I made some mistakes earlier. The node in question is named "+/-Vbiasdrivecasc" on the .asc file that now goes attached. This one with 44V supplies and still no clipping. The .four command says 0.005%THD. I still don't like the noise floor. I will reevaluate my common-base buffer for the differential current output.
Edith: apparently, biasing with power LEDs gives lower THD. Biasing +/-Vbiascasc, the node for the output cascode with one diode and one LED (instead of a string of 3 diodes at 2.1V) set at +/-3.38V gives 0.0024%THD...more experiments to come.
On the real amp I still have the bootstrapped VAS as I respect DX's opinion about sound Q. Can I use baker clamp in my bootsatrapped VAS???
The real amp makes superior sound, IMHO.
Edith: apparently, biasing with power LEDs gives lower THD. Biasing +/-Vbiascasc, the node for the output cascode with one diode and one LED (instead of a string of 3 diodes at 2.1V) set at +/-3.38V gives 0.0024%THD...more experiments to come.
On the real amp I still have the bootstrapped VAS as I respect DX's opinion about sound Q. Can I use baker clamp in my bootsatrapped VAS???
The real amp makes superior sound, IMHO.
Last edited:
Confession: the real amp has solved the oscillation bursts problem with a configuration of cascoded EF VAS plus cascoded drivers exactly as depicted in diagram from post# 101, with 2N5551-->2SC4793 cascoded with BD139 and then drivers MJE15028/15031 cascoded with BD139/149.
When I tried to make the VAS an EF by adding 2N5551 it went into heavy oscillation, even if the RC from the zobel was transplanted to the speaker binding posts (BTW the connectors from the speaker protection circuit, IF outboard, are also useful for that move) with the only difference being that the drivers on that channel are still BD139/149 cascoded with BC546/556!
When I tried to make the VAS an EF by adding 2N5551 it went into heavy oscillation, even if the RC from the zobel was transplanted to the speaker binding posts (BTW the connectors from the speaker protection circuit, IF outboard, are also useful for that move) with the only difference being that the drivers on that channel are still BD139/149 cascoded with BC546/556!
Working Amnesis and sounding good. Tests continue.
I added notes. The Hagerman mod is not yet done.
I think MJE15028/15031 are superior drivers. And they can handle parallel outputs, I imagine.
When I am satisfied I will prototype my 44V version, with all mods from .asc file included: Hagerman VBE; CCS loaded VAS.
There are many mods still to try:
Input-> increase current from 3.2mA to 4 or 5mA. Cascode the current mirror. Add the common-base buffer to the current output from the differential "pair". Transform the J FET cascode element into a compound CFP...
VAS-> independent regulated supply with increased value to "cascode the cascode" as suggested by Peufeu. Compare boostrap VAS with CCS loaded VAS.
Drivers-> find the ideal biasing for the common-base BJTs and find the ideal BJT combination.
Output-> once the previous section get stable, continue to experiment with crazy outputs to find the optimal one that can cope with the increased resolution of the low TMD amp: VFET; CFP; cascoded CFP internal and external; CFP cascoded CFP, internal and external...other that I cannot imagine yet. As you see, lots of work and experimenting.
Cheers,
M.
I added notes. The Hagerman mod is not yet done.
I think MJE15028/15031 are superior drivers. And they can handle parallel outputs, I imagine.
When I am satisfied I will prototype my 44V version, with all mods from .asc file included: Hagerman VBE; CCS loaded VAS.
There are many mods still to try:
Input-> increase current from 3.2mA to 4 or 5mA. Cascode the current mirror. Add the common-base buffer to the current output from the differential "pair". Transform the J FET cascode element into a compound CFP...
VAS-> independent regulated supply with increased value to "cascode the cascode" as suggested by Peufeu. Compare boostrap VAS with CCS loaded VAS.
Drivers-> find the ideal biasing for the common-base BJTs and find the ideal BJT combination.
Output-> once the previous section get stable, continue to experiment with crazy outputs to find the optimal one that can cope with the increased resolution of the low TMD amp: VFET; CFP; cascoded CFP internal and external; CFP cascoded CFP, internal and external...other that I cannot imagine yet. As you see, lots of work and experimenting.
Cheers,
M.
Attachments
What great inventions LEDs are!
The real amp (with 32V supplies) is now stable with LED regulating Vbias for the drivers' cascode, decoupled with 20R-100nF to ground. I could test it until clipping with no oscillation whatsoever. The cascodes for the outputs must be checked for bursts of oscillation: none found.
