The AMNESIS amp: a good amplifier, like a gentleman, has no memory.

This is better.
 

Attachments

  • AMPC2002 CASCODEDVAS2 SZIKLAI HIGHV.asc
    14.1 KB · Views: 124
  • AMPSLAB C200.2 SZIKLAI LTMD HIGH V.png
    AMPSLAB C200.2 SZIKLAI LTMD HIGH V.png
    87.5 KB · Views: 314
This is the beauty of the bootstrap cap for the drivers: look how the signal rests between 0 and (+/-) 50V, for 32V supplies.
Apparently, this increases driver section input impedance...anybody?

Cheers,
M.

PS: for 32V supplies, R55 better be lowered from 12K5 to 10K, to reduce gain...
 

Attachments

  • AMNESIS SZIKLAI OUTPUT 32V3.png
    AMNESIS SZIKLAI OUTPUT 32V3.png
    63.8 KB · Views: 170
Last edited:
The Sziklai output into de AMNESIS (the chaotic one) experiment was rather interesting...

The amp showed signs of oscillation: heatsink for the positive driver was hot; some scratching noises at volume peaks. The bias was stable though, at +/-50mA. My scope is bad so I do not have more details/insight.
Anyway, I tried many tricks and made some re-readings, including the excellent http://hifisonix.com/ovation-e-amp/ site, which I understand is dear Bonsai's site. These investigations led to improvement and reducing of the scratching noises (that afterwards I discovered were at least due to the messy build of the input section :p ) and allowed for sound comparison.
That particular QUAD Amnesis, with modded SMPS, is open and detailed but rather flat and with "thin" bass. The Sziklai ouput has rounder and darker bass/midbass but has recessed HF/details. That made me wonder...the DX-HRII and the C200 Ampslab have both a 1000uF feedback cap (with a similar sound presentation of "good bass/recessed highs"), plus the Ovation-E amp also uses it and Bonsai explains his choice with selected references. The AMNESIS retained the 220uF RubyconZL FB from Uncle Charlie's Blame. I decided to swap it for a 560uF RubyconZLH and, imagine my surprise, the bass/midbass of that particular AMNESIS improved a lot to a point that I cannot longer distinguish from that with the Sziklai output. I chose 560uF because the simulations showed equal result compared to 1000uF.

I won't detail every mod and permutation here but, suffise to say that, after the realization that the hot driver heatsink was not improving, I went back to the QUAD but retained some of the mods, including a revisiting of the TMC which seems to offer better stability to this particular amp variety. Now I will have to make prolonged audition to evaluate if the higher value FB cap has negative effect on the other AMNESIS qualities.

Another mod that improved the Sziklai experiment scratching was returning (was that way before placing the modules inside a "decent" metal box) to attaching the Zobel directly to the T-piece of the center power ground, also a Hifisonix tip. That reduces the possibility of HF injection near the input section.

On another topic, I always forget to point you to this site:

https://sites.google.com/site/fabaudio/designdetails
Interesting, isn't it?
The VAS is alike my preference for the Ampslab C200.2 LTMD.

Cheers,
M.
 

Attachments

  • AMNESIS hybridCFP in TMC QUAD.png
    AMNESIS hybridCFP in TMC QUAD.png
    39.8 KB · Views: 195
  • AMNESIS hybrid INPUT TMC QUAD.asc
    17.8 KB · Views: 97
Quick update.

Yes, increasing FB cap to 560uF in that particular QUAD Amnesis did improve bass/midbass, without reducing micro/macro-dynamics and HF detail and insight, plus midrange became more textured. I do not feel like swapping that cap in my other amps which have more balanced freq. response.

I returned to my new Darlington Out (the pro-PCB one) and noticed that, though it sounds great, current bias jumps from +/-40mA to +/- 100mA when trimming slowly: a sign of probable (parasitic) oscillation of low grade. From hints of the recent past, I tried small caps on the positive side drivers (opposite VAS transistor) from base to collector. One module needed 240p and the other needed 560p (the next I had on hand) to keep on a steady 50mA bias current after torture. I imagine R16 + C20 (please see attached schem.) form a low pass filter for re-entrant energies. Increasing C (and R) value guided by a scope reading of the oscillation frequency should be a more "scientific" way to calculate the values. I hope some wise man can enlighten me if my hypothesis is incorrect.

