Gerhard, you have made an important point. It is possible that the 220uf cap on the current source could throw things off. I think that a J203 fet could replace the 1K resistor. It is true that 2ma is a bit low, but with too much current would give too high a Vp and make for a higher voltage initial supply. It seems to me that the shunt fet is really an open loop follower, so the output impedance of the regulator is really 1/gm of the fet itself. The 220uf cap in that circuit is probably OK.
Dear Sir,
very interesting post of yours.
So the classic LM317 is not that bad.
Could you give me some hints about the topology you use with it?
I am very interested in +/- 30V dual supply.
Thank you so much.
And of course my sincere congratulations for your excellent creations !
Kind regards,
beppe
poobah said:
Poobah, My intent is to simultaneously have a low noise (RC filter after the TL431) and an open loop configuration, as I can get a quite low output impedance (2 Ohm or less) without feedback. I prefer this than the drawbacks of a feedback shunt regulation.
gerhard said:
Gerhard, you are absolutly right about the 220uF capacitor (ground to base in both figures).
Yes it ¡¥s more by practice... however concerning the fig2, I still need to think of it more in depth, the zener noise is shunted on the base side by the 220uF capacitor and on the other side (emitter-resistor) shunted to ground by the mosfet¡K and this is why I said that I preferred the fig2 circuit.
john curl said:
Yes John it's possible that this capacitor "throw things off", as I said to Gerhard, but for fig2 circuit, I have to think of it some more, in any case as soon as I would have more time I would make measurements on a real circuit, starting with my "old" test circuit.
Yes of course the 1k resistor must be replaced by a fet, the 2 mA value was more to use for leds but I have no led spice models to represent them, of course 5mA is best suited for zeners.
For the moment these circuits are rather intended to study the "direction" of the BJT current source related to the preregulator and the shunt device.
”…However, with the 10 ohm resistor you have dropped your potential noise contribution by the current source about 10/.125 or about 80 times.That 10 ohm resistor pads the transconductance of the transistor that much! Perhaps, you knew that, Richard, but I hope that I can point this out to others, who might be looking at these circuits.¨
About the preceding post, I suppose you speak about Gm=1/Re and Re=26/Ic, I knew this, so to optimize the noise contribution of the current source, John could you give us some noise values for say 2.2V, 4.7V and 15v zeners as well as for led's please
or even some hints on noise/impedance vs. current for these devices (I have read the Vishay datasheet on zeners but there is nothing on noise, and either for leds, where the impedance is not stated (I measured it and according to the colours and current, values goes from 40 to 80 Ohms, 1mA < If < 5mA).
Richard, another factor is the beta of the C45 transistor. I have been working with fets for so long, that I forgot to put that in, however it would seem that you need at least 2 ma to bias the transistor to its operating current. That would leave no current for the zener! It would seem that you need a low current j113 fet, let's say 10ma or so for a bias fet. I doubt that you could get away with 1K, in any case. Am I missing something else, Gerhard?
OF COURSE JOHN!!!
Don’t worry, for some simulations I used variable voltage sources, by the way here some simulations results:
Line rejection with and without the 220uF cap:
Response to line transients for the fig1 and fig2 circuits without the 220uF cap:
Response to line transients for fig1 and fig2 circuit with the 220uF cap:
In circuit 1 this capacitor worsens things, in circuit 2 results are better (! not same scale as above.
Load response for 0-100 mA steps:
In circuit 1 this capacitor worsens things, in circuit 2 results are a little better…
I don’t trust so much sims, but to get an idea how things work it's ok.
No noise values, John, please.
Don’t worry, for some simulations I used variable voltage sources, by the way here some simulations results:
Line rejection with and without the 220uF cap:
An externally hosted image should be here but it was not working when we last tested it.
Response to line transients for the fig1 and fig2 circuits without the 220uF cap:
An externally hosted image should be here but it was not working when we last tested it.
Response to line transients for fig1 and fig2 circuit with the 220uF cap:
In circuit 1 this capacitor worsens things, in circuit 2 results are better (! not same scale as above.
An externally hosted image should be here but it was not working when we last tested it.
Load response for 0-100 mA steps:
In circuit 1 this capacitor worsens things, in circuit 2 results are a little better…
An externally hosted image should be here but it was not working when we last tested it.
I don’t trust so much sims, but to get an idea how things work it's ok.
No noise values, John, please.
Hi, Upupa
Well said ! I am in a full agreement with you., Upupa... . (Rare, eh ? LOL !).
Some guys have little more answers to our questions (than others claim to...). Hence your valuable question... They are not really gurus... Are they ? But they are much more experienced as we are, as amateurs...
