Hi Jeepee,
Does it have to be amplified with the LNA? Measured with a Fluke 8920A shows about 500uV rms. In the range of 20Hz - 200kHz the analyzer shows a lot of little "mountains" of 1-5uV. At 1kHz the noise spectral density is measured at 90nV/rtHz.
Does it have to be amplified with the LNA? Measured with a Fluke 8920A shows about 500uV rms. In the range of 20Hz - 200kHz the analyzer shows a lot of little "mountains" of 1-5uV. At 1kHz the noise spectral density is measured at 90nV/rtHz.
National is rolling out a new 0.69nV/RtHz opamp == the LMH6629
http://www.national.com/ds/LM/LMH6629.pdf
Not exactly a low distortion opamp, however.
http://www.national.com/ds/LM/LMH6629.pdf
Not exactly a low distortion opamp, however.
Hi Ikoflexer,
thanks for doing the measurements.
Referring to the circuit in post #238, the 90nV/rtHz will inject a current of 90 nV/130 Ohm=700pA/rtHz into the input of the transimpedance amplifier. (the + input of U1 is referred to ground, so the - input is a virtual ground. The supply ripple occurs directly across R3, injecting the current into the very low input impedance of the transimpedance stage)
Now this 700pA/rtHz supply ripple induced current adds to the amplified drain current of the input JFET's. Taking optimisticly low JFET input referred noise of 0.2nV/rtHz, amplified by 9*gm (taking conservatively low 30mS), the total drain current noise will be 0.2nV/rtHz*9*30mS=54pA/rtHz, or 13x lower than the supply ripple induced current.
In order to have less than 1% error due to the supply ripple, the PSRR needs to be 1300x, or 62dB. Now check again post #355 😉
thanks for doing the measurements.
Referring to the circuit in post #238, the 90nV/rtHz will inject a current of 90 nV/130 Ohm=700pA/rtHz into the input of the transimpedance amplifier. (the + input of U1 is referred to ground, so the - input is a virtual ground. The supply ripple occurs directly across R3, injecting the current into the very low input impedance of the transimpedance stage)
Now this 700pA/rtHz supply ripple induced current adds to the amplified drain current of the input JFET's. Taking optimisticly low JFET input referred noise of 0.2nV/rtHz, amplified by 9*gm (taking conservatively low 30mS), the total drain current noise will be 0.2nV/rtHz*9*30mS=54pA/rtHz, or 13x lower than the supply ripple induced current.
In order to have less than 1% error due to the supply ripple, the PSRR needs to be 1300x, or 62dB. Now check again post #355 😉
Thanks Jeepee! Couple of small points. One, not so important: should it not be sqrt(9)*gm? Two: might look like the filtering is over the top, but I could measure large enough a difference in the noise when the Sziklai buffer was used and when the cap multiplier was used with the lab psu. This might be one of those cases in which the joke about the French applies: "it works in practice, but does the theory hold?" 🙂
Hi Ikoflexer,
The gain of 9 parallel JFET's is simply 9 times the gain of 1 JFET.
The noise of 9 parallel JFET's is not equal to the 9 times the noise of 1, since each JFET has its own uncorrelated noise source.
That's where the factor sqrt(9) comes in.
PS: Shouldn't your new logo read "Authorized Personnel Only 0.0000000003 Volts" ?
The gain of 9 parallel JFET's is simply 9 times the gain of 1 JFET.
The noise of 9 parallel JFET's is not equal to the 9 times the noise of 1, since each JFET has its own uncorrelated noise source.
That's where the factor sqrt(9) comes in.
PS: Shouldn't your new logo read "Authorized Personnel Only 0.0000000003 Volts" ?
Last edited:
actually, I don't know their parameter spreads. 😕
Have you tried FET opamps? Bipolars doesn't seemed to be first choice here.
regards
Unfortunately there is no opamp that I know of that has such low noise as a few paralleled jfets. Paralleling opamps is an idea that I don't find appealing because of the high cost.
nope, I was referring to the second stage opamp and the integrator
just curiosity, nothing else...
just curiosity, nothing else...
Last edited:
I have a question about PSRR.
I've seen this happen several times, where someone comes up with a circuit here or somewhere else, and others come see it, and they'll comment on how the circuit has low PSRR and is bad engineering. Two different amplifiers get powered by the same power supply, so they both see the same noise. Circuit A can reject only half the PSU noise. Circuit B has all the mechanism to reject twice the PSU noise circuit A can, within the amplifier. Let's assume that a simple capacitor multiplier can filter enough of the PSU noise that we will measure the same noise density value at the output of both amplifiers.
So if the noise at the output is equal, then why is approach A worse than B?
1) One can't claim simplicity makes circuit B better because improving the PSRR greatly seems to be quite complex.
2) Is it that improving the PSRR of an amp involves some sort of canceling circuitry and that is inherently superior? I think it can be done also outside the amplifier, in a simple regulator.
In my opinion, moving some complexity out of the amplifier and into the power supply/filter should be a good thing. There are enough issues to deal with in the amplifier, like stability, phase shift, distortion, for instance, so why put even more effort to improve the PSRR within the circuitry of the amplifier, at the possible cost of having all the other factors affected.
I can understand someone selling the amplifier without the power supply, as is the case with opamps. Sure, then high PSRR is definitely a feature, to appeal to the widest audience, including those who don't care to look into a good psu.
But here dealing with this specialty preamp it seems to me a better approach to have the two items separate: low noise power supply, and low noise amplifier.
