I have no objections whatsoever to the publications of you many observations, and I certainly find them interesting.
But not everyone here is as experienced as you and me, and they might jump to the wrong conclusions without understanding fully what lies behind your observations. I merely wish to clarify those.
Look forward to further results from you, as already mentioned. And I would be interested to see the effect of load resistance (e.g. 1k, 10k & 100k). Open circuit output connected to a probe only probably behaves like a pure capacitive load ??
Patrick
But not everyone here is as experienced as you and me, and they might jump to the wrong conclusions without understanding fully what lies behind your observations. I merely wish to clarify those.
Look forward to further results from you, as already mentioned. And I would be interested to see the effect of load resistance (e.g. 1k, 10k & 100k). Open circuit output connected to a probe only probably behaves like a pure capacitive load ??
Patrick
Okay, I have prepared more data for all to use.
I've got the B1 on a regulated +/-5V supply, the results are as follows:
Quiescent current: 14.9mA
14mV offset (input grounded)
Frequency response is within 2dB from 100kHz to 10MHz (measured with HP 654A sig. gen and Tek 2247A scope)
Frequency response graph from 1Hz to 100kHz will follow in an attached image along with graph comparing distortion of HP signal source against amplifier output.
Gain at 1kHz:
Loaded by Fluke 179 DMM: 0.990
Loaded by Fluke 179 DMM and 1000Ω resistor: 0.967
Loaded by Fluke 179 DMM and 470Ω resistor : 0.943
Loaded by Fluke 179 DMM and 100Ω resistor : 0.790
Output Impedance at 1kHz: 25.2Ω (Something weird about that, as if transconductance is very low or something's wrong. Maybe my Iq is too low and I have JFETs at the low end of the spec.) (Or I did the math wrong three times).
I did very thorough testing to determine the source of the aforementioned oscillation. The oscillation turned out to be within the B1 and not my function generator. I did some tests to see what it took to make it oscillate:
Load: 1.5m RG56 Coaxial cable, unterminated.
Input: 470Ω resistor to ground and Tek P6109 10x probe with 2247A scope.
Power supply: Regulated +/-5V bypassed each side to ground with 1.5µF poly caps within 2cm of B1 board.
Results:
With B1 direct to supply no oscillation was detected but there was peaking at 10MHz when driven with HP 654A oscillator.
With B1 V+ coupled through 1mH inductance to supply oscillation occurred at 864kHz at 1.18Vp-p.
With B1 V+ coupled through 50µH inductance to supply oscillation occurred at 6.06MHz at 6.44Vp-p.
To reduce susceptibility to oscillation 100Ω resistance was added in series with output to decouple the reactive load; results follow:
With B1 v+ coupled through 1mH inductance to supply oscillation occurred at 864kHz at 50mVp-p
With B1 V+ coupled through 50µH inductance to supply oscillation did not occur.
The above data suggests that this circuit should be stable in most audio applications provided the power supply is adequately bypassed and the output decoupled or kept within certain tolerances of load reactance (no big long runs of coax).
Everybody please comment here; I hope others do some tests to see what happens for them; make sure I'm not fudging anything up seriously.
I might attempt a noise measurement with the analyzer if I have time tonight but who knows if I'll get to it.
I've got the B1 on a regulated +/-5V supply, the results are as follows:
Quiescent current: 14.9mA
14mV offset (input grounded)
Frequency response is within 2dB from 100kHz to 10MHz (measured with HP 654A sig. gen and Tek 2247A scope)
Frequency response graph from 1Hz to 100kHz will follow in an attached image along with graph comparing distortion of HP signal source against amplifier output.
Gain at 1kHz:
Loaded by Fluke 179 DMM: 0.990
Loaded by Fluke 179 DMM and 1000Ω resistor: 0.967
Loaded by Fluke 179 DMM and 470Ω resistor : 0.943
Loaded by Fluke 179 DMM and 100Ω resistor : 0.790
Output Impedance at 1kHz: 25.2Ω (Something weird about that, as if transconductance is very low or something's wrong. Maybe my Iq is too low and I have JFETs at the low end of the spec.) (Or I did the math wrong three times).
