Go Back   Home > Forums > >
Home Forums Rules Articles diyAudio Store Blogs Gallery Wiki Register Donations FAQ Calendar Search Today's Posts Mark Forums Read

Solid State Talk all about solid state amplification.

The AMNESIS amp: a good amplifier, like a gentleman, has no memory.
The AMNESIS amp: a good amplifier, like a gentleman, has no memory.
Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 7th February 2019, 06:23 PM   #481
Hans Polak is online now Hans Polak  Netherlands
diyAudio Member
 
Join Date: Mar 2005
Location: Blaricum
Max,

I had a closer look at the cascoded OPS.
I compared with a non cascoded version, using for as far as possible the same components, still using the 50Watt@20Khz test as before.
First of all I tried to find the optimal Quiescent current. For both versions this was at 125mA, giving resp 0.017%/Cascode and 0.022%/Straight.
When going to 75mA Quiescent current, THD20K became resp 0,047% and 0,037% or +8dB and +4.5db. The cascoded version proves to be a bit more sensitive for the right adjustment.

Next I measured the input impedance of both versions.
The straight version has a slightly higher LF impedance at 115KOhm versus 105KOhm for the cascoded version.
However as can be seen in the image below in red, the cascoded version has a significantly higher BW, which is to be preferred.

I could not find any effect caused by C36 and R39, which can be both omitted to my opinion.
As a conclusion: Cascoding the driver of the OPS, results in a better HF load for the VAS, which in turn leads to less distortion helping the Amnesis to become a top performer.



Hans
Attached Images
File Type: jpg OPS.jpg (441.9 KB, 127 views)
  Reply With Quote
Old 9th February 2019, 02:23 PM   #482
maxlorenz is offline maxlorenz  Chile
diyAudio Member
 
maxlorenz's Avatar
 
Join Date: Oct 2003
Location: osorno , Chile
Quote:
Originally Posted by Hans Polak View Post
Max,

I had a closer look at the cascoded OPS.
I compared with a non cascoded version, using for as far as possible the same components, still using the 50Watt@20Khz test as before.
First of all I tried to find the optimal Quiescent current. For both versions this was at 125mA, giving resp 0.017%/Cascode and 0.022%/Straight.

Luckily, I have bigger heatsinks!

When going to 75mA Quiescent current, THD20K became resp 0,047% and 0,037% or +8dB and +4.5db. The cascoded version proves to be a bit more sensitive for the right adjustment.

Next I measured the input impedance of both versions.
The straight version has a slightly higher LF impedance at 115KOhm versus 105KOhm for the cascoded version.
However as can be seen in the image below in red, the cascoded version has a significantly higher BW, which is to be preferred.

I could not find any effect caused by C36 and R39, which can be both omitted to my opinion.

Those proved to be useful to tame oscillation at one stage. Can be viewed as optional, probably...

As a conclusion: Cascoding the driver of the OPS, results in a better HF load for the VAS, which in turn leads to less distortion helping the Amnesis to become a top performer.

Hans
Dear Hans,

I don't know how to thank you for all your interest and energy invested in our little project. You find me in a personal low, though, with much real life interference, exhausted and with only one week ahead to summer vacation...it is hard for me to envisage complex re-organization of the circuit. For that, I simulated a quick and dirty version, with errors as you pointed out, and, as now I realize, with a fishy behavior on square wave test...trying to debug it proved hard.

I fully understand all your advices and the rationale behind them but I find it hard at the moment to grab the energy and time required to proceed to the orderly modification and testing of the circuit, especially when I consider the present state of both versions to be fully satisfactory in musical terms (I may be positively inclined to a certain amount of distortion and "dirt" ) and that my primordial concern was not THD in the first place...

I was hopping that the "first edition" of the amp gained enough acceptance to motivate enough people to investigate further and develop a more refined version, which I confess, may be beyond my limited capacity...I feel already payed by having a working amp, which sounds nice in addition.

In the following days I plan to do easy tries:
1) Look for my "improved current sources" article and try to find the best CCS I can for the input section.
2) Try the EF VAS again...
3) Cascode the bootstrap for the bootstrapped input section

I am glad that you analyzed the output section also. The present configuration of the "QUAD" was selected only in sonic terms as it seems not to limit the benefits of the putative LTMD organization of the previous sections, in the aspects of transparency and dynamics, which are my main goals, as I always repeat...

If you have time, can you analyze the Darlington output on post #468, which seems dearest to JOSI1?

Gratefully yours,
M.
Attached Files
File Type: asc AMNESIS ALLBJT CASCODED INPUT (1).asc (18.3 KB, 12 views)

Last edited by maxlorenz; 9th February 2019 at 02:40 PM.
  Reply With Quote
Old 9th February 2019, 07:01 PM   #483
Hans Polak is online now Hans Polak  Netherlands
diyAudio Member
 
Join Date: Mar 2005
Location: Blaricum
Hi Max,

As in your previous version, Q11 and Q12 have a Vce of 0Volt and cannot function as transistors.