Samuel, you were one of the people that insisted that I try to improve the PSRR, perhaps I can have your opinion on this?
Anyone else?
I've seen this happen several times, where someone comes up with a circuit here or somewhere else, and others come see it, and they'll comment on how the circuit has low PSRR and is bad engineering. Two different amplifiers get powered by the same power supply, so they both see the same noise. Circuit A can reject only half the PSU noise. Circuit B has all the mechanism to reject twice the PSU noise circuit A can, within the amplifier. Let's assume that a simple capacitor multiplier can filter enough of the PSU noise that we will measure the same noise density value at the output of both amplifiers.
So if the noise at the output is equal, then why is approach A worse than B?
1) One can't claim simplicity makes circuit B better because improving the PSRR greatly seems to be quite complex.
2) Is it that improving the PSRR of an amp involves some sort of canceling circuitry and that is inherently superior? I think it can be done also outside the amplifier, in a simple regulator.
In my opinion, moving some complexity out of the amplifier and into the power supply/filter should be a good thing. There are enough issues to deal with in the amplifier, like stability, phase shift, distortion, for instance, so why put even more effort to improve the PSRR within the circuitry of the amplifier, at the possible cost of having all the other factors affected.
I can understand someone selling the amplifier without the power supply, as is the case with opamps. Sure, then high PSRR is definitely a feature, to appeal to the widest audience, including those who don't care to look into a good psu.
But here dealing with this specialty preamp it seems to me a better approach to have the two items separate: low noise power supply, and low noise amplifier.
Samuel, you were one of the people that insisted that I try to improve the PSRR, perhaps I can have your opinion on this?
Anyone else?
Well, let me disabuse you. 😀
All things being equal, higher PSR is better than lower PSR. But things aren't usually equal. One can design a great circuit, high bandwidth, low distortion, low noise, but lousy PSR. Now, the downside is that the power supply must be built heroically. But there are lots of examples of well-designed gear using that philosophy.
Others (I'm one of them) shoot for high PSR and simpler power supplies. One could argue that it's just transferring the complication from one part of the circuit to another. And that's a reasonable argument (though I think my circuits aren't all that complicated).
That's why engineering is interesting- there's lots of choices and trade-offs. In your example, there WILL be a noise difference. You can't use the same supply with both circuits and expect the same results.
All things being equal, higher PSR is better than lower PSR. But things aren't usually equal. One can design a great circuit, high bandwidth, low distortion, low noise, but lousy PSR. Now, the downside is that the power supply must be built heroically. But there are lots of examples of well-designed gear using that philosophy.
Others (I'm one of them) shoot for high PSR and simpler power supplies. One could argue that it's just transferring the complication from one part of the circuit to another. And that's a reasonable argument (though I think my circuits aren't all that complicated).
That's why engineering is interesting- there's lots of choices and trade-offs. In your example, there WILL be a noise difference. You can't use the same supply with both circuits and expect the same results.
I don't think I can help you much more here without just showing a complete solution--all points which I consider important for such a design have been mentioned. Hence I'd like to keep my contribution to this thread to a minimum.
It is not true that it is not possible to improve PSRR without significantly affecting complexity or other performance aspects. In this thread two easy solutions have been given; one considers the reduction of the input stage bias and the other decouples the integrating opamp to the positive rail. The key point is that you understand why the PSRR of your current plan is so bad (it has been explained in this thread)--then you can look for solutions.
Samuel
I'm sorry but I think you misunderstood my question. I did not ask which points you consider important, since you have already said. I asked why improving PSRR within the preamp is considered a superior solution to improving the power supply.
I understand why the PSRR of the current plan is low.
About the reduction of the input stage bias, which was your proposed solution. You suggested to run the first stage at low current (around 1mA). A larger load resistor, and another resistor below the source resistor, bypassed by a capacitor would accomplish this. But to improve the PSRR of this stage significantly the resistor values have to be quite large, so the self-noise of the first stage would increase.
I will look at how much the PSRR improves with decoupling the integrator to the positive rail and get back.
Please do not give me a complete solution, this is not what I need. And thank you for your understanding; perhaps you remember that six years ago you were a beginner too.
It is not clear to me how you came to this conclusion; biasing the JFETs at 0.1*Idss instead of Idss typically increases their voltage noise by say 10%. This is not much of an issue if we consider that:
* PSRR increased by 20 dB
* noise contribution from the first stage drain resistor is reduced
* noise contribution from the voltage noise of the second stage opamp is much reduced
* operating temperature of the JFETs is greatly reduced
* battery life went up almost 10x
All in all this will at least partially make up for the 10% voltage noise increase or even lead to an overall lower noise amp.
1 mA per JFET. So 9 mA for 9 transistors. High enough to get well below 0.5 nV/sqrt(Hz). An IF3601 does 0.3 nV/sqrt(Hz) at 5 mA (according to the datasheet).
It seems much better to me to bias the gate somewhat negative as suggested earlier.
Samuel
Last edited:
I tried the circuit in spice because I thought it should be a decent approximation of the thermal noise involved. I noticed that you did not include the load resistor.
Regarding the first stage, is its current noise of more concern than its voltage noise?
Yes, I misunderstood and tried 1mA total current (111uA each jfet). Have no access to IF3601.
I will try it.
I don't understand--where did I not include which resistor?
Again I'm not sure I understand you correctly--whether amplifier voltage or current noise is more of a concern depends on source impedance. But this is very well covered in the opamp literature and not worth repeating here.
Samuel