I did very thorough testing to determine the source of the aforementioned oscillation. The oscillation turned out to be within the B1 and not my function generator. I did some tests to see what it took to make it oscillate:
Load: 1.5m RG56 Coaxial cable, unterminated.
Input: 470Ω resistor to ground and Tek P6109 10x probe with 2247A scope.
Power supply: Regulated +/-5V bypassed each side to ground with 1.5µF poly caps within 2cm of B1 board.
Results:
With B1 direct to supply no oscillation was detected but there was peaking at 10MHz when driven with HP 654A oscillator.
With B1 V+ coupled through 1mH inductance to supply oscillation occurred at 864kHz at 1.18Vp-p.
With B1 V+ coupled through 50µH inductance to supply oscillation occurred at 6.06MHz at 6.44Vp-p.
To reduce susceptibility to oscillation 100Ω resistance was added in series with output to decouple the reactive load; results follow:
With B1 v+ coupled through 1mH inductance to supply oscillation occurred at 864kHz at 50mVp-p
With B1 V+ coupled through 50µH inductance to supply oscillation did not occur.
The above data suggests that this circuit should be stable in most audio applications provided the power supply is adequately bypassed and the output decoupled or kept within certain tolerances of load reactance (no big long runs of coax).
Everybody please comment here; I hope others do some tests to see what happens for them; make sure I'm not fudging anything up seriously.
I might attempt a noise measurement with the analyzer if I have time tonight but who knows if I'll get to it.
Well,
a) +/- 5V supply is for me very low. I never tested anything below +/-9V; most of the time +/- 12V. If I am not wrong, the capacitances of the JFET increase quite a bit at such low voltages. Worse is that they get very non-linear and vary with Vds significantly (when you supply a voltage swing, e.g.); this might explain also the oscillation you observed.
b) 14mV offset is large. You matching does not seems to be very good. I regularly get below 2mV with little effort. I suggest you measure Idss at the same voltage (Vds) as you later use as rail voltage in the circuit.
The rest looks more or less OK. The output impedance I expect to be around 25 ohm. 1/Yfs for each JFET is about 30R, so 2 in parallel gives 15R; plus 10R degeneration (?) -> 25R.
See also :
http://www.diyaudio.com/forums/showthread.php?postid=1774452#post1774452
Again your low rail voltages might lower the Yfs further at Iq.
Patrick
a) +/- 5V supply is for me very low. I never tested anything below +/-9V; most of the time +/- 12V. If I am not wrong, the capacitances of the JFET increase quite a bit at such low voltages. Worse is that they get very non-linear and vary with Vds significantly (when you supply a voltage swing, e.g.); this might explain also the oscillation you observed.
b) 14mV offset is large. You matching does not seems to be very good. I regularly get below 2mV with little effort. I suggest you measure Idss at the same voltage (Vds) as you later use as rail voltage in the circuit.
The rest looks more or less OK. The output impedance I expect to be around 25 ohm. 1/Yfs for each JFET is about 30R, so 2 in parallel gives 15R; plus 10R degeneration (?) -> 25R.
See also :
http://www.diyaudio.com/forums/showthread.php?postid=1774452#post1774452
Again your low rail voltages might lower the Yfs further at Iq.
Patrick
I cannot draw any conclusions from the spectrum analyser outputs. The output has about 20dB higher noise floor than the input (coming from power supply?), but the 2nd harmonics from both the input and the output are about the same amount higher than the noise floor (about 20dB for both).
The FR is essentially flat, as I think 0.01dB is within measurement accuracy ?? Also the dip at LF is due to the coupling cap you are using ? Did you measure the input before or after the coupling cap ?
See also
http://www.diyaudio.com/forums/showthread.php?postid=1774455#post1774455
Patrick
The FR is essentially flat, as I think 0.01dB is within measurement accuracy ?? Also the dip at LF is due to the coupling cap you are using ? Did you measure the input before or after the coupling cap ?
