Dear Everyone,
Some of you are probably familiar with the works of AMB Labs. One of their project is called the M3 headphone amp which is a 3 channel hybrid opamp-MOSFET based headphone amp. It is a pretty decent amplifier which I have been using for 15years now. I decided to build a fully differential DAC which could directly drive a fully diff headphone amp (https://www.amb.org/forum/my-gamma3-build-with-a-built-in-fully-balanced-m3-amp-t4735.html) without issues with volume control and the low input impedance of the fully diff headphone amp. This is where my struggle began.
I found a design where a user on head-fi designed a fully diff amp based on the M3. He called it the M4: M4 head-fi link I decided that I wanted to put this into my Gamma 3 DAC until I get an AMB B22 or something similar done (there nothing similar I know, but I have something in mind).
I designed and built the module but I cannot figure out what is wrong with it. I get stuck at the very first initial setup step where the opamp bias current needs to be adjusted. On one side of the differential amp, the one which has the positive input, everything seems fine. On the other side, the R5 resistor (R3 in my design) is stuck at 2.5V = 25mA no matter what I do. The trimpot U$2_1k has no effect on the voltage. If I remove the R3 resistor, the trimpot adjust the voltage on the JFET current source's leg (although it's not a current source at that time). When I removed the MOSFET stage from the amp, I got the same symptoms. I tried to bias the JFET current source (CW pin of U$2 trimpot) to the positive PSU_OPA (V+) instead of the negative, same result. When I removed the Q2 MOSFET bias transistor, the currents were the same and were correctly showing 0.5ish voltage on both R5 resistors (R3 and R10 in my design).
Could someone take a look at my circuit and the PCB design if some error is screaming at you instantly?
I already found the following issues, but resolving them didn't resolve the issue with the CSS or the fact that I couldn't force any current on the MOSFET:
One solution I was thinking about is to remove the two constant current sources because the JFETs have high tolerance with the Idss and probably ruins the symmetry but I don't know what would happen to the OPAMPs behavior if it was not biased to class A (AMB is saying that is the CSS for but I have never seen any other design having this feature).
Please don't bash on me for not knowing what I am doing. I already know that although I am trying to get up to speed with the fully diff opamp behavior and basic principles.
This is how the module looks like IRL (the PSU doesn't have a GND wire to avoid the ground loop, would come from the same GND as the balanced input):
I would appreciate if someone could help me out. Also, after we find the error(s), if you are interested, I could send you a pair of PCBs for free so someone else could build this, as well.
Some of you are probably familiar with the works of AMB Labs. One of their project is called the M3 headphone amp which is a 3 channel hybrid opamp-MOSFET based headphone amp. It is a pretty decent amplifier which I have been using for 15years now. I decided to build a fully differential DAC which could directly drive a fully diff headphone amp (https://www.amb.org/forum/my-gamma3-build-with-a-built-in-fully-balanced-m3-amp-t4735.html) without issues with volume control and the low input impedance of the fully diff headphone amp. This is where my struggle began.
I found a design where a user on head-fi designed a fully diff amp based on the M3. He called it the M4: M4 head-fi link I decided that I wanted to put this into my Gamma 3 DAC until I get an AMB B22 or something similar done (there nothing similar I know, but I have something in mind).
I designed and built the module but I cannot figure out what is wrong with it. I get stuck at the very first initial setup step where the opamp bias current needs to be adjusted. On one side of the differential amp, the one which has the positive input, everything seems fine. On the other side, the R5 resistor (R3 in my design) is stuck at 2.5V = 25mA no matter what I do. The trimpot U$2_1k has no effect on the voltage. If I remove the R3 resistor, the trimpot adjust the voltage on the JFET current source's leg (although it's not a current source at that time). When I removed the MOSFET stage from the amp, I got the same symptoms. I tried to bias the JFET current source (CW pin of U$2 trimpot) to the positive PSU_OPA (V+) instead of the negative, same result. When I removed the Q2 MOSFET bias transistor, the currents were the same and were correctly showing 0.5ish voltage on both R5 resistors (R3 and R10 in my design).
Could someone take a look at my circuit and the PCB design if some error is screaming at you instantly?
I already found the following issues, but resolving them didn't resolve the issue with the CSS or the fact that I couldn't force any current on the MOSFET:
- MOSFETs are mirrored
- Capacitance multiplier is wired wront at the JFET
- VOCM on the OPAMP shouldn't be connected to GND
One solution I was thinking about is to remove the two constant current sources because the JFETs have high tolerance with the Idss and probably ruins the symmetry but I don't know what would happen to the OPAMPs behavior if it was not biased to class A (AMB is saying that is the CSS for but I have never seen any other design having this feature).
Please don't bash on me for not knowing what I am doing. I already know that although I am trying to get up to speed with the fully diff opamp behavior and basic principles.
This is how the module looks like IRL (the PSU doesn't have a GND wire to avoid the ground loop, would come from the same GND as the balanced input):
I would appreciate if someone could help me out. Also, after we find the error(s), if you are interested, I could send you a pair of PCBs for free so someone else could build this, as well.