| bocka |
I've developed a discret voltage stabilisator without a feedback loop. The output impedance I measured is about 0.03 Ohms from DC to 100kHz, slightly rising above that frequency because of the inductivity of the caps. Noise or something else I cannot measure: belows the limits of my measurement equipment because there's much rf around me. The static performace is not worse than a LM317/LM337 but a big improvement dynmically.
The 0.47R resistors in the rectifier lines together with the 100nF capacitors forms a low pass filter witch effectivly filters the recovery noise generated by the HEXFRED diodes. Another CRC filter lowers the ripple from the unregulated DC. Cut off frequency is about 25Hz so the 100 Hz ripple and yet more the harmonics are filtered in a simple but effective way.
Output voltage is about 27V, slightly arising with temperature. Voltage range can modified quite easy, just change the ZDiode and the current limiting resistors for the LEDs. For 15V just use a 16V Zener and decrease the 4k7 resistors to 2k2. The current for the LED should be about 8ma. The 0.47R resistors are 2W types, do not use any metall-oxide resistors here wire wound or metal foil is impotant to use. Other resistors are metal film. Caps are Panasonic FC-Type (150µF) and TPSH (6800µ). 100nF caps are WIMA MKP4. LED are 3mm red ones.
Output voltage :
Va = Vz - 0.6V
Vz is the Zener Voltage
The soundstage is quite good in my preamp, warm and good colored. No harsh sound or something else. And dynamically very good...
 |
|
|
| Bricolo |
| how did you calculate or measure the output impedance? |
|
|
| bocka |
| Feed a signal into the output and calculate the ratio between signal (driving) level and measured level. This is a simple resistor divider. Or pulse the output signal with a dynamically load and look to the ripple. With a FFT you can calculate the impedance. |
|
|
| Bricolo |
I simulated the reg with simmetrix.
I run a simulation with 0.2A load, and one with 1A load
then, I calculater R=delta(U)/delta(I)
I found +- 0.1R
Does someone know how to make a current source that vary it's drawn current during the simulation? (to calculate more rapidly the resistance, but also the ripple when the load vary)
bocka, I was looking for a tutorial about discrete voltage regulators. Yours measures very good, can you explain me how it works, why you designed it this way... ? |
|
|
| bocka |
My simulation shows about 0.015 Ohms (@0.1A), but that was not the design goal...
Some years ago I experimented with shunt regs (no TL431, discrete circuit) very good sounding for a preamp but getting awfully hot. So I tried to make an improved reg, better said stabilisator because I do not feed back anything here.
First it's necessary to get a stable reference. Nothing complex, a zener will do this job very well. I tried to use a resistor to set up the correct current but with some calculations I could see, that the regulation factor is quite bad, only 1/50 refered to the input signal because the intrinsic resistance of the zener is much too high. This will give a bad rejection of the 100Hz ripple. Next i tried to filter out tis voltage with a RC-Filter in the signal path of the zener. Good at high frequencies but bad in the low frequency region. Of course I can use a large cap but I won't. Seems not a good solution.
So let's make the resistor in the series with the zener higher, much higer, because the ratio between this resistnace and the zener resistance is the regulation factor. This can be done with a CCS, a simple diode (LED in this case), transistor and a resistor from emitter to the input voltage. This forms a 5 ma CCS - quick and simple. :idea:
To get rid of the zener noise I use a simple RC-Filter, 150µF are laying around 220 Ohms I need to adjust the current through the zener, thats my filter. With some experience I expect that the cut-off frequency should quite low, low enough to filter nealy any noise out of the zener voltage.
Typically anyone would now take a multi emitter follower as current amplifier. Well, in the middle of the 90'th I made amps with CFB output stages because these have a lower output resistance compared to the standard emitter follower. Seems to be good in this stabilisator too. Another advantage is that ther is only one junction between the base and the output so the output voltage is less dependant from temperature changes. It is important to drive the first stage in a CFP - the 2SA1209 and 2SC2911 transistors - with quite a large current, this minimises the output resistance. If any change in the output current has no effect of modulating the collector current of the first transtistor of a CFP then the output voltage would be independant from the output current. In this way the output resistance is... 0!
