Gerhard, my question was more about the 100R shunt resistor, this value is to high if you expect something like 200mA from your current source, because in this case you get 20V across this resistor, and your "unreg" source is 20V... so you need at least 27-28V with a 100R resistor, or lower his value.
john curl said:
As I said in my last post, you are absolutely right John, about the 1K resistor (too low to set the zener current), putting it to ground now that you pointed it, is obvious! In this case the value for this resistor must be higher around 2.5K-3K (17V/3K=5.6mA) and for fig1. During all these discussions I wondered more especially about the pro's and con's of the current source transistor "direction" and trying it in an unusual way, so just like you (as my crossover design is fet's based), I forgot Ib!.
john curl said:
I am a little confused ??? Didn’t you said that you use shunt regulation after the IC series regulator and before the fet followers?
When I asked for noise values in a previous post, it was about Zener and led devices, have you (or someone here) some noise values on them? (2.2V, 4.7V and 15v zeners and some led's), I would like to find the best ratio between the zener/leds reference in matter of noise vs. Vz-If values and the emitter resistor for the current source compound. According to what you said previously, for an Ic around 200-400mA, I should use a Zener lower than 4V7 (real Zener not avalanche) or 2 leds and an emitter resistor around 10R-20R, right?
JUSTCALLAMEDAD
Voltage references noise : see post # 75 and the whole thread :
http://www.diyaudio.com/forums/showthread.php?s=&threadid=35821&perpage=10&highlight=&pagenumber=8
Results are surprising.
~~~~~~ Forr
§§§
Voltage references noise : see post # 75 and the whole thread :
http://www.diyaudio.com/forums/showthread.php?s=&threadid=35821&perpage=10&highlight=&pagenumber=8
Results are surprising.
~~~~~~ Forr
§§§
Forr, I looked at #75, but I found it confusing to me. However, I did find it useful.
Justcallmedad, I am following a pretty old data book by Motorola. It was made in the days when the military insisted that they measure these things.
It is called: 'Rectifier and Zener Diodes Data' 1984. It seems that the very first quality zener diodes had a noise curve measured on them. So, if you took a particular zener diode series, you could estimate the noise contribution.
One series is the 1N746-759 and 1N957-1N986.
On page 4-8 they show a 'typical noise density' graph.
Below about 4V, the noise is so low, that it is off the graph. However, the optimum operating current of these devices is about 20ma. 2 ma operation will severely effect the operating impedance, changing it from 10ohms to perhaps over 100 ohms. Not too good. Above about 4.5V the graph turns upward fast! At perhaps 1uV/rt Hz at 4.7V, to 1000uV/rt Hz at 10V! It peaks at about 15V to about 2500uV/rt Hz, and then DROPS to about 50uV/rt Hz at 40V or so, then starts up again and peaks at 1500uv/rt Hz at 90V. What a curve!. Is this typical for today's devices? I wouldn't chance it, and ignore this curve.
In any case, for low noise, stay below 5V or so. I should think that leds would actually be a better choice. This is where #75 could be valuable. Also, any input by Walt Jung would be useful. However, wouldn't noise add as the sqrt of the sums of the squares, rather than linearly? So, 2 identical diodes in series would have 1.4 times the noise.
I only glanced at #75, so this may be the same thing. I'll go back an look.
Justcallmedad, I am following a pretty old data book by Motorola. It was made in the days when the military insisted that they measure these things.
It is called: 'Rectifier and Zener Diodes Data' 1984. It seems that the very first quality zener diodes had a noise curve measured on them. So, if you took a particular zener diode series, you could estimate the noise contribution.
One series is the 1N746-759 and 1N957-1N986.
On page 4-8 they show a 'typical noise density' graph.
Below about 4V, the noise is so low, that it is off the graph. However, the optimum operating current of these devices is about 20ma. 2 ma operation will severely effect the operating impedance, changing it from 10ohms to perhaps over 100 ohms. Not too good. Above about 4.5V the graph turns upward fast! At perhaps 1uV/rt Hz at 4.7V, to 1000uV/rt Hz at 10V! It peaks at about 15V to about 2500uV/rt Hz, and then DROPS to about 50uV/rt Hz at 40V or so, then starts up again and peaks at 1500uv/rt Hz at 90V. What a curve!. Is this typical for today's devices? I wouldn't chance it, and ignore this curve.