See also
http://www.diyaudio.com/forums/showthread.php?postid=1774455#post1774455
Patrick
As far as I know, the distortion is compared to the fundamental frequency of stimulus and not the noise floor when dealing with an amplifier.
Correct me if I'm wrong; and I will research this to be sure.
It seems to me as if the amplifier simply has a higher noise floor than the signal source, and also some distortion. Would others please comment?
I tested the circuit with +/-9V supply and noted worse offset and more proneness to oscillation.
I will measure distortion later when I have more time.
Correct me if I'm wrong; and I will research this to be sure.
It seems to me as if the amplifier simply has a higher noise floor than the signal source, and also some distortion. Would others please comment?
I tested the circuit with +/-9V supply and noted worse offset and more proneness to oscillation.
I will measure distortion later when I have more time.
> As far as I know, the distortion is compared to the fundamental frequency of stimulus and not the noise floor when dealing with an amplifier.
That is correct. But since your input signal has already a -80dB 2nd harmonics, and the noise floor is at about -100dB, at most one can say is that the circuit adds 20dB 2nd, but to what (in the case of an ideal input signal, I would say to -120dB noise fllor, which means -100dB actual 2nd). I do not think one can say the circuit itself has a second harmonics of -60dB, because it is not getting a pure sine at the input. The performance you measured is essentially identical to what I posted, and I did not draw any conclusions then because of the same reason, namely the input source is not good enough. A THD meter, which substrates the input from the output, gain adjusted, might improve things a bit, but one needs to have a good, calibrated one around. Not everyone has. And the result is still limited by the level of the noise floor.
Since I did not experience any of your oscillation and DC offset problems, I cannot make any useful comments on those, other than to say that I suspect not good match. As Nelson said in his article, best DC and distortion performance requires good match.
Perhaps some measurements repeated with local coupling (say 1000uF within 20mm of the circuit) would be more enlightening ??
Patrick
That is correct. But since your input signal has already a -80dB 2nd harmonics, and the noise floor is at about -100dB, at most one can say is that the circuit adds 20dB 2nd, but to what (in the case of an ideal input signal, I would say to -120dB noise fllor, which means -100dB actual 2nd). I do not think one can say the circuit itself has a second harmonics of -60dB, because it is not getting a pure sine at the input. The performance you measured is essentially identical to what I posted, and I did not draw any conclusions then because of the same reason, namely the input source is not good enough. A THD meter, which substrates the input from the output, gain adjusted, might improve things a bit, but one needs to have a good, calibrated one around. Not everyone has. And the result is still limited by the level of the noise floor.
Since I did not experience any of your oscillation and DC offset problems, I cannot make any useful comments on those, other than to say that I suspect not good match. As Nelson said in his article, best DC and distortion performance requires good match.
Perhaps some measurements repeated with local coupling (say 1000uF within 20mm of the circuit) would be more enlightening ??
Patrick
Actually, -80dB represents 1/10,000th of the input. But -60dB represents 1/1000th of the output, so it clearly has 10 times more 2nd harmonic than the input, which is much more than the fundamental. Since the two components were not amplified an equal amount, it clearly added some along the way, which we call distortion. In fact, it added about 9/10,000. Since this is almost ten times more than the input had, we can say with fairly good certainty that it has -60dB distortion under this condition.
As far as oscillation goes, JFETs are good out to UHF. You may not even be able to see the oscillation except by the quirky operating point and distortion resulting from it. I've experienced this before, using a 2N5486 (an RF JFET with similar parameters). That was partly my fault, since I was building a wideband front end; such is to be expected. Insane bandwidth is not wanted or needed in audio, so you will have to reduce it. A suitable combination of RLC on the JFETs should help that. To compensate for supply inductance, for instance, a bypass capacitor is required. Frankly, I'm amazed you got useful results at all with an uncompensated circuit!
Tim
As far as oscillation goes, JFETs are good out to UHF. You may not even be able to see the oscillation except by the quirky operating point and distortion resulting from it. I've experienced this before, using a 2N5486 (an RF JFET with similar parameters). That was partly my fault, since I was building a wideband front end; such is to be expected. Insane bandwidth is not wanted or needed in audio, so you will have to reduce it. A suitable combination of RLC on the JFETs should help that. To compensate for supply inductance, for instance, a bypass capacitor is required. Frankly, I'm amazed you got useful results at all with an uncompensated circuit!