But this is not the reality. Of course output current changes are reflected to the basis of the second Transistor Ib = 1/ß * Ic. So a transistor with a high ß is needed. The larger the ratio between the current in the first transistor and modulation of this current by output current is, the better this circuit will work.
This is why maybe the differences come from: This is not a 1A regulator, it delivers about 300 ma, above this current ß of the 2SA1930/2SC5171 goes down, at 1A it's only the half @0,1A and this causes a higher output impedance at 1A. Additonally ß is higher at high temeratures and my Transistors get quite hot @3W dissipation power. All in all this results in a higher output impedance of the simulation than that I measure. 3 times higher seems to be right in this way.
Bonne nuit ;) |
|
|
| Jocko Homo |
Bocka:
Have you been taking apart commercial products designed by me? [joke]
Sure looks that way.
For those of you who like to run simulations: look at output z vs frequency. That may answer some of the questions that you guys have.
Jocko |
|
|
| Bricolo |
I understand the voltage reference (led+2sa1209+zpd27+220r+4k7r), that's a classical voltage reference.
OK for the RC filter on the zener, even if a C seems to measure better on my sims.
But why is the 2sc2911's base connected between R and C and not directly at the zener? Is this the CFB? (regulating the output voltage according to the ripple that is filtered on the voltage reference?)
I'm also trying to understand the output stage.
As said before, the 2SC2911 seems to be controlled by the reference-ripple, and outputs current according to this.
This output current creates a voltage drop across the 33R, and this voltage controls the 2SA1930's current.
Am I right?
BTW, do you have some tutorials to teach me the basics about discrete regultaors, so I could be able to design mine myself?
Jocko: do you know how to use simmetrix? I'm still trying to find how to do mathematical operations on the results (even the basic + - * / ). For the momment I can only measure I and U.
So if you know how to measure output Z vs Frequency, please tell me
Thanks
Alex |
|
|
| ALW |
| quote: | | Jocko: do you know how to use simmetrix? |
Don't mention sim's to Jocko - he'll get excited ;)
In fact I bet his lightning fast soldering iron hand would have this circuit built faster than most of you can find the models...
The o/p Z you've simulated at present seems to be a DC spec, which is almost irrelevant for audio, try adding an AC current source in Simetrix as a load for the reg, then run an AC sim.
Set the current source to 1mA, then using the add trace function (From Graphs and Data, Add Curve) in Simetrix plot: -
abs(1000*Vout), where Vout is the vector of the voltage at the reg o/p.
You'll get a nice trace showing the output impedance, directly displayed in Ohms!.
Andy.
P.S. if you attach the file here, I'll show you how! |
|
|
| janneman |
| quote: | Originally posted by Bricolo
[snip]But why is the 2sc2911's base connected between R and C and not directly at the zener? Is this the CFB? (regulating the output voltage according to the ripple that is filtered on the voltage reference?)[snip]Thanks
Alex |
Alex, look at a power amp with CFB output stages. This is just one half of it.
Bocka, nice circuit. Hadn't seen a CFB in this app. Looks like you really optimised it. Maybe you can get further by carefull selection of output transistors with sustained-beta.
Jan Didden |
|
|
| Bricolo |
I've attached my simmetrix file. But I used the transistors I have in the evaluation version's database, the input voltage isn't also the same.
Janneman: did you recieve my email? |
|
|
| bocka |
jocko:
| quote: | | Have you been taking apart commercial products designed by me? [joke] |
Of course I'm always looking at comercial designs and posts it as my own...
Yes I've seen many commercial designs and the more I see the less I like. There are only a few good designs with good ideas. Most build that stuff with a diff input pair, a VAS stage and output stage sometimes with emitter follower, sometimes a CFP pair aor a MOSFET output stage. I think, about 90% are build in that way. Typical OpAmp design.
So it's important to use my own brain. The last years I made some designs together with my local dealer. I mesured and build this circuits, he listened. Takes a lot of time but it was very useful. And never these designs get into commercial products.