And here lies the problem : on the forum full beginners are crossing end of professional life engineers, soldering technicians cross full chief engineers, digital world crosses analog. world, and so on... Where is the problem ?
It is around serious and possible terms... 60 years old guys claiming that they can still hear 16 Khz at 0 dB are still astonishing per my data (ever with the Loudness button pushed on !!!). Or both !!!! Eh, Elso ?... (Both Fletcher & Munson and Baxendall would be wrong at the same time)... But LOL !
No chance that those people really wnat to exchange data here... They are just NOT in their own universe... And far too prone to shout one out.. Only PRIVATE messages around make the forum go seemingly round...
That's why I am slowly but consistently ripping off this forum... And it is just my 2 cents...
But, anyway, thanks for valuable inputs (as it made my time taking English lessons not fully unuseful !).
Jbaudiophile
Well said ! I am in a full agreement with you., Upupa... . (Rare, eh ? LOL !).
Some guys have little more answers to our questions (than others claim to...). Hence your valuable question... They are not really gurus... Are they ? But they are much more experienced as we are, as amateurs...
And here lies the problem : on the forum full beginners are crossing end of professional life engineers, soldering technicians cross full chief engineers, digital world crosses analog. world, and so on... Where is the problem ?
It is around serious and possible terms... 60 years old guys claiming that they can still hear 16 Khz at 0 dB are still astonishing per my data (ever with the Loudness button pushed on !!!). Or both !!!! Eh, Elso ?... (Both Fletcher & Munson and Baxendall would be wrong at the same time)... But LOL !
No chance that those people really wnat to exchange data here... They are just NOT in their own universe... And far too prone to shout one out.. Only PRIVATE messages around make the forum go seemingly round...
That's why I am slowly but consistently ripping off this forum... And it is just my 2 cents...
But, anyway, thanks for valuable inputs (as it made my time taking English lessons not fully unuseful !).
Jbaudiophile
Listen!
JB.
I never said I could hear till 16kHz. This month I will be 59.........I am happy with 10kHz.......
The issue is what can you hear and which circuit or circuit detail affect this.
All these power supply proposals...... Build them folks, LISTEN!
After all all these efforts are done to please the ear!
JB.
I never said I could hear till 16kHz. This month I will be 59.........I am happy with 10kHz.......
The issue is what can you hear and which circuit or circuit detail affect this.
All these power supply proposals...... Build them folks, LISTEN!
After all all these efforts are done to please the ear!
Richard, what I am most concerned with is the 1K resistor. Looking more carefully at your complete schematic, I see that you have only a 2V differential across the 1K resistor and this implies that all the current from the resistor goes into the transistor to supply the base current. Is the 3V device even turned on? Anyway, why not put the 1K resistor to ground. Wouldn't that work OK?
As far as design philosophy is concerned, I like to think about new ways of making circuits. Sometimes, I see a new way of looking at a circuit on this website. I find that interesting. However, why don't I like normal IC regulators that much? Well, they are noisy, have bad transient response, and fall apart at high frequencies. I prefer, myself, to use an active IC regulator in series with a simple high speed fet cap follower. This gives me the DC stability of the feedback regulator with the high speed response and low noise of the fet follower. That is my approach. I know that some circuits may be less critical of power supply effects, but I do know that complementary cascode circuits are VERY sensitive to power supplies. That I why I put so much effort into power supplies.
As far as design philosophy is concerned, I like to think about new ways of making circuits. Sometimes, I see a new way of looking at a circuit on this website. I find that interesting. However, why don't I like normal IC regulators that much? Well, they are noisy, have bad transient response, and fall apart at high frequencies. I prefer, myself, to use an active IC regulator in series with a simple high speed fet cap follower. This gives me the DC stability of the feedback regulator with the high speed response and low noise of the fet follower. That is my approach. I know that some circuits may be less critical of power supply effects, but I do know that complementary cascode circuits are VERY sensitive to power supplies. That I why I put so much effort into power supplies.
Here we have a quick simulation of fig1. The current source provides voltage gain for the input ripple when loaded with 100 Ohm.
I used components only that are in the distributed library of LTspice.
There was no 3V zener so I used a 6V one, I doubled the 10R.
The Input voltage is 20V with 1% ripple, or 200 mV. The output AC voltage is 1.8V. This results in a voltage gain of 9. The only node that's pure is the base.
