The simulation follows the measured result closely, so its fair to say the model is good enough for further discussion and optimization.
The attached is the LTSPICE file for my version of the J-Mo II. It has higher gain than the regular version, and lower bias current, optimized for 300 ohm loads.
It's good for 70 mW at 1% THD into 300 ohms. Distortion falls away as the signal voltage gets smaller. At typical listening levels, its much lower.
The attached is the LTSPICE file for my version of the J-Mo II. It has higher gain than the regular version, and lower bias current, optimized for 300 ohm loads.
It's good for 70 mW at 1% THD into 300 ohms. Distortion falls away as the signal voltage gets smaller. At typical listening levels, its much lower.
Attachments
Technical discussion:
The Mosfet source follower output stage used in this and all Szekeres-based headphone amplifiers basically sucks in terms of distortion, and the problem goes from bad to worse as the output load increases. If high THD figures bother you, this is terrible circuit element to use.
So, similar to single-ended tube amp design, isn't spending inordinate effort finessing this circuit to minimize the distortion is a bit ... backwards? Well, yes and no I guess. Its human nature to try and optimize the chosen circuit, even for a choice which is inherently poor. Even if we chose it for the way it sounds, part-and-parcel of which is the very harmonics we are busily trying to get rid of!
My compromise position runs something like this: yes, single-ended MOSFET source followers are going to have substantial harmonic components, and yes this is part of the reason we went with this topology, but it's not the whole reason (others being low noise, no crossover distortion, and very simple distortion characteristics) and within the given power supply voltage and current/thermal limitations its desirable to tweak the circuit to keep the distortion as low as possible, at as high an output power as possible.
Fiddling around with the simulation, I conclude that the cleanest signal is generally obtained with the Mosfet gate voltage biased well above half the power supply. This is done by adjusting the trim pot on the jfet drain. Overreach however and the jfet current drops below a critical threshold... if this happens the Sziklai element has to be reconfigured, perhaps substituting a jfet with higher Vgs0, or alternatively increasing the gain.
This kind of optimization can drop the second harmonic component 5-10 dB, so its not trivial. How much different it makes to the sound though is something I'll discover shortly.
At the moment, at -70 dB THD at typical max listening levels, I would not characterize the sound to be at all "fat", like you might expect for excessive 2nd harmonic. The amp though does have a very definite character. It's very quiet, for one. At the same time the music itself sounds slightly hushed, even when playing very loud. Like being in the midst of gently falling snow.
I now have four high-quality headphone amps in house. Two with push-pull bipolar output stages, two with single-ended mosfets. Some weekend I'll have to do a shoot out...
The Mosfet source follower output stage used in this and all Szekeres-based headphone amplifiers basically sucks in terms of distortion, and the problem goes from bad to worse as the output load increases. If high THD figures bother you, this is terrible circuit element to use.
So, similar to single-ended tube amp design, isn't spending inordinate effort finessing this circuit to minimize the distortion is a bit ... backwards? Well, yes and no I guess. Its human nature to try and optimize the chosen circuit, even for a choice which is inherently poor. Even if we chose it for the way it sounds, part-and-parcel of which is the very harmonics we are busily trying to get rid of!
My compromise position runs something like this: yes, single-ended MOSFET source followers are going to have substantial harmonic components, and yes this is part of the reason we went with this topology, but it's not the whole reason (others being low noise, no crossover distortion, and very simple distortion characteristics) and within the given power supply voltage and current/thermal limitations its desirable to tweak the circuit to keep the distortion as low as possible, at as high an output power as possible.
Fiddling around with the simulation, I conclude that the cleanest signal is generally obtained with the Mosfet gate voltage biased well above half the power supply. This is done by adjusting the trim pot on the jfet drain. Overreach however and the jfet current drops below a critical threshold... if this happens the Sziklai element has to be reconfigured, perhaps substituting a jfet with higher Vgs0, or alternatively increasing the gain.
This kind of optimization can drop the second harmonic component 5-10 dB, so its not trivial. How much different it makes to the sound though is something I'll discover shortly.
At the moment, at -70 dB THD at typical max listening levels, I would not characterize the sound to be at all "fat", like you might expect for excessive 2nd harmonic. The amp though does have a very definite character. It's very quiet, for one. At the same time the music itself sounds slightly hushed, even when playing very loud. Like being in the midst of gently falling snow.
I now have four high-quality headphone amps in house. Two with push-pull bipolar output stages, two with single-ended mosfets. Some weekend I'll have to do a shoot out...
Last edited:
What the previous post says, in graphical format. This is real data, not simulation. By way of comparison the sapphire amp has no measurable distortion, i.e. no points above the black dashed line. However it's worth noting that real world listening power levels are not even represented on the chart, they are down at 0.01 mW where the THD is in any case about 0.03~0.05% and almost entirely 2nd harmonic...
Attachments
Last edited:
Simulated results with LTSpice, for the standard BOM version driving a 16 ohm load.The BOM version is optimized at about 40 ohms, the idea being it will be reasonably comfortable driving anything from 15 to 120 ohms. 16 ohms is a "worst case" situation, more power / less distortion will be available into 30 or 60 ohms for example.
As shown, it will do 40 mW into 16 ohms at 1% THD.
As shown, it will do 40 mW into 16 ohms at 1% THD.
Attachments
Last edited:
The general layout of the PCB (which is single-sided) can be used for a P2P build also.