Nice to see that I've stolen your ideas ;)
bricolo:
I don't think there are any books available which describes transistor based stabilisators. The circuit I've posted is my own design, of course similar to the basic stabilisator. I'm only using these basic circuits and modify them a bit. Hmm, try to use the Tietze/Schenck Halbleiter Schaltungstechnik. It seems that this book is also available in English, its called Electronic Circuit Design.
This is called DIY. So I post simple circuits. Feel free to build it. Measure as much as you can. Modify the values. And so on. This will feed into success and a lot of acknowledge. I started this way nearly 30 years ago. Of course a EE degree will help (and I have), but it's not nesessary. Basical mathematics are enough. And if something blows up... It doesn't matter there are only penny parts inside.
jan:
Another output Transistors? I don't know. I've tried these because they are laying around. I'm using these ones in my DIY-amps in the driver stage too. For more power a MJL1302 should work. DC output resistance of about 0.03 ohms is enough for any preamp.
And some more measurements:
I've changed the zener for some more common voltages to 2 x 6.2V. With a 220 ohm load and 18.1V input DC voltage I get an output voltage of 11.87V, with an additional 33 ohm load this voltage drops to 11.86V. In my opinion there's no need for further improvement. When I'loaded the regulator with a 6.8 ohm resistor the output voltage fall to about 8V. Too much current for these transisors. BTW a the 2SA1837/2SC4793 and 2SA1538/2SC3952 pairs also work very well.
PS: with a MJL1302 this stabi delivers 2A DC with 50mV voltage drop. And my load resistor is smoking...
PPS: Nelson - the one and only - posted that this is DIY not DIS (do it simulate). Best statement I've heard ever |
|
|
| Jocko Homo |
Put a diode in series with it. Will add some temperature compensation, and add one diode drop to offset the diode drop due to the folower.
Right......DIS......I hate simulators. Never used one, never will. Only problem, not all guys have tons of test equipment to measure how this stuff works. So, for them, it may provide some insight to how things work.
Just as long as they don't believe everything that it tells you. Yes, 30+ years of doing this will turn you into me in more ways than one.
Jocko |
|
|
| analog_sa |
Bocka
As you may have noticed, in a different thread i desribed a simple shunt regulator. I intended to follow this with several other types of regulators and compare the sound. A variation of your circuit was also on the list so after you posted i built it immediately :) Used a 2SK170 as a current source for the zeners (3x5v6) as i think it provides better performance plus it's simpler. Outputs are TIP34C.
The sound is significantly better than the 431 based circuit in every respect: freer, better flow, less electronic hash, more natural. I really like it, thanks! I'll probably also try something similar with a MOSFET pass device before i get to Jung/ALW types.
cheers
peter |
|
|
| Jocko Homo |
Don't think so. The impedance, and therefore the PSRR, of a bipolar CCS will be higher. Taking your reference from the output may help bring it back up.
Jocko |
|
|
| Charles Hansen |
| quote: | Originally posted by bocka
I've developed a discret voltage stabilisator without a feedback loop. |
Hello -
Unfortunately, we don't have a very precise terminology for different types of feedback in audio amplifiers. However, this circuit uses a CFP, or complementary feedback pair. Of the three transistors, one is a current source for the zener, and the other two form the CFP. Clearly there is feedback around the CFP, and it is in the form of a loop. It is a relatively short loop, as it is only around 2 transistors, but a loop nonetheless.
I'm sure that it sounds far superior to high feedback 3-pin regulators, but I've prefer the sound of regulators made only with emitter-followers.
Best regards,
Charles Hansen |
|
|
| bocka |
peter:
yes I've noticed your thread also before I started this. Because this is not a shunt regulator I decided to start a new one.
Nice that you like the sound. To my experience this regulator is also getting better with better parts, similar to what's changing in the amp. I'm thinking using a bipolar is better than a FET too. But a FET is a good solution when you have to get gid of the FETs you've selected for a single FET stage which did not match there.
jocko:
Yes a diode will perform some temerature compensation. But when I looked into the datasheet I noticed that I need about 10 diodes co compensate the zener. This is because ot the high voltage of the zener I use. At lower voltages this is a good method of compensating the voltage depencies.