I keep testing/simulating alternatives to increase stability margins. The schematics presents a crazy idea that seems to increase both closed loop frequency response extension and stability. Please see attachments.

Finally, and my intention is not to overload the readers, but I found many interesting articles on the Audiophile Index I posted before. Here, a sample from a Cascoded Sziklai output amp (alike my version of the SONY TA5650) in which the author introduces a zener diode after the pre-driver, to avoid saturation of the driver. The description of the realization of the amp is very complete indeed.

http://6bm8.lab.free.fr/Documentations/Revues/Audiophile/1977-1988/03/AMPLI/AMPLI.html
Cheers and Happy Christ-mass.
M.
 

Attachments

  • AMNESIS ALTERN COMP HYBRID INPUT CASCODED DARLINGTON BOOTSTRAPPED OUTPUT V2.asc
    16.7 KB · Views: 83
  • AMNESIS DARLINGTON out ALTERNATE COMP2.png
    AMNESIS DARLINGTON out ALTERNATE COMP2.png
    37.1 KB · Views: 194
  • AMNESIS DARLINGTON out ALTERNATE COMP2 BODE.png
    AMNESIS DARLINGTON out ALTERNATE COMP2 BODE.png
    11.9 KB · Views: 184
  • CASCODED SZIKLAI AUDIOPHILE.jpg
    CASCODED SZIKLAI AUDIOPHILE.jpg
    80.2 KB · Views: 197
Last edited:
Hi guys, hope you're doing well.

Regarding the Bootstrapped Darlington output, I was forced to increase the value of the Zener from 8V2 to 12V in order to avoid premature clipping for that particular configuration. But there is another way, and I present it here the comparison in the attachment. By referencing the Vb for the bootstrapping of the driver to the power rails (instead to the Vc), with bootstrap-caps, simulations show decent clipping behavior with 3V2 to 3V6 Vb (for example 2X Red LEDs in series) , which allows for reduction to a 8V2 Zener diode to the gate of the Mosfet. Look at the reduction of Vce for the drivers and for the output T. That way junctions become thinner and charging/discharging processes should be faster. Hopefully, there shall be an improvement in sound with little effort. Simulations show comparable results especially in stability margins.

Cheers,
M.
 

Attachments

  • AMNESIS DARLINGTON ALTERNATIVE OUTPUT COMPARISON.png
    AMNESIS DARLINGTON ALTERNATIVE OUTPUT COMPARISON.png
    32.1 KB · Views: 116
  • AMNESIS ALTERN COMP HYBRID INPUT CASCODED DARLINGTON BOOTSTRAPPED OUTPUT V3.asc
    16.9 KB · Views: 69
Hi again,

The other advantage of the latest bootstrapped driver configuration relates to using VFET (best option) instead of MOSFET at the output: it should allow for more liberal use of VFETs since one can choose units with a lower (roughly 4V5 to 8V) VGS at your preferred current bias :cool: without reduction of the power output ceiling.

Cheers,
M.
PS: prosperous and abundant NY.
PS2: imagine your ideal World!
 

Attachments

  • AMNESIS DARLINGTON ALTERNATIVE OUTPUT COMPARISON VFET.png
    AMNESIS DARLINGTON ALTERNATIVE OUTPUT COMPARISON VFET.png
    34.1 KB · Views: 96
Hi guys,
I hope you are all doing perfectly well...

Instead of an update of my adventures, I want to quickly post this old trick that I missed from dear Hans Polak: a limiting sub-circuit ( 33V Zener in series with 1N4148 diode from nodes to ground) to avoid the VAS from overloading the output.
I hope I interpreted it well...
Simulations show +/-35V output, with a soft clipping of the negative peak at least, instead of >36V...

Cheers,
M.
 

