Hi dear JOSI1,
Great to hear from you.
Agreed. Darlington squared is good enough to beat any amp.
It will be the new "approuved" AMNESIS amp output.
My faulty amp was due most probably by a bad solder because I revisited the simulations and the amp should work OK with no early clipping with VRefs as low a 3.2V (+/-) for the drivers cascodes.
BTW, the VAS cascode will work with VRef as low as 2.4V, as per simulation, but the lowest I tried is 3.6V, with selected yellow LEDs. I do not recomend going so low.
Both, my VFET and MOSFET amps are working OK with great sound and stability. when I finish my speaker protection I will be able to slowly compare amps.
Agreed. That was my own impression back then.
Remember to try good quality (low inductance; low capacitance) 47R inter-transistor resistor VAS if you can.
That would be "bias current", right?
I use 110mA for my VFET Darlington. Honestly, I did not make critical comparison between different levels of bias: it sounded equaly good wherever I put the goal...
My new MOSFET amp is runing on very low bias before I add speaker protection, and it sounds great.
The QUADs are using something like 65mA for one and +/-90mA for the other. I set them where I found stability met good sound.
Simulation show 5R//1uH are also good. Mine are Vishay Draloric (3W/5W?) if I remember correctly...
My tester cannot measure so low inductance so I make 10 turns of 1.5cm diameter of AWG18 cable, more or less, which was a recommended receipe. Now that you mention it, I definitely must try an OCC copper inductor!
I am so thankful for all your effort and support (and courage) for my crazy project.
At this time I can say all the goals of the experiment are achieved and we have an amp that is transparent, open and dynamic enough to get a satisfactory listening experience from classical music, which is the most demanding genre. Obviously, it also will satisfy in other genres.
Some day, I will try the Sicklay output, but not for now...
Best wishes of peace and prosperity for all,
M.
Great to hear from you.
Agreed. Darlington squared is good enough to beat any amp.
It will be the new "approuved" AMNESIS amp output.
My faulty amp was due most probably by a bad solder because I revisited the simulations and the amp should work OK with no early clipping with VRefs as low a 3.2V (+/-) for the drivers cascodes.
BTW, the VAS cascode will work with VRef as low as 2.4V, as per simulation, but the lowest I tried is 3.6V, with selected yellow LEDs. I do not recomend going so low.
Both, my VFET and MOSFET amps are working OK with great sound and stability.
when I finish my speaker protection I will be able to slowly compare amps.Agreed. That was my own impression back then.
Remember to try good quality (low inductance; low capacitance) 47R inter-transistor resistor VAS if you can.
That would be "bias current", right?
I use 110mA for my VFET Darlington. Honestly, I did not make critical comparison between different levels of bias: it sounded equaly good wherever I put the goal...
My new MOSFET amp is runing on very low bias before I add speaker protection, and it sounds great.
The QUADs are using something like 65mA for one and +/-90mA for the other. I set them where I found stability met good sound.
Simulation show 5R//1uH are also good. Mine are Vishay Draloric (3W/5W?) if I remember correctly...
My tester cannot measure so low inductance so I make 10 turns of 1.5cm diameter of AWG18 cable, more or less, which was a recommended receipe. Now that you mention it, I definitely must try an OCC copper inductor!
I am so thankful for all your effort and support (and courage
) for my crazy project.At this time I can say all the goals of the experiment are achieved and we have an amp that is transparent, open and dynamic enough to get a satisfactory listening experience from classical music, which is the most demanding genre. Obviously, it also will satisfy in other genres.
Some day, I will try the Sicklay output, but not for now...
Best wishes of peace and prosperity for all,
M.
Last edited:
Speaker protection circuits (with relays) are working now. I have to complete the box and then I will be able to perform critical listening test...well, not so critical as who can critisize objectivelly his new-born child?
I promised myself to rest BUT, I lied (to myself) as I have here unused good +/-60V power supplies (from another project) and I think I owe the audience an amp with higher power headroom...
I simulated a version with two pairs of outputs, probably not the most correct technically, but it works. With 1.85V input signal it devellops +/-50V output without clipping and with 0.00093% THD. The MOSFETs are dissipating under 8W.
I have to modifie the input to an "ALLBJT" version to handle the increase in PS voltage. That could alter the THD, but not too much.
Cheers,
M.
