Which CCS is to prefer?
There is an article which compares different ways to make a constant current source.
But I can't find this article.
I need help to find it.
Thank you!
There is an article which compares different ways to make a constant current source.
But I can't find this article.
I need help to find it.
Thank you!
https://nathaliebeimler.com/tech/audio_amplifier_current_source_comparison.html
Among them, this is the best.
Among them, this is the best.
Here is a simulation I did.
The two first use LT1004 reference diode. First is not cascoded. The second is cascoded.
The cascoded with LT1004 has the lowest impact from the signal at supply 1kHz.
TLV431 has also rather good result.
Two diodes 1N4148 has the poorest result.
One LED and the two transistors are not good either.
The two first use LT1004 reference diode. First is not cascoded. The second is cascoded.
The cascoded with LT1004 has the lowest impact from the signal at supply 1kHz.
TLV431 has also rather good result.
Two diodes 1N4148 has the poorest result.
One LED and the two transistors are not good either.
Split each 6.8K resistor into two 3.3K resistors in series. Then attach a 47uF electrolytic cap between (the middle node where the 3.3K resistors connect) and (ground), for every current source.
Or, install a LM7809 voltage regulator. +15V input, +9V output. Bypass the 9V rail to ground with 100uF. Then attach each of the 6.8K resistors to +9V instead of +15V. Perhaps think about whether you wish to change those resistors to 6800 * 9 / 15 = 3.9K instead of 6.8K. "Mountain View" does this except the voltage regulator isn't an IC, it's just a resistor and a zener diode.
Or, install a LM7809 voltage regulator. +15V input, +9V output. Bypass the 9V rail to ground with 100uF. Then attach each of the 6.8K resistors to +9V instead of +15V. Perhaps think about whether you wish to change those resistors to 6800 * 9 / 15 = 3.9K instead of 6.8K. "Mountain View" does this except the voltage regulator isn't an IC, it's just a resistor and a zener diode.
Here is the current generator i use to test VAS stages. It has really high output impedance even without the fet but it gives zero blocking time at overload.
It is an ideal simulated OP so dont expect it to work in the real world.
The output impedance is way over 100 Megaohms but limited by the output capacitance of the 2N7002.
It is an ideal simulated OP so dont expect it to work in the real world.
The output impedance is way over 100 Megaohms but limited by the output capacitance of the 2N7002.
I was looking into the "Best" VAS when I was developing an amp with folded cascoded VAS. The open loop voltage gain of those folded cascoded VAS is gm*Rout, where gm is transconductance of the input stage, and the Rout is the output impedance of the VAS. Assuming the gm is 1/100, to get 100dB open loop gain, the Rout has to be greater than 10MOhm. Nothing is better than the "NFB assistanted CCS", shown in #5.
When using no cascode the one in post #5 is clearly the best.
When using one BJT as cascode, most any method is very good.
Look here:
The old fashion standard current generator has also about 10Mohm out impedance.
A fun thing is that you can connect a link to the current sence resistor to cancel the output impedance of the transistor T1.
C2 for the capacitance and R7 for the early effect. They can even cancel input impedance of the output stage.
Sorry only that these effects are voltage and current dependent so it is only in simulation you can really cancel.
In the real world is it sometimes possible to reduce distortion to 1/3.
I have only seen this technik in one single amplifier.
A fun thing is that you can connect a link to the current sence resistor to cancel the output impedance of the transistor T1.
C2 for the capacitance and R7 for the early effect. They can even cancel input impedance of the output stage.
Sorry only that these effects are voltage and current dependent so it is only in simulation you can really cancel.
In the real world is it sometimes possible to reduce distortion to 1/3.
I have only seen this technik in one single amplifier.
This is a positive feedback. I would be caution with this approach. If you over do it, the performance may get worse or unstable.
10MOhm Rout is very good already. If you need more, as you have done already, cascode a MOSFET on top of it.
I was surprised when i noticed that it was not so sensitive as i thought. It seems as if only the negative feedback is stronger it works.
I measured the signal current in the VAS and it actually increased when it was to much positive feedback to the current sense.
But the distortion didn´t sink as much. It was mostly the clean undistorted signal i cancelled out. The R6 is important to limit the bandwith. I tested only for one decade.
Just a few minutes ago i got the idea to replace the feedback capacitor with a transistor of the same type and shortcircuit B-E just to have the capacitance B-C.
Seems to work very good.
I measured the signal current in the VAS and it actually increased when it was to much positive feedback to the current sense.
But the distortion didn´t sink as much. It was mostly the clean undistorted signal i cancelled out. The R6 is important to limit the bandwith. I tested only for one decade.
Just a few minutes ago i got the idea to replace the feedback capacitor with a transistor of the same type and shortcircuit B-E just to have the capacitance B-C.
Seems to work very good.