Tim
> Frankly, I'm amazed you got useful results at all with an uncompensated circuit!
John Curl has been using this with 2SK170 for years. Grey Rollins also published a version with parallel JFETs, again without PLC compensation. And they both advocate using no gate resistors or at most below 100R.
I am already using 100R gate resistors. If you wish to experiment, or do need more stability, then you might wish to try 1k instead. The bandwidth is still way too high for most audio use. But for me, supply rail decoupling is a much better solution.
Patrick
John Curl has been using this with 2SK170 for years. Grey Rollins also published a version with parallel JFETs, again without PLC compensation. And they both advocate using no gate resistors or at most below 100R.
I am already using 100R gate resistors. If you wish to experiment, or do need more stability, then you might wish to try 1k instead. The bandwidth is still way too high for most audio use. But for me, supply rail decoupling is a much better solution.
Patrick
2SK170 has about twice the capacitance as the BF862, so 2x BF862 in parallel will have about the same as one 2SK170. The RF response is, I suppose, largely determined by the capacitances of the JFETs, or not ? So I do not agree than the frequency response of the 2SK170 is vastly different.
As to noise, Scott Wurcer has already published some time ago noise measurements of the BF862 down to 10Hz, with very favorable results. I guess someone who has been designing opamps for analog Device for the last 25 years knows what he is doing and has better equipment than any of us. Please take time to have a look at his results over the Solid State forum.
If I remember correctly, someone at the solid state forum has been using BF862 for input stage of his phono preamp.
Patrick
As to noise, Scott Wurcer has already published some time ago noise measurements of the BF862 down to 10Hz, with very favorable results. I guess someone who has been designing opamps for analog Device for the last 25 years knows what he is doing and has better equipment than any of us. Please take time to have a look at his results over the Solid State forum.
If I remember correctly, someone at the solid state forum has been using BF862 for input stage of his phono preamp.
Patrick
Here are a few links of what Scott has published over the BF862 :
http://www.diyaudio.com/forums/showthread.php?postid=1815084#post1815084
http://www.diyaudio.com/forums/showthread.php?postid=1648400#post1648400
http://www.diyaudio.com/forums/showthread.php?postid=1532323#post1532323
http://www.diyaudio.com/forums/showthread.php?s=&postid=1721852&highlight=#post1721852
You can find a lot more posts within the Blowtorch thread.
http://www.diyaudio.com/forums/showthread.php?postid=1817439#post1817439
Quote :
"If you haven't checkout syn08's 32-BF862 phono stage a noise record (without magnetics) I think."
http://www.diyaudio.com/forums/showthread.php?postid=1720268#post1720268
Quote :
"Next, down to under 0.2nV/rtHz with BF862, ...."
http://www.diyaudio.com/forums/showthread.php?postid=1603576#post1603576
Patrick
http://www.diyaudio.com/forums/showthread.php?postid=1815084#post1815084
http://www.diyaudio.com/forums/showthread.php?postid=1648400#post1648400
http://www.diyaudio.com/forums/showthread.php?postid=1532323#post1532323
http://www.diyaudio.com/forums/showthread.php?s=&postid=1721852&highlight=#post1721852
You can find a lot more posts within the Blowtorch thread.
http://www.diyaudio.com/forums/showthread.php?postid=1817439#post1817439
Quote :
"If you haven't checkout syn08's 32-BF862 phono stage a noise record (without magnetics) I think."
http://www.diyaudio.com/forums/showthread.php?postid=1720268#post1720268
Quote :
"Next, down to under 0.2nV/rtHz with BF862, ...."
http://www.diyaudio.com/forums/showthread.php?postid=1603576#post1603576
Patrick
I don't really mean to say that the BF862 is particularly noisy. Simply we have noise showing up here and its source must be determined with more testing. I will find out more about it later when I have time.