I promised myself to rest BUT, I lied (to myself) as I have here unused good +/-60V power supplies (from another project) and I think I owe the audience an amp with higher power headroom...
I simulated a version with two pairs of outputs, probably not the most correct technically, but it works. With 1.85V input signal it devellops +/-50V output without clipping and with 0.00093% THD. The MOSFETs are dissipating under 8W.
I have to modifie the input to an "ALLBJT" version to handle the increase in PS voltage. That could alter the THD, but not too much.
Cheers,
M.
Attachments
Last edited:
A little mistake with the last example: I tried to split R from the VBe multiplier but I wrongly put there a 600R (VAR) which must be 400R(VAR) meaning, in practice, splitting the 680R into one fixed +/-300 and one 500R or 1K variable multiturn resistor. I know you know that...
Attachments
Interestingly, the simulation showed unstability on the square wave test. I discovered that C8(5p) the cap // to FB resistor (12.5K) was the culprit.
I seems to be working fine now...
I use 4*49K9 as FB resistor as this was shown to reduce THD significantly on real world experiments. It also should add some capacitance. Luckily, using series-parallel arrangement should attenuate this capacitance effect without affecting THD too much.
I seems to be working fine now...
I use 4*49K9 as FB resistor as this was shown to reduce THD significantly on real world experiments. It also should add some capacitance. Luckily, using series-parallel arrangement should attenuate this capacitance effect without affecting THD too much.
Attachments
Quick update:
A couple of months ago I got rid of the bootstrap caps for the drivers on the QUAD amps:
Before experimenting with a combination of //caps installed in that position again, I decided to try an easiest and more logical test: change the zeners (6V5 or 7V5) from the bootstraps (Q8-9) of the drivers for LEDs. Simulations show that with LEDs, 5V2 is enough for full output. I installed Two White LEDs which gives me 5V5. Easy mod. Instant effect. Gone is the thinness and back is the fullness of the Amnesis sound, chesty instrument voices and more energy on the mid-bass/bass, more "tubey" sound. Of course, it appears as there is less HF detail...I need more listening but first impressions are very obvious.
It is fascinating how the bootstrap works. The LEDs not only set the Vref for the base but also provide a sample of the output signal to that base superimposed to it. I can't but think that LEDs' kind of galvanic insulation between the output and the base of the drivers' bootstrap has some play on the welcome sound effect. It is also possible that LEDs make less HF noise than zeners injected there...
The next obliged test would be replacing the zeners from the MOSFETs for a chain of LEDs...
Cheers and much Love,
M.
A couple of months ago I got rid of the bootstrap caps for the drivers on the QUAD amps:
Before experimenting with a combination of //caps installed in that position again, I decided to try an easiest and more logical test: change the zeners (6V5 or 7V5) from the bootstraps (Q8-9) of the drivers for LEDs. Simulations show that with LEDs, 5V2 is enough for full output. I installed Two White LEDs which gives me 5V5. Easy mod. Instant effect. Gone is the thinness and back is the fullness of the Amnesis sound, chesty instrument voices and more energy on the mid-bass/bass, more "tubey" sound. Of course, it appears as there is less HF detail...I need more listening but first impressions are very obvious.
It is fascinating how the bootstrap works. The LEDs not only set the Vref for the base but also provide a sample of the output signal to that base superimposed to it. I can't but think that LEDs' kind of galvanic insulation between the output and the base of the drivers' bootstrap has some play on the welcome sound effect. It is also possible that LEDs make less HF noise than zeners injected there...
The next obliged test would be replacing the zeners from the MOSFETs for a chain of LEDs...
Cheers and much Love,
M.
Attachments
Last edited:
Update on the QUAD:
After leaving the LED fed bootstrap drivers burn-in for quite a few days, the sound stabilized and got more even through the whole spectrum, all in all better thank the zener based sub-circuit but with still a dry and upfront presentation.
So I went to my original experiment of re-installing the bootstrap caps but this time a combination of (the original) Rubycon ZL 150uF electros, Wima 1uF MKS4 and Arcotronics 0n1/630V said to be polystyrene. Quite bulky but I achieved to install them. The sound went from dry and upfront to liquid and more recessed instantly, a type of sound which I like, but which may change with further burn-in. If this trend persist, the amp will keep sounding more tubey with each mod, though this time I did not perceive any loss in detail and HF, on the contrary, the highs are very palatable.