In any case, for low noise, stay below 5V or so. I should think that leds would actually be a better choice. This is where #75 could be valuable. Also, any input by Walt Jung would be useful. However, wouldn't noise add as the sqrt of the sums of the squares, rather than linearly? So, 2 identical diodes in series would have 1.4 times the noise.
I only glanced at #75, so this may be the same thing. I'll go back an look.
Thank you forr for the link and John to spend time to seek this information, really thank you . Apparently between the thread on the Zener noises measures and your research, Zener are less noisy when they are fed by a "high" current (about 20mA), but what about fet's current sources at this Id? I suppose they are more noisy too? so an J203 or an SK246 feeding 2 leds would be less noisy?
. Apparently between the thread on the Zener noises measures and your research, Zener are less noisy when they are fed by a "high" current (about 20mA), but what about fet's current sources at this Id? I suppose they are more noisy too? so an J203 or an SK246 feeding 2 leds would be less noisy?
Long gate fets, like the j203 or the k246 are better current sources, because they have LESS transconductance and have generally higher output impedance. However, they are limited in current. I would think that an led or two in series, would be best.
Zeners are very susceptable to operating current, not because of noise, , but because of impedance, which reduces their effectiveness in regulating anything. Noise in zeners is facinating, because it varies so much with operating voltage.
Looking through my book further, I found that the 1N4099 thru 1N4135 series of 1/4W zener diodes gave a slightly different spec.
For example: 4.3V gave 1uV/rt Hz, 4.7V gave 1uV/rt Hz, 5.1V gave 2uV/rt Hz, 6.2V gave 5uV/rt Hz, and 6.8V and above, gave 40uV/rt Hz! This is better than common zeners, but the trend is the same: Use zeners below 5V, or else you must expect noise.
Again, I would stick to leds, but there could be a 3V zener out there that is actually quieter than 2 diodes, but I expect you would have to run it at a higher current than the leds to get the same performance.
Zeners are very susceptable to operating current, not because of noise, , but because of impedance, which reduces their effectiveness in regulating anything. Noise in zeners is facinating, because it varies so much with operating voltage.
Looking through my book further, I found that the 1N4099 thru 1N4135 series of 1/4W zener diodes gave a slightly different spec.
For example: 4.3V gave 1uV/rt Hz, 4.7V gave 1uV/rt Hz, 5.1V gave 2uV/rt Hz, 6.2V gave 5uV/rt Hz, and 6.8V and above, gave 40uV/rt Hz! This is better than common zeners, but the trend is the same: Use zeners below 5V, or else you must expect noise.
Again, I would stick to leds, but there could be a 3V zener out there that is actually quieter than 2 diodes, but I expect you would have to run it at a higher current than the leds to get the same performance.
Searching at Google "Zener noise" I find this : It could be useful, and some more tests for Christer ;-)
http://www.knoxsemiconductor.com/onlinecatalog.htm#z
Series 1N6082-6091
http://www.knoxsemiconductor.com/onlinecatalog.htm#z
Series 1N6082-6091
In fact I need 150 mA for the audio circuit and I think that a current source about 200 -250mA would be appropriate, and would allow a suitable Rds on the shunt mosfet, the problem with this circuit is to get a high current (with low noise) on the pass device (current source), with the requirement of a low noise reference (zener or led) so 3 or 4V and a “high” emitter resistor (but not to high) in my case something between 10-20R. In short I need the highest V reference together with the lowest noise to don't amplify it by a low emitter resistor value.
Justcallmedad, I think that we are on the right track. I don't know which color led is the quietest, or if the size of the led matters. Still, the effective impedance of the leds, even 2 or 3 in series is pretty good at low currents because it follows the 'diode equation'. So 2 ma should be about 13 ohms per diode. Not bad! A 3V zener might be 100 ohms or more, at 2ma. Not too good. Seems that leds could win in regulation, and probably in noise, in most cases.
john curl said:
In [1] the author spends 2 pages on Zener diode noise and shows some good and bad examples with respect to the frequency distribution of the generated noise. Having the noise at low frequencies is especially bad because 1. we are interested in audio frequencies and 2. lower frequencies are harder to filter.
