After learning Jung's paper about CCS and some discussing in this forum, I pieced together this circuit. Choosing 2cs2705 is because of its high linearity, high hfe, and low cob. Choosing crd can increase the psrr of ccs, but it needs 10v to work well will increase the voltage drop. Using LED series connection to reduce noise. Another CRD keeps the operating current of the LED constant while improving the PSRR. Any suggestions?
You mean, adding another cascode.
In post #1 the lower left "E-562" is the current source, while the 2SC2705 is a cascode. "CRD" is how some manufacturers label their Current Regulating Diode product line.
You could improve PSRR in the audio band by adding a large electrolytic capacitor between the base and VEE. Panasonic EEU-FS product line of capacitors have especially low ESR.
In post #1 the lower left "E-562" is the current source, while the 2SC2705 is a cascode. "CRD" is how some manufacturers label their Current Regulating Diode product line.
You could improve PSRR in the audio band by adding a large electrolytic capacitor between the base and VEE. Panasonic EEU-FS product line of capacitors have especially low ESR.
Biased at 5.6 mA of forward current, the dynamic impedance of each LED will be, very approximately, 20 ohms. So if you connect a 330 uF electrolytic capacitor in parallel with the string of 5 series LEDs, your PSRR will increase. PSRR starts to improve at very approximately 3.4 Hertz (the "corner frequency"), then PSRR gets better and better as frequency rises. Until capacitive reactance falls below ESR, then PSRR stops getting better and flattens out. Around 12 kilohertz or so.
Last edited:
Perhaps an opamp driving a FET? Then sensing the source resistor will accurately measure the output current as a FET isolates the current path from the gate. You'd need to figure out how to power the opamp though, and have an accurate voltage reference.
You could use a FET in the circuit given of course, although it might not really gain much if the BJT's gain is high enough.
You could use a FET in the circuit given of course, although it might not really gain much if the BJT's gain is high enough.
I don't know if you are interested in the performance at high frequencies, but Microcap tells me that jfets will help a bit in that regard. Not sure this will do anything for the sound (maybe in digital applications?). Maybe it's just circuit nitpicking, which can be interesting.
The parameter that will determine high frequency rejection is Cgd, also called reverse transfer capacitance (usually Crss in datasheets). It varies with Vgd, you can see the capacitance curves in Toshiba datasheets for example.
The standard spice model for jfets defines Cgd as zero-bias junction capacitance (Vgd=0). In the simulation below, for the cascode jfet Cgd=5.3 pF (I used the J112 model included with Microcap). With the dc operating point, Vgd=17 V and this capacitance term becomes 1.9 pF. This will be the average value of this capacitance during AC simulation. The improvement is 13dB at 1 MHz compared to the Elvee floating ccs with BC550 / 560. It will depend on the voltage headroom that you have for the ccs.
It would be interesting to measure the HF performance of built circuits, and experiment with RF jfets such as those for FM tuner (which have very low Cgd).
The parameter that will determine high frequency rejection is Cgd, also called reverse transfer capacitance (usually Crss in datasheets). It varies with Vgd, you can see the capacitance curves in Toshiba datasheets for example.
The standard spice model for jfets defines Cgd as zero-bias junction capacitance (Vgd=0). In the simulation below, for the cascode jfet Cgd=5.3 pF (I used the J112 model included with Microcap). With the dc operating point, Vgd=17 V and this capacitance term becomes 1.9 pF. This will be the average value of this capacitance during AC simulation. The improvement is 13dB at 1 MHz compared to the Elvee floating ccs with BC550 / 560. It will depend on the voltage headroom that you have for the ccs.
It would be interesting to measure the HF performance of built circuits, and experiment with RF jfets such as those for FM tuner (which have very low Cgd).
Last edited:
That many LED's will not thermally cancel the NPN CCS tempco (way negative - 5X the NPN !). Just one red led bypassed with 10uF + .1uf is ideal.
PS - a 2Q CCS is the most "constant" ...