Well, I like simulators, but it's important what do to. And you have to compare the results with the reality. Yet more: One have really to understand the theory (of semicons as well of SPICE in
itself) and the practice. I fully agree with you it's a good tool when haven't tons of test equipment. But the best test equipment anyone (i believe so) has on the left and the right side of his head.
Charles:
Yes, a CFP uses some feedback more precisely I should have written over all feedback. A simple emitter follower is also not free from feedback mechanism, because is called voltage controlled current source and any load performs that feedback mechanism.
I don't want to get a acdemical here, which is the one and only correct description of a topology. I can live with the feedback in the CFB.
2 all:
My intention was to post a circuit which sounds good to my opinion, which is simple and costs only a few bucks. Anyone should be free to test it. If it sounds better than that you've used before use this. If not just throw this CFP stabilisator away, it's not worth it.
This easy stabi is also usable for that guys which haven't much experience with electronics. It also works on my experimentation board without oscillation. It simulates good, it measures good and it sounds good (to my opinion).
I've also tried this with a MOSFET as pass transistor. Does not work well.
I'm very interested of any feedback of the sound of this and other regulators. My experience shows to make a better sound it's important to make the circuits a simple as possible without performance loss. Try to minimise the feedback paths and the amout of feedback. Feedback in itself is not bad, too much kills the sound no feedback results in a indifferent soundstage quite often. Feedback is like salt in the soup: none tastes bad and too much is terrible. :dead:
Use good quality components, compare them sonically to theses you've used before. There is no automatism that expensive components are better than the cheap ones, although some of the very expensive parts are the best. For example: Vitrohm 526 Series resistors are cheap and sounds really good. They are capless, no NOS stocks, pretty small and so on. I use them everywhere in the signal path.
Cheers |
|
|
| Jocko Homo |
He knows that it sounds good......he just does it that way to be different from me. [inside joke]
Jocko |
|
|
| Bricolo |
Nice new avatar, Jocko ;)
"Bite My Shiny Metal ***" |
|
|
| capslock |
Hi Bocka,
nice circuit! Just a few comments:
- temperature compensation of the Zener: I believe 6V8 Zeners actually have a TC of near zero
- I believe I have read that Zeners of this value also have the lowest dynamic impendance, but I don't know if stacking them will still result in lower impedance than one large-value component
- another improvement of the current source: placing an electrolytic across the LED would shunt its noise and probably give you some more line rejection
- ripple rejection would also improve if you take the reference from the output, as Jocko suggested, but I am not sure about the startup behavior of this circuit
- you like high fT, sustained beta TO-126 transistors, don't you? unfortunately, these are hard to get here (where do you buy yours?), and if you get them at all, there is a high chance they will be dubious second source devices or outright fakes. An adequate replacement in my eyes are devices with an enlarged TO92 case. Reichelt carry both A1145/C2704 and A124/C2632 at reasonable prices, and their current stock are original Toshibas and Matsushitas.
- Isn't CFP stability an issue with either strong load transients or low impedance caps on the output? I played with them in power amps for a while, and they were temperamental, to say the least.
Greetings,
Eric |
|
|
| bocka |
Hi capslock,
you can get the Toshibas from Spoerle. Maybe you can get the other transistors from heho elektronik in Germany.
6.8 - 9.1V zeners have the lowest dynamic resistance. In this way the are best. It's very difficult to obtain the same DC performance from a discrete regulator compared to an integrated one. But I can't see any problems when the voltage drifts slightly with the temperature. When using as 5V reg just put a diode in series with a 5.1V zener. Works very well without any temerature drift.
Maybe a cap across the LED helps to reduce some high frequency noise. It has no influence of the ripple rejection because the LED has a very low dynamic impedance (< 10 ohms). Noise filtering is effectivly obtained by the 220R and 150µF RC Filter after the zener.
Low impedance of the output caps is not a problem, because they also have some intrinsic resistance (at least 10 milliohms) and inductivity of about 6-30nH which is enough to make the CFP stable in an extraordonary way. The transistors I've used are at least 10 times faster than any high power bjt in an output stage. Simulation and my tests shows a very good transient resonse probably much better than you expect. The input caps becoming the problem now...