Attachments

  • AMNESIS overload CONTROL.png
    AMNESIS overload CONTROL.png
    30 KB · Views: 123
Hi guys,

I've been busy with other fun projects and I posponed update.
I am in the process of modifying my Amnesis amps to the latest interesting findings and lessons learned:

1) 1nF ceramic cap from the ring of the "ground/return" of the RCA input connector to chassis (Thanks Bonsai!) like I've seen on Phono input connectors. Silence with the tweeter "inside" the ear... :D

2) Zobel to Star or T-Ground. Thanks again Bonsai.

3) 560pF Silver-Mica (I suppose ceramic could work also) cap from base to collector of the positive driver. If I had my scope I could calculate better the value.

4) TMC plus that other mod with the series caps//FB resistor, with a parallel R to ground, whose name, if it has one, I do not know. See attached schematic.

5) Increase of the FB cap from 220 to 560uF Rubycon ZLH. I suppose 470uF would also sound good.

My VFET amp now let my DAC fully express, meaning my DAC is better than I thought.

Well, to save time I decided to saw windows on the base plate of the case to easier access to amplifier PCBs: 4 accidents! :p
Firstly, I am becoming faster at detecting and correcting my own faults :cool::LOL:
Secondly, after an accidental hit to the heatsink (HS), that channel suffered all the sings of heavy oscillation: I discovered that my sawing disconnected one of the 47nF cap from HS to ground (!!!), that I remember perhaps dear Symon recommended long ago, and which, after re-soldering, protected the amp from reproducing the hitting-oscillation! That made wondering and remembering our conversation about separate HS between output units, but this time not because of thermic but for capacitive considerations! Maybe transistor (read: transistor/Mosfet/FET) to HS AND transistor to transistor capacitance is/are the main cause(s) for the Bootstrapped outputs tendency to oscillate... so next I must try putting the BJTs, which dissipate low power (mW range at clipping, with 60mA bias) , on its own HS, separated from the main HS but connected remotelly to ground.
What do you guys think?

Pray and prepare.
M.
 

Attachments

  • AMNESIS DARLINGTON out ALTERNATE COMP2.png
    AMNESIS DARLINGTON out ALTERNATE COMP2.png
    42.9 KB · Views: 117
Last edited:
Hi Max,

after enjoying my Amnesis Cascoded Darlington 2 for many months
without any stability problems I think it is time for an update
to try to further improve sound quality.
My present design is very close to the schematic shown in post 727
(cascoded darlington) but with supply voltage +-34V insteadt of +-44V

The schematic in latest post 790 shows the following changes:

output section
emitter resistors R29/R30 connected now to the emitter of the
output transistors (TTC2/TTA2).
2.2R resistors replaced by a wire

driver section bootstrapped
base of Q3 connected via R31=15K and R6=5k to positive supply
bootstrapp Cap C22=100u to output
base of Q5 connected via R47=15K and R46=5k to negative supply
bootstrapp Cap C27=100u to output Q5

Changes marked with orange circle
feedback network
C11=20p and C24=20p R20=2k to GND in parallel to R55=12.5k

VAS section
C25=82p C10=240p R37=1.5k to output

Can you tell me what are the reason for these proposed changes
(improved sound quality, increased stability, else...)

Which of these changes do you think are most important and should be tried first.

Best Regards
 
Hi dear JOSI1,
How do you do?
I hope all is well.
after enjoying my Amnesis Cascoded Darlington 2 for many months
without any stability problems I think it is time for an update
to try to further improve sound quality.
My present design is very close to the schematic shown in post 727
(cascoded darlington) but with supply voltage +-34V insteadt of +-44V