Cascodying still has secrets for me...the drivers are good with the cascodes' bases at PS+/-3V while the ourtput need the bases of the cascodes at +/-2.1-2.8V from ground...
Now I have the basis for re-exploring the cascode-CFP and then the CFP cascoded-CFP.
The simulations with 44V, LEDs and the buffer for the differential output seem good enough, with extended response and good stability margins (80°; -18db) and is now a worthwhile experiment.
Cheers,
M.
The real amp (with 32V supplies) is now stable with LED regulating Vbias for the drivers' cascode, decoupled with 20R-100nF to ground. I could test it until clipping with no oscillation whatsoever. The cascodes for the outputs must be checked for bursts of oscillation: none found.
Cascodying still has secrets for me...the drivers are good with the cascodes' bases at PS+/-3V while the ourtput need the bases of the cascodes at +/-2.1-2.8V from ground...
Now I have the basis for re-exploring the cascode-CFP and then the CFP cascoded-CFP.
The simulations with 44V, LEDs and the buffer for the differential output seem good enough, with extended response and good stability margins (80°; -18db) and is now a worthwhile experiment.
Cheers,
M.
Attachments
Don't get stuck in the past...
Don't worry about the future...
Just enjoy the present as musical bliss with Amnesis LTMD.
Yeah, I already have the promotional ideas to go into production, hehe.
The difficult part will be making a PCB that is versatile enough to accept many output configurations...
The hard part of this experiment, apart from fixing instability, was powering-off the amps to try new mods...there were afternoons where I simply could not stop listening to it, even knowing it was not working perfectly. Mind you, these were cold-freezing days outside "in the lab".
Cheers,
M.
PS: I am looking for buyers to my other amps.
Don't worry about the future...
Just enjoy the present as musical bliss with Amnesis LTMD.
Yeah, I already have the promotional ideas to go into production, hehe.
The difficult part will be making a PCB that is versatile enough to accept many output configurations...
The hard part of this experiment, apart from fixing instability, was powering-off the amps to try new mods...there were afternoons where I simply could not stop listening to it, even knowing it was not working perfectly.
Mind you, these were cold-freezing days outside "in the lab". Cheers,
M.
PS: I am looking for buyers to my other amps.
I dear. Thanks for your interest.
I have two projects in one so to speak: the real amp is the circuit posted above with bootstrapped VAS. The other amps are only in .asc simulation format and those have active CCS VAS. Many variations have been tested.
The biasing I posted latelly is a modified Hagerman: it could a a diode or a transistor in diode config. It is supposed to be much lineal than the classical.
Be aware that I am no expert and that I progress giving "blind sticks" here and there, trial and error...until I catch the attention of the big dogs.
I dear. Thanks for your interest.
Be aware that I am no expert and that I progress giving "blind sticks" here and there, trial and error...until I catch the attention of the big dogs.
I think : Most of the magasine readers have followed the articles and updated their amp designs along the way.....
As i read your tread you're improving along reading the artikles, probably ending up with a design close to one of the articles series end design ?
-Maybee this wont catch the attention to The Dog until it is capable of grilling a hot dog?
Sinse i am only a smal cat and like to play with sertain things:
What is your ultimate design today ?
What would it bee tomorrow ?
Is the continus variable the way to go ?
-If so look at the BIG Dogs tail and improove it, if not make a BASTARD !
"The Bastard" has to bee better than the other dogs, as you make your BASTARD to bee the ultimate for the living room, for all the cats and dogs thats going too tuch it, smell it and having its drag race with it.
Now its gonna rain cats and dogs! It has to stay firm knowing this is easy handling.
A lot of ppl dont like Bastards, therfore your BASTARD has to give a **** abt what few ppl like or not.
Then your abt ready to pull out the GRILL, make a grill party! Ya know all the dogs has too bee grilled to bee hot dogs thats when the real taste com to life.
After all we are doing this for fun and audio mania
I think : Most of the magasine readers have followed the articles and updated their amp designs along the way.....
As i read your tread you're improving along reading the artikles, probably ending up with a design close to one of the articles series end design ?
-Maybee this wont catch the attention to The Dog until it is capable of grilling a hot dog?
Sinse i am only a smal cat and like to play with sertain things:
What is your ultimate design today ?
What would it bee tomorrow ?
Is the continus variable the way to go ?
-If so look at the BIG Dogs tail and improove it, if not make a BASTARD !