I will test some 2SK170 when I find mine to see how fast they are. I only said "most likely"; it is entirely possible that they are fast, I don't yet know.
The fact is this thing still oscillates easily if perturbed. It works great as long as you make sure to bypass the supply. I like the circuit and its simplicity; I'll probably choose it for some of my projects as well. I don't mean to put it down or put anyone's ideas or work down; but I will do my best to obtain empirical data on the matter and refine my results in order to know what's going on.
I will test some 2SK170 when I find mine to see how fast they are. I only said "most likely"; it is entirely possible that they are fast, I don't yet know.
The fact is this thing still oscillates easily if perturbed. It works great as long as you make sure to bypass the supply. I like the circuit and its simplicity; I'll probably choose it for some of my projects as well. I don't mean to put it down or put anyone's ideas or work down; but I will do my best to obtain empirical data on the matter and refine my results in order to know what's going on.
> I don't mean to put it down or put anyone's ideas or work down
I wasn't suggesting that you were.
It remains a mystery to me the problems you seemed to encounter. And I tend to think it has to do with the power supply you are using, especially the low rail voltage. So I was hoping to clear some of that up.
If you look at the datasheet, at low Vds, not only does the capacitances change sharply, but also the Id (its variation with Vds).
Actually this is probably the one area where I expect the 2SK170 might be better -- that the Id is less dependent on Vds variations.
Patrick
I wasn't suggesting that you were.
It remains a mystery to me the problems you seemed to encounter. And I tend to think it has to do with the power supply you are using, especially the low rail voltage. So I was hoping to clear some of that up.
If you look at the datasheet, at low Vds, not only does the capacitances change sharply, but also the Id (its variation with Vds).
Actually this is probably the one area where I expect the 2SK170 might be better -- that the Id is less dependent on Vds variations.
Patrick
The thing is the increased power supply voltage I tried only exacerbated the problems.
Properly set up; the circuit functioned properly though. (Aside from the offset, which I know is due to a matching issue as I don't think I waited long enough for the devices to thermally stabilize during Idss measurement.)
Properly set up; the circuit functioned properly though. (Aside from the offset, which I know is due to a matching issue as I don't think I waited long enough for the devices to thermally stabilize during Idss measurement.)
I would recommend you try the following :
1) Increase the supply rails to +/-9V, and use local decoupling caps (say 100uF min.)
2) To trim the offset, you can do 2 things :
a) If you have a positive offset, attach a trimpot (say 50R) to pin 4 & 5, and adjust the trimpot until you get zero. Give it time to stabilise. A heatsink will help a lot here though I understand it might be difficult to get in your end of the world. Replace the trimpot with fixed resistor once you have it working properly.
b) If you have a negative offset, you could replace the degeneration resistor (0603) with a higher value one, until you get a positive offset, then repeat fine trimming as in (a).
Alternatively, you can also use connect the output to either the positive rail (in case of negative offset) or vice versa with an adjustable constant current diode (I would guess somewhere between 10uA to 2mA). This can be done easily using a JFET (e.g. 2SK246 and a trimpot as degeneration. But triming with the 0603 resistor is the simplest solution, and is there by design.
Patrick
1) Increase the supply rails to +/-9V, and use local decoupling caps (say 100uF min.)
2) To trim the offset, you can do 2 things :
a) If you have a positive offset, attach a trimpot (say 50R) to pin 4 & 5, and adjust the trimpot until you get zero. Give it time to stabilise. A heatsink will help a lot here though I understand it might be difficult to get in your end of the world. Replace the trimpot with fixed resistor once you have it working properly.
b) If you have a negative offset, you could replace the degeneration resistor (0603) with a higher value one, until you get a positive offset, then repeat fine trimming as in (a).
Alternatively, you can also use connect the output to either the positive rail (in case of negative offset) or vice versa with an adjustable constant current diode (I would guess somewhere between 10uA to 2mA). This can be done easily using a JFET (e.g. 2SK246 and a trimpot as degeneration. But triming with the 0603 resistor is the simplest solution, and is there by design.
Patrick