The schematic in latest post 790 shows the following changes:

output section
emitter resistors R29/R30 connected now to the emitter of the
output transistors (TTC2/TTA2).
2.2R resistors replaced by a wire
A more classical (maybe faster) way to connect the Darlington. I did not experience significant changes.
driver section bootstrapped
base of Q3 connected via R31=15K and R6=5k to positive supply
bootstrapp Cap C22=100u to output
base of Q5 connected via R47=15K and R46=5k to negative supply
bootstrapp Cap C27=100u to output Q5
This configuration I have not tested yet. Note that this one can use a 3.2V to 3.5V reference to the bootstrapped driver's base and a 8V2 zener for the Mosfet. The idea is to have the lowest possible VCollector for both the ouput and driver BJTs, or the VFET if one choses to use that optimal option.
The potential benefits are:
1)Increased Driver section input impedance from the boostrapped driver BJT with split Rs and Cap C27(may be higher than 100uF), which can be //with a good film. This increased input impedance should make the VAS happier and it is needed to full swing of the output or avoiding premature clipping...
2)Reduced heat stress for the output BJT: one can try the "independent Heat-Sink" idea to eliminate parasitic BJT-HS-BJT capacitance, potantially reducing oscillation risk.
3)I have noticed easier, more dynamic and transparent sound every time I reduce VCE., through bootstrapping.
I attach a diagram with working voltages to show VCEs...

Changes marked with orange circle
feedback network
C11=20p and C24=20p R20=2k to GND in parallel to R55=12.5k

VAS section
C25=82p C10=240p R37=1.5k to output
This one I have tried in 3 of my 5 Amnesis with good effect. Simulations show increased HF extension with better stability margins, from my humble interpretation of the graphs...I kind of find it sounds a bit better this way, but I cannot be 100% sure as my subjetive listening is getting better also...
Remember the FB cap (I remember you use Black Gate?) may be increased to good quality 470uF or 560uF (or higher) for better bass, after the previous mod and it still should be stable.
Can you tell me what are the reason for these proposed changes
(improved sound quality, increased stability, else...)

Which of these changes do you think are most important and should be tried first.

Best Regards

And best wishes to you,
M.
 

Attachments

  • AMNESIS DARLINGTON out ALTERNATE COMP2.png
    AMNESIS DARLINGTON out ALTERNATE COMP2.png
    45.1 KB · Views: 84
Hi guys,

After long months (6 to be exact) I found the enthousiasm to do the mod I posted above (#786) regarding the bootstrap of the driver section, which is a very simple mod but it took me some time as I used a DIY copper plate PCB and space was scarce. For a couple of resistors and capacitors (I didn't have the time to do the zener for the Mosfets from 12V to 8V2) the change in sound is significant: it reminds me of the QUAD with capacitor-boostraps for drivers + outputs. Very promising, with bigger depth and more"wooden box" sound from stringed instruments, more energy on midrange and midbass. I have to perform more tests with other speakers. The V reference for the bootstrapped driver went from 5V4 (2X white LEDs) to 3V4 (one white LED +one 1N4007 diode)...

On another note, the cascoded output for the Ampslab C200 I posted above is alike the output that Hephaïstos preferred...

Cheers,
M.
 

Attachments

  • AMNESIS DARLINGTON ALTERNATIVE OUTPUT COMPARISON.png
    AMNESIS DARLINGTON ALTERNATIVE OUTPUT COMPARISON.png
    32.1 KB · Views: 115
  • HEPHAISTOS SORTIE CASCODE.png
    HEPHAISTOS SORTIE CASCODE.png
    461.3 KB · Views: 110
OK, guys, this is the typical "why didn't I make this mod earlier?" kind of mod...
I moved the modified Bootstrapped Darlington Amnesis I showed before to the "main system n°2" which has ECDesigns Mosaic UV DAC plus the big Tannoy Autograph Clones (Beyma Coax has 98db sens. before horn amplification) and the music is flowing with big soundstage (go figure) ansd lots of bass/midbass and dynamics, but more importantly it has a refinement and depth which I didn't know it lacked before...I removed the QUAD Amnesis that was powering that system and trasnformed it into the Bootstrapped Darlington, which was easier that I preview, and that amp has the bootstrapped capacitors as a //150uF Rubycon ZL//1uF Red Wima film//270nF Russian Polysttyrene on its own daughter-board. That combination is very transparent and the sound is not overwelming like the previous amp so this confirms that the coloration comes from the quality of the bootstrapping caps! That second amp has 32V supplies and I attach diagrams for the vales required and measured. That amp has few minutes on it but it sounds very refined, soft, deep and involving. I had the same experience as with the first one as both original configurations sounded flat and up-front by-comparison, which were qualities not considered to be in...quite shocking, as the presentation of the new and "official configuration" is extreme in its musical qualities, with instant emotional connection and relaxation into the experience.
I had a good time re-designing the PCB, but my previous one (first ever) is easy to mod to accomodate the new configuration.
I will mod my other amps and I will attack again the Sziklai output but this time with the new boostrapped driver section.