The heat dissipation by the MOSFET is quite low, 1~2 W. You could, however, arrange the MOSFET to bolt to the case instead of using a heatsink.
In line with the original Szekeres, this is a "fun" circuit, more for how it sounds than for being the last word in fidelity, so feel free to change values as you want.
The heat dissipation by the MOSFET is quite low, 1~2 W. You could, however, arrange the MOSFET to bolt to the case instead of using a heatsink.
In line with the original Szekeres, this is a "fun" circuit, more for how it sounds than for being the last word in fidelity, so feel free to change values as you want.
@RCruz You can use IRF610 and 2N5457 for sure 
I've built the Szekeres amp with LM317 as current source (at the place of R9) and there is something special about single ended class A sound. It has been reported many times that the SQ in this topology greatly improves upon using a low-impedance buffer before the MOSFET and I had the idea to try an OpAmp + MOSFET but I'd better try J-MoII first.
@RJM - I hope you don't mind etching our own boards at home for the sake of non-scientific measurement experience, do you?
I've built the Szekeres amp with LM317 as current source (at the place of R9) and there is something special about single ended class A sound. It has been reported many times that the SQ in this topology greatly improves upon using a low-impedance buffer before the MOSFET and I had the idea to try an OpAmp + MOSFET but I'd better try J-MoII first.
@RJM - I hope you don't mind etching our own boards at home for the sake of non-scientific measurement experience, do you?
@Gloop
You can use and or modify the .brd files to make your own PCBs. That's fine with me.
I think the op-amp + MOSFET has merit, but it is best paired with a CCS on the MOSFET source. If you want to stick with the resistive load of the original Szekeres, then the jfet-bjt input stage of the JMo2 is a nice matching front end. Another route is the triode + MOSFET mu-follower.
@RCruz
"Can I use IRF610 and 2N5457 instead ? (Or a 2sk170 at 8mA)" Yeah, whatever takes your fancy, really. The pnp transistor, too, can be substituted with just about anything. In the worse case you find that the trim resistor is insufficient to properly adjust the source bias current. In that case have to adjust the values of R2,R3 up or down (keeping their ratio R2/R3 constant). High values reduce the current in the transistor, increasing the current through the jfet and lowering the MOSFET source current, lower values do the opposite. If you are unsure and can't run it through LTSpice to check give me a shout and I'll sim it for you.
You can use and or modify the .brd files to make your own PCBs. That's fine with me.
I think the op-amp + MOSFET has merit, but it is best paired with a CCS on the MOSFET source. If you want to stick with the resistive load of the original Szekeres, then the jfet-bjt input stage of the JMo2 is a nice matching front end. Another route is the triode + MOSFET mu-follower.
@RCruz
"Can I use IRF610 and 2N5457 instead ? (Or a 2sk170 at 8mA)" Yeah, whatever takes your fancy, really. The pnp transistor, too, can be substituted with just about anything. In the worse case you find that the trim resistor is insufficient to properly adjust the source bias current. In that case have to adjust the values of R2,R3 up or down (keeping their ratio R2/R3 constant). High values reduce the current in the transistor, increasing the current through the jfet and lowering the MOSFET source current, lower values do the opposite. If you are unsure and can't run it through LTSpice to check give me a shout and I'll sim it for you.
Schematic is on the web page, and included below.
Please note the LTSpice simulation part numbers are not consistent with the formal schematic and BOM.
R2 is 470 ohms and R3 is 1.5k in the BOM version.
A quick check, with the IRF530 standing in for the 640 showns no significant degradation of the bandwidth. (even the IRFP240(chosen at random) is -3dB at 2 Mhz) The BC560C should be able to drive it. I don't have a model for the 2SK170 but with the 5486 subbing the circuit works great with the BOM values for R2,3.
Please note the LTSpice simulation part numbers are not consistent with the formal schematic and BOM.
R2 is 470 ohms and R3 is 1.5k in the BOM version.
A quick check, with the IRF530 standing in for the 640 showns no significant degradation of the bandwidth. (even the IRFP240(chosen at random) is -3dB at 2 Mhz) The BC560C should be able to drive it. I don't have a model for the 2SK170 but with the 5486 subbing the circuit works great with the BOM values for R2,3.
Attachments
Last edited:
Ok... I finally got it working (thanks to rjm good posts... read it all and than reread)
I used a J310 and a BC560C for the input.
R2 240ohm
R3 1500ohm
I can easily set collector voltage from 1.4 to 13.9v using the 1k trimmer.
Current on the fet is 2.6mA and current on the bjt is 5.1mA
Should I expect this contraption to be able to drive the mosfet ?
I used a J310 and a BC560C for the input.
R2 240ohm
R3 1500ohm
I can easily set collector voltage from 1.4 to 13.9v using the 1k trimmer.
Current on the fet is 2.6mA and current on the bjt is 5.1mA
Should I expect this contraption to be able to drive the mosfet ?
Last edited:
Now about some détails.... I believe R9 (47r) forms a filter with caps C3 and C4.... can it be removed ?
Please bear in mind that I am only doing this as a Learning method... not trying to "deoptimize" your splendid work... And I am working with whatever I have in my spares box.....
Please bear in mind that I am only doing this as a Learning method... not trying to "deoptimize" your splendid work... And I am working with whatever I have in my spares box.....
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.