Because the output voltage is smaller than the zener voltage there's no way to feed the zener from the output unless you give the CFP some gain.
Feel free to experiment with the stabi. |
|
|
| peranders |
| quote: | Originally posted by bocka
6.8 - 9.1V zeners have the lowest dynamic resistance. In this way the are best. It's very difficult to obtain the same DC performance from a discrete regulator compared to an integrated one. But I can't see any problems when the voltage drifts slightly with the temperature. When using as 5V reg just put a diode in series with a 5.1V zener. Works very well without any temerature drift. |
You could also use a LM431 , but it's integrated..... You can pretend that the 431 is a regular zener, much better(?) than any discrete device. |
|
|
| capslock |
Exactly because the 1930/5171 are so fast I was afraid about oscillation.
One very efficient way to get low ESR in the output caps is to use 100 nF 1206 ceramic caps. There are some that can stand 63 V, at least in theory. I have had one of them explode at 48 V...
Spoerle is an idea, but then they only ship to businesses, so I'd have to talk somebody into ordering for me.
Don't know Heho. Got an URL? Are they reliable? Some others may ship you counterfeit junk.
Greetings,
Eric |
|
|
| capslock |
| I wouldn't dream of using ceramics to carry a signal, but IMO they are the best choice for decoupling a supply. My DAC is full of them! |
|
|
| bocka |
Technically right. As decoupling caps for digital parts I'm using 100nF 0603 X7R ceramic caps. They have half the inductivity as the large 1206 caps. And 0603 are much cheaper than 1206.
WIMA has also some different opinion to ceramics. ;)
You can obtain the superior performace of the ceramics only, if you are using a proper (I mean really proper!) layout. 95% of any layout seems not designed in a good (decoupling) way. Look to the return parth to nearly any PCB what I mean. If you take not care of the return path, you can use a film capacitor as well without any performance degration.
Again: sonically I prefer MKPs to ceramics, not only in the signal path also as (analog) decoupling caps |
|
|
| capslock |
Well, I got my board layout right :)
Got any theory as to why MKP would be sonically superior in a
supply decoupling application? |
|
|
| bocka |
| Unfortunately not :bawling: |
|
|
| Terry Demol |
| quote: | Originally posted by Charles Hansen
Hello -
Unfortunately, we don't have a very precise terminology for different types of feedback in audio amplifiers. However, this circuit uses a CFP, or complementary feedback pair. Of the three transistors, one is a current source for the zener, and the other two form the CFP. Clearly there is feedback around the CFP, and it is in the form of a loop. It is a relatively short loop, as it is only around 2 transistors, but a loop nonetheless.
I'm sure that it sounds far superior to high feedback 3-pin regulators, but I've prefer the sound of regulators made only with emitter-followers.
Best regards,
Charles Hansen |
Charles, Bocka, Jocko,
Good design, I have been using variations of this for years.
Charles you are right on WRT CFP FB loop. There is a loop
however small. And so it goes, you can do a 3 stage CFP
and get lower OP Z, more OLG and worse HF performance.
You could also replace 33R with current source, same effect
more OLG, worse HF performance. Seems simple works
best.
I've tried em all in follower mode and in shunt mode, 1, 2
and 3 stage versions.
Interesting experiment, once I built these for clients DAC
in shunt mode and left a shorting link for the CFP tranny
to turn it into simple 1 transistor follower. He didn't know
anything about electronics, just that it sounded different with
the link shorted (EF). Always preferred simple 1 tranny (EF)
config. So you are not alone Charles!
Terry |
|
|
| jam |
Bocka,
You said you tried a Mosfet as a pass element and you did not like it, what mosfet did you use, and did you try it in a CFP mode?
Terry and Charles,
I have found the same result as you have. I find feedback regulators and CFP configured as the pass element might sound a tad more dynamic during initial listening but you soon realise they lack musicality, smear transients and are flat in comparison.
Have any of you found that loading the regulator at it's output helps the sound?