Cheers,
M.
 

Attachments

  • AMNESIS BOOTSTRAPPED DARLINGTON PCB_3D.png
    AMNESIS BOOTSTRAPPED DARLINGTON PCB_3D.png
    83 KB · Views: 115
  • AMNESIS BOOTSTRAPPED DARLINGTON PCB_3D2.png
    AMNESIS BOOTSTRAPPED DARLINGTON PCB_3D2.png
    69.1 KB · Views: 105
  • AMNESIS SZIKLAI OUTPUT 32V2.png
    AMNESIS SZIKLAI OUTPUT 32V2.png
    33.3 KB · Views: 120
  • AMNESIS DARLINGTON out ALTERNATE COMP2.png
    AMNESIS DARLINGTON out ALTERNATE COMP2.png
    38.8 KB · Views: 107
Last edited:
I removed the QUAD Amnesis that was powering that system and transformed it into the Bootstrapped Darlington...and that amp has the bootstrapped capacitors as a //150uF Rubycon ZL//1uF Red Wima film//250nF Russian Polystyrene on its own daughter-board. That combination is very transparent and the sound is not overwhelming like the previous amp so this confirms that the coloration comes from the quality of the bootstrapping caps! That second amp has 32V supplies and I attach diagrams for the values required and measured. That amp has few minutes on it but it sounds very refined, soft, deep and involving
Liquid is the description I was looking for, as in music as a fluid substance instead as in music as a pressure perturbation of the medium...

Here, a pic of the perfect capacitor combination to produce this result...I will experiment with Electro//Wima film alone as it is significantly smaller. Both film are 250V caps, but there are lower V and smaller, as you know.

Cheers,
M
 

Attachments

  • CAPACITOR BOARD.jpg
    CAPACITOR BOARD.jpg
    26.8 KB · Views: 61
Last edited:
I removed the QUAD Amnesis that was powering that system and transformed it into the Bootstrapped Darlington...and that amp has the bootstrapped capacitors as a //150uF Rubycon ZL//1uF Red Wima film//250nF Russian Polystyrene on its own daughter-board. That combination is very transparent and the sound is not overwhelming like the previous amp so this confirms that the coloration comes from the quality of the bootstrapping caps! That second amp has 32V supplies and I attach diagrams for the values required and measured. That amp sounds very refined, soft, deep and involving.

Liquid is the description I was looking for, as in music as a fluid substance instead as in music as a pressure perturbation of the medium...

Cheers,
M
So we have two amps modded with the new bootstrap configuration : first was Bootstrapped Darlington (amp 3 henceforth; chronologically) which sounded very "powerful" and bassy, though refined, and a QUAD transformed into the new Bootstrapped Darlington (amp1; chronologically) which sounded "liquid" and also very refined, though it sounds leaner, as if a layer of fat was removed from it. The differences of them were 44V supplies with remaining 12V zeners for the Mosfet's Gates (higher, 8.2Vc for the Darlington's Driver and Output) and simple ElCap for the Bootstrap Driver arrangement for Amp 3 compared to 32V supplies with 8V2 zeners for the Mosfet's Gates (lower, 4.2-4.4Vc for the Darlington's Driver and Output) and parallel ElCap//Film for the Bootstrap Driver arrangement for Amp 1.

I also used diverse Vrefs for the Bootstrap Driver arrangement to the lowest possible or reasonable values, without achieving premature clipping: 3V4 for Amp 3 and 2V75 for Amp 1, accounting for the difference in supply V, as depicted in the attached diagrams.

The causes for the different sound presentations could be speculated to be due to the difference in Vc for the Darlington's Driver and Output or due to Caps for the Bootstrap Driver arrangement, or due to both, or due to neither of them...

