohms (???), before the resistors that produce the floating ground, make any positive difference, i.e. would it isolate the regulator enough to improve regulator performance?

As noted above, I simply don't know another variable, lower power, regulator. If one has a part number, one can usually find some specs on the web. Does anyone know a source that allows one to compare regulators, transistors, opamps, etc to help one choose a reasonable part number for a project?

SteveG · Location: United Kingdom

AndyH · Posted - Wed May 28, 2003 3:58 pm

Thanks again. I already had a PCB layout started with 4 diodes and a 3300uf filter. The single diode was just for initial testing, to get an idea what I might expect for voltages.

It never ends, even for a simple project. Now I have to learn about voltage references and their uses. I have found, in addition to the BUF634, some TI products: TLE2426, TPS70102, TLC272. I don't yet have any idea what I might actually find for sale around here. Mail order is often a problem for small parts because so many companies have minimum orders that far exceed my needs of the moment.

SteveG · Location: United Kingdom

The reason that he's suggested the BUF634 (apart from being told to!) is because its output drive capability is a little better than a lot of op-amps. This means that your 0v reference is more likely to stay at 0v! Its quiescent current seems to be lower as well, so if you can get a BUF634, I think that there are good reasons to stick with it.