On other topic, I downloaded Kicad and started practicing with a viable amp board. No need to say that I am not experienced in this matter. I especially have troubles with finding the right connectors and holes, not to mention basic good practice guidelines relating to noise/EMF.
Any advice is welcome. PS: oops! Invalid files... Edit: only images...
Thanks and happy Christ-mass,
M.
After leaving the LED fed bootstrap drivers burn-in for quite a few days, the sound stabilized and got more even through the whole spectrum, all in all better thank the zener based sub-circuit but with still a dry and upfront presentation.
So I went to my original experiment of re-installing the bootstrap caps but this time a combination of (the original) Rubycon ZL 150uF electros, Wima 1uF MKS4 and Arcotronics 0n1/630V said to be polystyrene. Quite bulky but I achieved to install them. The sound went from dry and upfront to liquid and more recessed instantly, a type of sound which I like, but which may change with further burn-in. If this trend persist, the amp will keep sounding more tubey with each mod, though this time I did not perceive any loss in detail and HF, on the contrary, the highs are very palatable.
On other topic, I downloaded Kicad and started practicing with a viable amp board. No need to say that I am not experienced in this matter. I especially have troubles with finding the right connectors and holes, not to mention basic good practice guidelines relating to noise/EMF.
Any advice is welcome.
PS: oops! Invalid files... Edit: only images...Thanks and happy Christ-mass,
M.
Attachments
Last edited:
Happy and Prosperous New Year to All!
I quickly updated my other QUAD's bootstrapped Drivers to the same quimeric paralleled caps but instead of 100nF y used 250nF Russian Military surplus polystyrene caps (best kept secret in audio) for a great effect. As was my first intention, the VAS Bootstrap caps were also upgraded to this wonderful combination of Rubycon ZL 150uF electros + Wima 1uF MKS4 + 250nF Russian Military surplus polystyrene caps, though my VFET amp uses 15uF Vishay film instead of the Wima, now burning-in. This solution is bulky but necessary to achieve the best sound. Now I have ampleur as French say, dynamics, details, separation, relief, textures and transparency. Very tubey, indeed.
What I have learn so far:
1) Thermal Memory Distortion seems to be real in audio reproduction and the lower the better for sound. My Amnesis amp addresses this problem in a satisfactory way.
2) Positive Feedback seems to be very important for good sound, even more important than negative FB.
3) Paralleling is a good solution for signal capacitors.
4) Driver section is of paramount importance in allowing the LTMD input and VAS sections to be fully expressed at the output.
Cheers,
M.
I quickly updated my other QUAD's bootstrapped Drivers to the same quimeric paralleled caps but instead of 100nF y used 250nF Russian Military surplus polystyrene caps (best kept secret in audio) for a great effect. As was my first intention, the VAS Bootstrap caps were also upgraded to this wonderful combination of Rubycon ZL 150uF electros + Wima 1uF MKS4 + 250nF Russian Military surplus polystyrene caps, though my VFET amp uses 15uF Vishay film instead of the Wima, now burning-in. This solution is bulky but necessary to achieve the best sound. Now I have ampleur as French say, dynamics, details, separation, relief, textures and transparency. Very tubey, indeed.
What I have learn so far:
1) Thermal Memory Distortion seems to be real in audio reproduction and the lower the better for sound. My Amnesis amp addresses this problem in a satisfactory way.
2) Positive Feedback seems to be very important for good sound, even more important than negative FB.
3) Paralleling is a good solution for signal capacitors.
4) Driver section is of paramount importance in allowing the LTMD input and VAS sections to be fully expressed at the output.
Cheers,
M.
Guys, I forgot to post a corrected version of the KiCAD board. There was a mistake at the output.
I have trouble finding the icon for "Faston "quick-disconnect" spades" for the power rails, ground and speaker outputs. Anybody?
All the best,
M.
I have trouble finding the icon for "Faston "quick-disconnect" spades" for the power rails, ground and speaker outputs. Anybody?
All the best,
M.
Attachments
Hello guys,
Gongrats for this very interesting topic Maxlorenz. I like your approach to build amplifier with low memmory and thermal distortion.
I consider to try this amplifier with some minor addons. Here is my preliminary layout. I don't has time to check the schematic for mistakes yet, but here is how look like.
Gongrats for this very interesting topic Maxlorenz. I like your approach to build amplifier with low memmory and thermal distortion.