In order to study the cause or reason of the differences heard, I lowered first the Zener for the Mosfet's Gates to 8V2 for Amp 3 in order to match as close as possible that aspect for both amps...
Guys, I think I confirmed my suspiction about the way to achieve liquidity (liquidity is the word in vogue these days, hehe) in a Solid State amp, comparable to that of Valve amps! Amp 3 (44V supplies) has now quite signifficant liquid presentation to it.
Or better said, the Low VC supplied to the BJTs of the Bootstrapped Darlington Output configuration Allows the Liquidity achievable for the LTMD Input and VAS sections to be expressed...
I won't dare to explain why this is so as I have a very limited understanding of the Physics involved but I presume it is related to the width of the CB and BE Junctions... 🙏

On the not so positive side, both amps have now a leaner sound (less prominent in Amp 3) though they reach deep notes effortlessly, but I miss the previous "fullness", so I will experiment with Feed-Back capacitor (both these amps have the original 220uF caps) values, and other...

For the 32V supplies, simulations show that Vref for the Mosfet's Gate can be as low as 7V3, which can be a 7V5 Zener, without premature clipping...in fact it will Soft-clip at +/-25VAC output...

Cheers,
M.
 

Attachments

  • AMNESIS DARLINGTON out ALTERNATE COMP2.png
    AMNESIS DARLINGTON out ALTERNATE COMP2.png
    39.3 KB · Views: 56
Last edited:
This one also works but does not add the bootstrap caps' signal to the base...Post#787
Guys, I think I confirmed my suspiction about the way to achieve liquidity (liquidity is the word in vogue these days, hehe) in a Solid State amp, comparable to that of Valve amps! Amp 3 (44V supplies) has now quite signifficant liquid presentation to it.
Or better said, the Low VC supplied to the BJTs of the Bootstrapped Darlington Output configuration Allows the Liquidity achievable for the LTMD Input and VAS sections to be expressed...
I think this one is worthy of experimentation too, as it avoids C on the "signal path"...
 

Attachments

  • AMNESIS DARLINGTON ALTERNATIVE OUTPUT COMPARISON2.png
    AMNESIS DARLINGTON ALTERNATIVE OUTPUT COMPARISON2.png
    40.8 KB · Views: 66
On the not so positive side, both amps have now a leaner sound (less prominent in Amp 3) though they reach deep notes effortlessly, but I miss the previous "fullness", so I will experiment with Feed-Back capacitor (both these amps have the original 220uF caps) values, and other...
I was premature in my opinion on this topic as as "burn-in" progressed bass and "fulness" improved to a level where I consider it to be very satisfactory, deep and articulated. I am using 3.5%Ag solder which may explain the observed effect...

C20 from base of positive side driver to collector of its bootstrapping BJT (220p to 560p; halfway to highest unharming value) is needed in some square wave test simulations of different versions (I know, too many) to avoid ringing and possibly avoid oscillations...

Cheers,
M.

PS: Amp1 now sounds more liquid than AmnesisVFET...
 

Attachments

  • AMNESIS BOOTSTRAP OUTV3.png
    AMNESIS BOOTSTRAP OUTV3.png
    43.2 KB · Views: 61
  • AMNESIS ALTERN COMP HYBRID INPUT CASCODED DARLINGTON BOOTSTRAPPED OUTPUT V3.asc
    17.5 KB · Views: 52
Last edited:
Hi dear all, more funny stories from your humble narrator...