Removing the capacitor (just set it to 1p ) makes it much better, but still not great. The Z diode + 1K is too easy a path for the noise. John's FET current source is definitely required.
store this in ccs_fig1.asc:
------
Version 4
SHEET 1 880 680
WIRE -224 160 -224 80
WIRE -224 352 -224 240
WIRE -80 0 -224 0
WIRE -80 160 -80 0
WIRE -48 0 -80 0
WIRE -32 160 -80 160
WIRE 80 0 32 0
WIRE 128 160 32 160
WIRE 128 160 128 64
WIRE 144 160 128 160
WIRE 304 0 176 0
WIRE 304 160 224 160
WIRE 304 160 304 0
WIRE 432 0 304 0
WIRE 432 80 432 0
WIRE 432 352 -224 352
WIRE 432 352 432 160
SYMBOL res 48 -16 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R1
SYMATTR Value 20
SYMBOL res 240 144 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R2
SYMATTR Value 1000
SYMBOL pnp 176 64 M270
SYMATTR InstName Q1
SYMATTR Value 2N2907
SYMBOL zener 32 144 R90
WINDOW 0 0 32 VBottom 0
WINDOW 3 32 32 VTop 0
SYMATTR InstName D1
SYMATTR Value BZX84C6V2L
SYMATTR Description Diode
SYMATTR Type diode
SYMBOL voltage -224 -16 R0
WINDOW 0 44 53 Left 0
WINDOW 3 -26 -48 Left 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V1
SYMATTR Value SINE(0 0.1 1000 0 0 0 10)
SYMBOL voltage -224 144 R0
WINDOW 0 42 53 Left 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V2
SYMATTR Value 20
SYMBOL res 416 64 R0
SYMATTR InstName R3
SYMATTR Value 10
------
This should be enough to recreate the simulation. Just get the free LTspice from the Linear Technology web site.
Regarding noise:
Has anybody here measured the 1/f noise corner frequencies of power MOSFETs? I fear they might come as bad surprise. JFETs should be OK.
BTW I find it kind of unfair to say that current sources amplify their noise. If we do our best to construct a thing that behaves like a large resistor, we should be ready to accept that it's voltage noise is like a large resistor...
I speculate for symmetry reasons that the same holds true for active voltage sources and current noise?
regards, Gerhard
I used components only that are in the distributed library of LTspice.
There was no 3V zener so I used a 6V one, I doubled the 10R.
The Input voltage is 20V with 1% ripple, or 200 mV. The output AC voltage is 1.8V. This results in a voltage gain of 9. The only node that's pure is the base.
Removing the capacitor (just set it to 1p ) makes it much better, but still not great. The Z diode + 1K is too easy a path for the noise. John's FET current source is definitely required.
store this in ccs_fig1.asc:
------
Version 4
SHEET 1 880 680
WIRE -224 160 -224 80
WIRE -224 352 -224 240
WIRE -80 0 -224 0
WIRE -80 160 -80 0
WIRE -48 0 -80 0
WIRE -32 160 -80 160
WIRE 80 0 32 0
WIRE 128 160 32 160
WIRE 128 160 128 64
WIRE 144 160 128 160
WIRE 304 0 176 0
WIRE 304 160 224 160
WIRE 304 160 304 0
WIRE 432 0 304 0
WIRE 432 80 432 0
WIRE 432 352 -224 352
WIRE 432 352 432 160
SYMBOL res 48 -16 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R1
SYMATTR Value 20
SYMBOL res 240 144 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R2
SYMATTR Value 1000
SYMBOL pnp 176 64 M270
SYMATTR InstName Q1
SYMATTR Value 2N2907
SYMBOL zener 32 144 R90
WINDOW 0 0 32 VBottom 0
WINDOW 3 32 32 VTop 0
SYMATTR InstName D1
SYMATTR Value BZX84C6V2L
SYMATTR Description Diode
SYMATTR Type diode
SYMBOL voltage -224 -16 R0
WINDOW 0 44 53 Left 0
WINDOW 3 -26 -48 Left 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V1
SYMATTR Value SINE(0 0.1 1000 0 0 0 10)
SYMBOL voltage -224 144 R0
WINDOW 0 42 53 Left 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V2
SYMATTR Value 20
SYMBOL res 416 64 R0
SYMATTR InstName R3
SYMATTR Value 10
------
This should be enough to recreate the simulation. Just get the free LTspice from the Linear Technology web site.
Regarding noise:
Has anybody here measured the 1/f noise corner frequencies of power MOSFETs? I fear they might come as bad surprise. JFETs should be OK.
BTW I find it kind of unfair to say that current sources amplify their noise. If we do our best to construct a thing that behaves like a large resistor, we should be ready to accept that it's voltage noise is like a large resistor...