I consider to try this amplifier with some minor addons. Here is my preliminary layout. I don't has time to check the schematic for mistakes yet, but here is how look like.
Attachments
Last edited:
Wow!
I am very impressed...and flatted.
Obviously, I am a beginner in these PCB SW so I cannot contribute much in that topic, layouts, best practices...etc.
Couple of quick comments:
Input: best sonics to me come from Hybrid CFP, meaning one JFET and one BJT (T41-42; T10-T13 in your SCH) either JFET-->BJT or vice versa (not talking about the bootstrapping JFET T8-9), so IF you have J74 P channels, better to use them as input and use other bootstrapping JFETs T8-9 not to waste a good audio JFET. As I said, LSJ289 seem good on paper; see my last schematic.
Output: I have tried D44-45H11 before and I did not like the sound. I use them on power regs with good effect, though.
Edit: please note that Darlington output needs 12V reference as VGS for Mosfet's gates. That would be one 12V zener or one 5 (white) LEDs strip. The resistor feeding them can be increased to 5K(1W), as per simulations.
Feedback: I use 4 good quality resistors in parallel for FR as it is said to lower THD significantly, and so I do not use C12(5p).
I have not experience with you compensation subcircuits, so I cannot help nor advice.
Power: I asume you checked it for offset and other issues. I also asume that you have a good scope cause this amp has positive feedback and likes to oscillate sometimes.
I will continue to check it.
Many thanks for your interest and confidence on my humble adventure.
M.
I am very impressed...and flatted.
Obviously, I am a beginner in these PCB SW so I cannot contribute much in that topic, layouts, best practices...etc.
Couple of quick comments:
Input: best sonics to me come from Hybrid CFP, meaning one JFET and one BJT (T41-42; T10-T13 in your SCH) either JFET-->BJT or vice versa (not talking about the bootstrapping JFET T8-9), so IF you have J74 P channels, better to use them as input and use other bootstrapping JFETs T8-9 not to waste a good audio JFET. As I said, LSJ289 seem good on paper; see my last schematic.
Output: I have tried D44-45H11 before and I did not like the sound. I use them on power regs with good effect, though.
Edit: please note that Darlington output needs 12V reference as VGS for Mosfet's gates. That would be one 12V zener or one 5 (white) LEDs strip. The resistor feeding them can be increased to 5K(1W), as per simulations.
Feedback: I use 4 good quality resistors in parallel for FR as it is said to lower THD significantly, and so I do not use C12(5p).
I have not experience with you compensation subcircuits, so I cannot help nor advice.
Power: I asume you checked it for offset and other issues. I also asume that you have a good scope
cause this amp has positive feedback and likes to oscillate sometimes.I will continue to check it.
Many thanks for your interest and confidence on my humble adventure.
M.
Thank you for the advices maxlorenz.
I edit the input CFP.
I am mostly concerned about amplifier stability. Bode Plot don't look like i want to be.
I attach pictures for comparison how my other amps Bode Plot look like.
Also i will try a little bit different VAS stage like the last attached schematic from the simulator.
If i get any success i will write...
Best Regards
Itco
I edit the input CFP.
I am mostly concerned about amplifier stability. Bode Plot don't look like i want to be.
I attach pictures for comparison how my other amps Bode Plot look like.
Also i will try a little bit different VAS stage like the last attached schematic from the simulator.
If i get any success i will write...
Best Regards
Itco
Attachments
Last edited:
Any advice is welcome.
Everybody here knows I do not know what I am doing...
Which reminds me a doubt I've always had: what is the best orientation of the output inductor w/r its output resistor?
M.
Hey Mauricio, i also don't have much theoretical background. Mostly my experience is practical and empirical. You are already on the good approach of making amplifier with good sonic performance. I will try only to improve stability.
About the paralel resistor with the inductor, i solder it on the oposite side (bottom side) of the pcb.
faston 2,8 mm and 4,8 mm
i once made these footprints and 3d-models for faston spade connectors, see attached. you may have to update the 3d-model path in the footprint editor.
model data from Vogtshop
2,8 mm - Vogt 3780A08.61
and
4,8 mm - Vogt 382505.68
regards, stv
i once made these footprints and 3d-models for faston spade connectors, see attached. you may have to update the 3d-model path in the footprint editor.
model data from Vogtshop
2,8 mm - Vogt 3780A08.61
and
4,8 mm - Vogt 382505.68
regards, stv