1) First I decided to mod the last amp (amp5) which I built on my first own Pro PCB to the last "drivers bootstrapping with caps", to evaluate which one sounds better, Cap-to-output or Cap-to-power rail...
First module went OK, and just when I was thinking -"Am I not the best amp designer or what???" :joker: the other channel, the unmodded one, after many months of faithful work, presented an intermitent catastrophic failure!!! Near power-rail (28VDC) at output...I checked everything, active and passive, and all looked OK. I remembered that channel gave me troubles at build-time, which I (wrongly; I know understand) attributed to gross oscillation. I decided to go from Miller to TMC hoping to solve the issue, when I discovered to my endless Shame that one of the leads of the 47R inter-BJT VAS resistor (R50) was totally loooose, not a cold solder but no solder, or better solder away from the soldering pad! So the amp was depending only on physical contact from lead to pad all these months...my fault was that I chose a "small R" footprint from KiCad, with a minuscule soldering pad, when designing the PCB and the bad solder totally escaped me during visual inspection...
I soldered the d**n thing and it works OK...Oh Shame, oh shame! :guilty:

2) As I de-soldered the output BJTs, I profited to test my hypothesis that the tendency to oscillate in this type of amps is (at least in part) depending on transistor-to-transitor parasitic capacitance, through common heat-sink: I soldered the big BJT in upright position without HS for now. They are not even warm to the hand and SOA for the 5200/1943 pair at 4V VCE/25°C is Yuge! :happy1:
I attached a pic (sorry for the quality) and you can see that even a small HS, like those rescued from damaged computer PSs would be good here.

3) Following with optimisation of the output section and given the the driver section has also low VCE and low power emission, I re-visited the BE resistor (R29-R30) which I considered to high. I changed them from 78 to 47R. By this time I stumbled upon a thread from dear Peufeu (very technical) where he proposed an interesting idea: mixing the Emitter-follower arrangement (like we had before) with the Darlington arrangement (present), which he found attractive. Please see his explanation and measurements (Post#72) for which I am making now sound test (39R instead of 47R) and for which I attach a diagram. That should pass >20mA and keep drivers on Class A. First listening is promissing, as I sense more textured and "warm" midrange and it seems as if a HF "glare", that I did not notice before, has dissapeared...

Edit: here is Peufeu's thread and no, I have not yet read it all...sorry.

https://www.diyaudio.com/community/...put-stage-measurements-shootout.374367/page-4

4) More optimisation: VAS.
Years ago...I decide to keep 4VDC (2* yellow LEDs) on the VAS arrangement as I had also a catastrophic failure with 3.2V...The minimum successful experiment was 3.6VDC but then I was using BC546 or something like that. Now, simulations with MPSA18, the preferred unit as VAS transistor, show good behavior even at 2.5VDC. One white LED at 2V75 should be a reasonable option to reduce its VCE. That experiment will be next. :xfingers:
Wish me luck as my scope is out of function and I have to rely on indirect signs...

So, optimisation of output requires reducing transistor VCE to their lowest usable and reliable value, which means the width of the BC zone/junction is minimal, or that the control-module is comparable to the power-module, as dear Hephaïstos would have said...

Cheers,
M

PS: I wish you all a happy Lion's Gate!
 

Attachments

  • AMNESIS DARLINGTON BOOTSTRAPPED OUTPUT V5.asc
    17.7 KB · Views: 49
  • OUTPUT 2022-08-02.jpg
    OUTPUT 2022-08-02.jpg
    44.1 KB · Views: 67
  • AMNESIS BOOTSTRAP OUTV5.png
    AMNESIS BOOTSTRAP OUTV5.png
    43.3 KB · Views: 71
Last edited:
...It works, but with the 10/15 Ohm stopper gate resistors for the MOSFETs it oscillated beautifully at +/-12.5MHz, so in my despair, I increased R43 (N channel stooper) to 330R and that stopped the overt oscillation. I may have to increase also the other one for a small tendency to widen the scope line at negative power peaks...and also the perfect (lowest) value must be explored. ...
Hi Max

Earlier in this thread, increasing the gate resistor R43 to 330R was necessary to stop oscillations. At that time, the gates were bootstrapped.
In the latest designs I've seen the gates to the MOSFETs are not bootstrapped and the gate resistors, R43, 44, are a low 39R. At such low values for gate stoppers I'd expect MOSFET self-oscillations in the 100 MHz to 200 MHz range. Is your scope fast enough to detect for potential oscillations or bursts this high? Or if not a fast scope, some other method?

It's normal to see 180R to 470R gate stoppers. Assuming that indeed there are no self-oscillations, what do you suppose is the reason this design can get away with such low values of 39R?