I speculate for symmetry reasons that the same holds true for active voltage sources and current noise?
regards, Gerhard
Attachments
What can I say? I know how to make cheap stuff as well as expensive designs. If I am to make an all-out design, then I will go all-out on the regulators as well. When I help Parasound with a cheap-to-make design, I am much more forgiving.
I am trying to make power supplies that will effectively be as good as a quality battery. Many here wonder why I don't use batteries, then. Well, batteries are unreliable, and I know this from experience. Also, I like to run my designs with higher current than many other designs, when I can. This is partially because the parts that I can get are really designed for higher current operation, and I like to operate closer to their optimum operating point, and I usually find that the distortion drops both in amount and in harmonic order, when I run at higher currents. This makes batteries impractical, except for really crazy people, (like some of my friends). Now, what can't I do as well as batteries? Usually, it is isolation from the power line, which I really think is pretty dirty.
I have measured my own power lines, and found them to have garbage in relatively high amounts, not only at low freqencies, but at mid frequencies 5K and above, as well as RF.
Why do any less?
I am trying to make power supplies that will effectively be as good as a quality battery. Many here wonder why I don't use batteries, then. Well, batteries are unreliable, and I know this from experience. Also, I like to run my designs with higher current than many other designs, when I can. This is partially because the parts that I can get are really designed for higher current operation, and I like to operate closer to their optimum operating point, and I usually find that the distortion drops both in amount and in harmonic order, when I run at higher currents. This makes batteries impractical, except for really crazy people, (like some of my friends). Now, what can't I do as well as batteries? Usually, it is isolation from the power line, which I really think is pretty dirty.
I have measured my own power lines, and found them to have garbage in relatively high amounts, not only at low freqencies, but at mid frequencies 5K and above, as well as RF.
Why do any less?
Now it's much better. The ripple gain is gone. Altough we have a 6.2V Zener, the voltage across the diode is only 4V. That means that the 1K resistor cannot deliver the current that is required.
Now if we insert a >6 mA FET CCS instead of the 1K we probably have something we can use.
BTW running Zeners with less than enough current makes them generate even more noise.
Electronics folklore factoid:
Professor Zener once sued semiconductor companies that they should not call high voltage (>6V) diodes Zener diodes, because they didn't use 'his' effect. They settled on calling them Z-diodes for the shape of their characteristic curve. But at the end of the day, Prof. Zener lost...
Gerhard
Now if we insert a >6 mA FET CCS instead of the 1K we probably have something we can use.
BTW running Zeners with less than enough current makes them generate even more noise.
Electronics folklore factoid:
Professor Zener once sued semiconductor companies that they should not call high voltage (>6V) diodes Zener diodes, because they didn't use 'his' effect. They settled on calling them Z-diodes for the shape of their characteristic curve. But at the end of the day, Prof. Zener lost...
Gerhard
Attachments
Phoebe Cates/James Bongiorno
I was a hippy, not a yuppy hahaha
I have irritated James Bongiorno, but I like his amps, have a Son of Ampzilla now.
Some of his suggestions are very good........I think James Bongiorno has acute hearing.
This month I hope to be 59.... With His will.
It's strange as James' face reminds me of my dad, though my dad did not wear that crazy coloured suits.
I bought my Ampzilla II in 1974. I was blown away and sold my SAE IV DM. Before that I had a Crown D-150. I tried many expensive bolides in the years following, never was satisfied.
http://www.allmovieportal.com/cgi-bin/frame/frame.cgi?phoebecates==http://www.phoebe-cates.com/
jacco vermeulen said:
I was a hippy, not a yuppy hahaha
I have irritated James Bongiorno, but I like his amps, have a Son of Ampzilla now.
Some of his suggestions are very good........I think James Bongiorno has acute hearing.
This month I hope to be 59.... With His will.
It's strange as James' face reminds me of my dad, though my dad did not wear that crazy coloured suits.
I bought my Ampzilla II in 1974. I was blown away and sold my SAE IV DM. Before that I had a Crown D-150. I tried many expensive bolides in the years following, never was satisfied.
http://www.allmovieportal.com/cgi-bin/frame/frame.cgi?phoebecates==http://www.phoebe-cates.com/
Justcallmedad said:
Caesar: Divide et impera (in English Divide and conquer IIRC)
For the moment I'm only interested in getting the current source right. The current that really flows is only 149 mA (light blue, right scale) because the Zener is current starved at 4V.
It's fun to play with the load resistor. In the first circuit, 1K makes the ripple clip. Or try 10 Ohm. Spend a virtual heat sink for the tiny 2N2907
Gerhard
- Status
- Not open for further replies.