"The Bastard" has to bee better than the other dogs, as you make your BASTARD to bee the ultimate for the living room, for all the cats and dogs thats going too tuch it, smell it and having its drag race with it.
Now its gonna rain cats and dogs! It has to stay firm knowing this is easy handling.
A lot of ppl dont like Bastards, therfore your BASTARD has to give a **** abt what few ppl like or not.
Then your abt ready to pull out the GRILL, make a grill party! Ya know all the dogs has too bee grilled to bee hot dogs thats when the real taste com to life.
After all we are doing this for fun and audio mania
Ha, ha! I like your style man.
Yes a bastard it is but it is my bastard. And it will become even more bastardized as I gain experience and skill. For now, second day of listening test on a humble system, it is stable and stunning. Tested with chamber music, not girl and guitar music.
The ultimate design goal at present, as I said before, is to find the most transparent output section to exploit the LTMD benefits of input and VAS.
Speaking of which, I will have to test cascoded-cascoded-VAS, as Peufeu recommended.
Cheers,
M.
In case you don't want to deal with potential oscillation problems, here a revisited version of my own splendid Low TMD amp with Tokin 2SK180 or THS51 VFET. This is the version with all-cascoded.
The VFET can be replaced by your prefered depletion Mosfet, but that would require adaptation of the Vbias, probably.
Even an enhancement Mosfet would do as cascode element, or a power BJT provided a change in biasing...trust me these Tokin amps are great.
This version is not tested yet, but will soon be, as I have the biggest heatsinks I could get...
The VFET can be replaced by your prefered depletion Mosfet, but that would require adaptation of the Vbias, probably.
Even an enhancement Mosfet would do as cascode element, or a power BJT provided a change in biasing...trust me these Tokin amps are great.
This version is not tested yet, but will soon be, as I have the biggest heatsinks I could get...
Attachments
Last edited:
This would be my Hagerman mod to the bias of the Blame PCB.
The extra compensation R* must be selected given the chosen BJT and its change in VBE from two selected points in the VBE v/s Ic chart, for example from 10mA to 100mA, as explained by the author.
I hope I got it well.
The extra compensation R* must be selected given the chosen BJT and its change in VBE from two selected points in the VBE v/s Ic chart, for example from 10mA to 100mA, as explained by the author.
I hope I got it well.
Attachments
Amplifier runs stable for 10days now. Tested on two systems.
I will not sing the merits of the circuit. I've done that enough. Who wants to explore it, welcome.
I am revisiting the diffamp for my VFET single ended as this is a tricky balance act. I wil post the present circuit. It gives enough un-clipping gain since the VFET gives a gain of 10.
I will not sing the merits of the circuit. I've done that enough. Who wants to explore it, welcome.
I am revisiting the diffamp for my VFET single ended as this is a tricky balance act. I wil post the present circuit. It gives enough un-clipping gain since the VFET gives a gain of 10.
Attachments
I discovered that one has to do frequent updates of LTspice for it to work properly...
Anyway, here go some variations of the simple cascoded EF, with inverted R for the input CFP, and with LED instead of R based Vbias for the VAS cascode which gives lower THD. Good stability margins, decent THD, though climbing at higher armonics...luckily they go down after 30K or so.
Remember: our goal is NOT THD...
It is very entertaining to look at the curves of the different sections and of the different cascodes...the Lavardin bias for the output cascode is very clever, but it still puzless mes...
Anyway, here go some variations of the simple cascoded EF, with inverted R for the input CFP, and with LED instead of R based Vbias for the VAS cascode which gives lower THD. Good stability margins, decent THD, though climbing at higher armonics...luckily they go down after 30K or so.
Remember: our goal is NOT THD...
It is very entertaining to look at the curves of the different sections and of the different cascodes...the Lavardin bias for the output cascode is very clever, but it still puzless mes...
I already started the built of the 44V version, as the 32V betters by far my best Current Feedback Amplifier, sound-wise. 
You don't know what you are missing until you listen to it....
Actual perceived noise floor is lower than simulations seem to show.
After the debugging and optimization period of the 44V version, I will order some HiQ PCBs. I will offer some to the mates that showed more interest.
Wish me luck.
M.
You don't know what you are missing until you listen to it....
Actual perceived noise floor is lower than simulations seem to show.
After the debugging and optimization period of the 44V version, I will order some HiQ PCBs. I will offer some to the mates that showed more interest.
Wish me luck.
M.