H
HAYK
Lsk489 comes smd or TO71. On your PCB it is Q1 and Q8 are to implement TO71?
Digikey asking $5+$20 shipment for smd.
Digikey asking $5+$20 shipment for smd.
Too expensive , unless you want bragging rights @ 1K/us.
The big power stages are what slows things down. My CFA driving a class A (BD139/140) output stage
can do 600+V/uS.
OS
The big power stages are what slows things down. My CFA driving a class A (BD139/140) output stage
can do 600+V/uS.
OS
H
HAYK
Isn't it too high running the jfets at 20v 4ma? I optioned for BF862 only $1 for 10 from Ali. The zener will be 8v and I don't know the current as it has 20mS VS 3mS T 2ma.
Do you really expect that BF862 from AliExpress are genuine?
Absolute zero chance.
If you do, then why restrain yourself?
For 0.1 $ a piece, you can purchase any exotic unobtanium (audiophile wet dream) part you can imagine. Ali will provide (at best, picture of part may be genuine).
H
HAYK
After 20 years, patents fall in public domain and components can be generic.
It costs $0.66+Ship.
It costs $0.66+Ship.
this is going offtopic but if the lsk489's are unavailable to you id convert to degenerated bjt input. reasults wil not be far off.
it will get warm to hot-ish. im running 3m2A now. Ive seen in simulation the more voltage the lower intermodulation.
H
HAYK
Do not convert your circuit, as you will apply 6 pin socket, it will be ready for any jfet. I find strange that with 2n7000 I had lower distortion 4db higher loop gain and higher slew rate. With BF862 I had to reach 7.5ma bias to get back 1kv/us and 100db NFB @20Khz.
To run with +/-36v supply, I decreased the CCS 330ohm resistors to100 to lower the dropout to 1.5v. I have now 30vp 1khz 0.00007% THD always with a gain of 34.
To run with +/-36v supply, I decreased the CCS 330ohm resistors to100 to lower the dropout to 1.5v. I have now 30vp 1khz 0.00007% THD always with a gain of 34.
Attachments
Last edited by a moderator:
The 2n7000 has much more transconductance. Im using fairly low capacitance low transconductance jfets. the vas is actually current mirror loaded to make it push pull so its not quite a constant current source.
H
HAYK
MOSFETs at low currents are perfect square law transfer function, that is the transconductance increases linearly. When used in differential mode, it results very linear.
For 2n7000, noise is not specified, different brands differ.
For 2n7000, noise is not specified, different brands differ.
For this project im not interested in using a mosfet input pair. That said, your build is yours.
Experimentation will continue when pcbs arrive on wednesday.
And excuse me I see now you were talking about the input load CCS's. Reducing those resistors does reduce VAS current and may increase thermal drift. Should work anyhow.
Experimentation will continue when pcbs arrive on wednesday.
And excuse me I see now you were talking about the input load CCS's. Reducing those resistors does reduce VAS current and may increase thermal drift. Should work anyhow.
Last edited:
H
HAYK
To respect your idea, I am not using MOSFET but jfet as you prescribed, only I chose cheaper BF862. The CCS resistors need to have 100mv min, as I increased the bias current 9ma to 15ma, 100 ohm gives 750mv. The VAS current is increased by 100 ohm common resistor to be 3.750ma for each VAS, maybe 75 ohm will be better. It decreases the positive dropout to be equal to the negative one. I adapted to my autobiased OPS without touching the compensation. I get without feedback 0.27% instead of 0.3% standalone I don't know why, full feedback it falls to 0.0006% keeping perfect square wave. I am gratefull to ingenious ideas that will alow me realize my last analog amplifier.
Last edited by a moderator:
Dear Hayk, you seem to know what youre doing indeed. In the end its the voltage over the resistors that matter; thermally speaking.
Interesting getting 0.27% instead of 0.3%. Is that with the exact same voltage over the load? Your front end does have some output impedance perhaps that affects the OPS a bit.
Interesting getting 0.27% instead of 0.3%. Is that with the exact same voltage over the load? Your front end does have some output impedance perhaps that affects the OPS a bit.
Be very suspicious of BF862 counterfeits. They were very nice FETs but obsoleted. We had to redesign a product that relied upon them. There was an industry-wide scramble to find remaining stock, so I'm very skeptical about your source.
P.S. This is the part we eventually made work. Performance seemed very comparable, but the application was not audio. https://www.onsemi.com/products/discrete-power-modules/jfets/cph3910
P.S. This is the part we eventually made work. Performance seemed very comparable, but the application was not audio. https://www.onsemi.com/products/discrete-power-modules/jfets/cph3910
Last edited:
That's my recollection as well. The Onsemi part was not as thoroughly characterized for noise performance, but on the bench seemed very good competition. We used it from about 50kHz up to about 30MHz, but I bet it was comparable at audio frequencies.
Here are some intermediate construction shots.
For now playing it safe with the high f compensation. The overcurrent protection components will be added and tested later. First I wanna get a baseline for performance. Also im shorting the output inductor. Untill i figure out how to turn neat coils that is.
Cheers,
Ruben
For now playing it safe with the high f compensation. The overcurrent protection components will be added and tested later. First I wanna get a baseline for performance. Also im shorting the output inductor. Untill i figure out how to turn neat coils that is.
Cheers,
Ruben
Nice ! Does not look like DIY ? Pro !
PS - why did you not just go all SMD. At least for small signal semi's ?
OS
PS - why did you not just go all SMD. At least for small signal semi's ?
OS