A bad example is the Moto 1N941 with 44 uVrms in 500 Hz bandwidth at 10 KHz, slowly falling to 15 at 1 MHz.
The better ones are the Moto 1N935: 2uV @ 10 KHz, 20uV @20 KHz and 45 uV at 30 KHz and above.
Moto 1N821: 1.5 uV @ 10 KHz, 12 uV @ 20 KHz, 45 uV @75KHz and above, everything in 500 Hz BW.
[1] James A. Crawford: Frequency Synthesizer Design Handbook, Artech House
Good book, but not centered around audio. Having clean supplies is merely looked upon as a self-understood prerequisite.
regards, Gerhard
Real circuit measures report.
I tested today in the “real world “ the 2 current sources circuits, the schematics are reported below. I set Ic to 100 mA.
I didn’t use any capacitor, and I measure the line AC rejection of those current sources.
The fig2 rejection is about 55 dB until 100 kHz then increases progressively towards 20 dB at 4MHz.
I get 10 to 20 dB better AC line rejection with the fig2 circuit. Adding a cap from base to ground improves it of 20dB.
I used a led in lieu of the zeners on the schematics
I tested today in the “real world “ the 2 current sources circuits, the schematics are reported below. I set Ic to 100 mA.
I didn’t use any capacitor, and I measure the line AC rejection of those current sources.
The fig2 rejection is about 55 dB until 100 kHz then increases progressively towards 20 dB at 4MHz.
I get 10 to 20 dB better AC line rejection with the fig2 circuit. Adding a cap from base to ground improves it of 20dB.
An externally hosted image should be here but it was not working when we last tested it.
I used a led in lieu of the zeners on the schematics
Has someone an explanation on these important differences between these 2 circuits on AC rejection ?
I measure also led impedances:
Green: HLMP 1503 AGILENT TECHNOLOGIES
Red: HLMP-NG07 AGILENT TECHNOLOGIES
They are the best on this matter I have on hand.
If . . . . Red Green
4.5 mA 12R 22R
3.4 mA 15R 27R
2.1 mA 20R 35R
1.3 mA 31R 50R
I am not able to measure the noise, I intend to arrange a test rig for this kind of measures including fets but, I do not have too much time at the moment, and I must mainly arrange my workshop at home!!!
john curl said:
I measure also led impedances:
Green: HLMP 1503 AGILENT TECHNOLOGIES
Red: HLMP-NG07 AGILENT TECHNOLOGIES
They are the best on this matter I have on hand.
If . . . . Red Green
4.5 mA 12R 22R
3.4 mA 15R 27R
2.1 mA 20R 35R
1.3 mA 31R 50R
I am not able to measure the noise, I intend to arrange a test rig for this kind of measures including fets but, I do not have too much time at the moment, and I must mainly arrange my workshop at home!!!
Re: Real circuit measures report.
Is this a result of a real, soldered circuit or a simulation? (Nice 5 digit voltmeter!)
If it's a soldered circuit, how are the current sources implemented? Are they potential energy sources?
Does "rejection increases from 55 towards 20 dB" mean that it's getting better or worse?
One would expect that fig. 2 is a little bit better because the beta of the npn is probably a little bit better. I did not check the data sheets. But a difference of 10 to 20 dB seems too much to be explained by beta.
If you insert a cap from base to gnd in fig. 2 you get something that looks much like a cap-multiplier AC-wise. This means it probably behaves more like a const. voltage source with a constant current load (I6 par 100R) for AC rather than a CCS. That's not bad by itself, but a second stage parallel regulator (as probably intended) won't find the Hi-Z environment it works best in.
best regards, Gerhard
Justcallmedad said:
Is this a result of a real, soldered circuit or a simulation? (Nice 5 digit voltmeter!)
If it's a soldered circuit, how are the current sources implemented? Are they potential energy sources?
Does "rejection increases from 55 towards 20 dB" mean that it's getting better or worse?