Regards,
Jam |
|
|
| jcarr |
Jam:
>you soon realise they (feedback regulators and CFP configured as the pass element) lack musicality, smear transients and are flat in comparison.<
Many feedback regulators and CFP configurations are problematic in a multitude of ways, but at least in my findings and opinion, some are not very problematic at all (although nothing is perfect). The outcome depends on the individual circuit (including the PCB layout and physical construction), what sort of load it is being asked to drive, and if we are to include subjective criteria, the remainder of the audio system, and the tastes of the individual listener(s).
OTOH, undoubtedly you should go with configurations that don't give you any problems (measureable or subjective) and you are personally happy with.
>Have any of you found that loading the regulator at it's output helps the sound?<
Auxiliary loads should be considered in combination with the behaviour of the circuit that the regulator is driving. In other words, I would look at the total load.
hth, jonathan carr |
|
|
| jam |
Hi Jonathan,
Where have you been hiding?
I might be generalizing but I is a lot harder to get a feedback regulator right. In my experiments I have built several feedback regulators and some have been very good but seem to always go back to the pass element referenced with a zener. The zener is usually bypassed with a small film cap and fed with a current diode.
Maybe one day I will find a feedback regulator that I am happy with. I have to admit that I fave a fondness for shunt regulators but they can be pretty complex.
Regards,
Jam |
|
|
| bocka |
| quote: | | You said you tried a Mosfet as a pass element and you did not like it, what mosfet did you use, and did you try it in a CFP mode? |
I'd used an IRF9530 for the positive regulator as pass element in CFP mode. The measurement behaviour is bad.
| quote: | | The outcome depends on the individual circuit (including the PCB layout and physical construction), what sort of load it is being asked to drive, and if we are to include subjective criteria, the remainder of the audio system, and the tastes of the individual listener(s). |
And it depends on the following amp someone is using. BTW it should have been said that the regulator requires 0,6V over the 33 ohms resistors. This regulator needs at least 20ma (better 30-40ma) of output current. Loading of the output in this way might not be an option it's necessary under some circumstances. My pre - where I'm using this regulator - requires about 120ma so I've no problems with it.
To my opinion higher driving currents in the first transisor sounds better.
I've used a (power-) shunt regulator before (and it sounds best) but the behaviour of my pre was more than a heating than an amp... |
|
|
| analog_sa |
Hi Bocka
Would you like describing your shunt reg?
thanks
peter |
|
|
| bocka |
| It's a differential bjt stage first followed by a VAS stage and shunt transistor. Vref is obtained by a LED and 6.2V zener. And it dissipates a lot of power. It's not (thermically) stable for no or very low output currents (it will burn, because it needs a well defined output current range), it's not a design you can simply modify or adapt to other circuits, because you must change nearly everything. That's why I do not post it here. |
|
|
| janneman |
| quote: | Originally posted by Charles Hansen
Hello -
Unfortunately, we don't have a very precise terminology for different types of feedback in audio amplifiers. However, this circuit uses a CFP, or complementary feedback pair. Of the three transistors, one is a current source for the zener, and the other two form the CFP. Clearly there is feedback around the CFP, and it is in the form of a loop. It is a relatively short loop, as it is only around 2 transistors, but a loop nonetheless.
I'm sure that it sounds far superior to high feedback 3-pin regulators, but I've prefer the sound of regulators made only with emitter-followers.
Best regards,
Charles Hansen |
Hello Charles, nice to meet you here!
Of course you are right about the feedback loop. But I'm sure you are not surprised if I tell you I stick to my feedback (overall) regulators..?
Jan Didden |
|
|
| Charles Hansen |
| quote: | Originally posted by janneman
Hello Charles, nice to meet you here!
|
Hello Jan,
Thank you! Just in case there is any confusion, there are two Charles Hansens in audio right now. One Charles Hansen lives in Maryland and contributes many articles to AudioXpress. I am the Charles Hansen that lives in Colorado and designs for Ayre Acoustics, Inc.
Best regards,
(the real!) Charles Hansen |
|
|
| paulb |
| Just as long as we don't get either of you mixed up with Charles Manson. |
|
|
| patriz |
| I want to have 24Vout I changed the zener for 24V and then 22V and lowered 4k7 resistor but I have always 25-25,9Vout. Secondary transformer are 25V. |
|
|
|