One would expect that fig. 2 is a little bit better because the beta of the npn is probably a little bit better. I did not check the data sheets. But a difference of 10 to 20 dB seems too much to be explained by beta.
If you insert a cap from base to gnd in fig. 2 you get something that looks much like a cap-multiplier AC-wise. This means it probably behaves more like a const. voltage source with a constant current load (I6 par 100R) for AC rather than a CCS. That's not bad by itself, but a second stage parallel regulator (as probably intended) won't find the Hi-Z environment it works best in.
best regards, Gerhard
Dear Mr Curl,
I am one of the audiophiles that buys Borbely's parts - yes, they are costly, but if they are the best, it is worth the money. Has anyone said that being an audiophile is a cheap hobby?
Could you let me know what FETs you are talking about?
With best regards,
Sigurd Ruschkowsk
I am one of the audiophiles that buys Borbely's parts - yes, they are costly, but if they are the best, it is worth the money. Has anyone said that being an audiophile is a cheap hobby?
Could you let me know what FETs you are talking about?
With best regards,
Sigurd Ruschkowsk
john curl said:
Gerhard, these results are not from a simulation, but from a solderless breadboard.
Rejection worsens from 55 dB towards only 20 dB at 4 Mhz.
Differences can’t be explain only by Beta mismatches, but rather as you pointed out by the Hi-Z Low-Z from one way to the other… that probably causes emitter degeneration in one case and not in the other or perhaps also, by a better channel separation (Base/Collector) in fig2 case.
I have tried to lower the shunt resistor to 10 Ohms but in this case my function generator doesn’t work properly (Zout 50 Ohms) I will carry on these tests later with a modified power amplifier for a good bandwidth until a few MHz.(Gv=1)
In any case it could be interesting to see which is the most suitable direction, knowing that we get line and load disturbances. The goal is to know which direction is best appropriate taking into account the different characteristics of these 2 disturbances.
A more complete data of various “Zeners- Leds” impedances:
It's obvious now that with Christer's noise measurements and these impedance values, red leds and to a lesser extent green ones are the best choice for references in audio, for lower values 1 or 2 1N4148 do the job.
Rejection worsens from 55 dB towards only 20 dB at 4 Mhz.
Differences can’t be explain only by Beta mismatches, but rather as you pointed out by the Hi-Z Low-Z from one way to the other… that probably causes emitter degeneration in one case and not in the other or perhaps also, by a better channel separation (Base/Collector) in fig2 case.
I have tried to lower the shunt resistor to 10 Ohms but in this case my function generator doesn’t work properly (Zout 50 Ohms) I will carry on these tests later with a modified power amplifier for a good bandwidth until a few MHz.(Gv=1)
In any case it could be interesting to see which is the most suitable direction, knowing that we get line and load disturbances. The goal is to know which direction is best appropriate taking into account the different characteristics of these 2 disturbances.
A more complete data of various “Zeners- Leds” impedances:
An externally hosted image should be here but it was not working when we last tested it.
It's obvious now that with Christer's noise measurements and these impedance values, red leds and to a lesser extent green ones are the best choice for references in audio, for lower values 1 or 2 1N4148 do the job.
Hi, John; H, Guys;,
Just have a oo on the Mark Alexender amplifier patent... Zener diode can be done withn reverse polarized bipolar transistor.... Just have a look on this wonderful amp... Doidn't try this ? But what where some guys just doing excepting criticize others, eh Forrr ?
It is onlya reverse polarized base-emitter junction... I measured 7.5 to 8.7 spread..... Please Forr and some other don't h anu message .. I won't reply... Just have some more study...
Jbaudiophile
Just have a oo on the Mark Alexender amplifier patent... Zener diode can be done withn reverse polarized bipolar transistor.... Just have a look on this wonderful amp... Doidn't try this ? But what where some guys just doing excepting criticize others, eh Forrr ?
It is onlya reverse polarized base-emitter junction... I measured 7.5 to 8.7 spread..... Please Forr and some other don't h anu message .. I won't reply... Just have some more study...
Jbaudiophile
- Status
- Not open for further replies.