The biggest advantage of your two stage OPS is the cascoding of the Drivers.
The Darlingtons in #476 do not have this cascoding, resulting in a higher Vas loading which is an absolute disadvantage.
From a THD point of view, cascoding the drivers seem to be much more important as cascoding the output transistors , but I have no idea what is does to the sound reproduction, a thing that cannot be measured with LTSpice.


Hans

Last edited by Hans Polak; 9th February 2019 at 07:09 PM.
  Reply With Quote
Old 10th February 2019, 02:28 PM   #484
maxlorenz is offline maxlorenz  Chile
diyAudio Member
 
maxlorenz's Avatar
 
Join Date: Oct 2003
Location: osorno , Chile
Quote:
Originally Posted by Hans Polak View Post
Hi Max,

As in your previous version, Q11 and Q12 have a Vce of 0Volt and cannot function as transistors.

Oops! I must stop working on a hurry...


The biggest advantage of your two stage OPS is the cascoding of the Drivers.
The Darlingtons in #476 do not have this cascoding, resulting in a higher Vas loading which is an absolute disadvantage.
From a THD point of view, cascoding the drivers seem to be much more important as cascoding the output transistors , but I have no idea what is does to the sound reproduction, a thing that cannot be measured with LTSpice.

In my only experiment, which lacked inter-emitter resistor, there was some wanting of easiness to the sound, by comparison to the QUAD. I wish I could find a medium power MOSFET to make a cascode arrangement for the drivers similar to the output...


Hans
Visitors arrived. My little brother and family. Little hope for experimenting.
Have a nice musical weekend.
M.
  Reply With Quote
Old 11th February 2019, 11:09 AM   #485
Symon is offline Symon  United Kingdom
diyAudio Member
 
Join Date: Mar 2002
Location: Herts, UK
Hi Max,
as Hans says cascoding the drivers is important in terms of reducing the load on hte VAS, this was covered in Dr Arto's book, although he considered it only nessecary to cascode the driver and not the Output device as well. But as always listening testing is what counts, and I recall you reported sonic advantages to the use of mosfet cascode.

As with the input stage, driving the cascode devices Q8 & Q9 from a voltage referenced from the drivers emmitter terminal will improve things. see attached drawing, although I've not checked all the effects of this, one problem here is that the boot strape current from R6, R20, R26 & R31 directly affect the bias setting.
Or perhaps again stealing from input stage, buffer the cascode reference voltage as in my second attachment. the nice thing here is we don't need high voltage devices and they are not exposed to large voltage swings.


Anyway just thoughts at this stage.

Regaards,
Symon
Attached Images
File Type: png Capture3.PNG (10.6 KB, 96 views)
File Type: png Capture1.PNG (12.1 KB, 94 views)
  Reply With Quote
Old 11th February 2019, 01:08 PM   #486
JOSI1 is offline JOSI1  Germany
diyAudio Member
 
Join Date: Mar 2006
Location: Bad Hersfeld
Hi Max and Hans,

allow me a short summary of the present status esp. relating to a soon layout

output / driver stage
the bootstrapped driver stage and the bootstrapped output stage (Quad) is working stable stable and provides a good sound quality.
Simpler configurations (Triple, Darlington) are possible with my latest layout by mounting options.
(R39 and C35 are kept optional)

VAS
a differential VAS is not realized yet possibly in a future step
the present state is extended by an EF option as proposed by Hans.
(With my test amp this option had no influence on stability)

JFET Input stage
the JFET bootstrapped version (post #440) is less complex and provides a good audio quality (???) if I remember correctly a statement from Max.
For J1/J2 the 2N5462 should be used. This seems to be the only JFET that provides a high value for parameter vto=-5.4.
This parameter mainly determines the negative voltage at the sources of J1/J2 and should not to be too low (J103 vto=-2.0, J74 vto=-0.54).
Fortunately the SMD version MMBF5462 is still available (e.g. at Mouser 0.30€). The optional SMD footprint is part of my layout.
(soldering the SMD device to a 3pole strip is also possible)


All BJT Input stage
is more complex and requires more components.
If I remember correctly it was a proposal from Max to increase the power supply to +-60V (JFET limit about40V).
Are there other reasons esp. sound quality to implement this more or less complex versions.
Up to now I tested successfully the less complex proposal from Hans (post #473) using BC547C/BC557C only with my test amp.
@ Hans: what was the reason to use 2N3906, 2N5550, 2N5401
(higher Uce voltage??)

For the more complex proposal of Max / Symon up to now there is no final solution in sight.

Conclusion
Diver/Output stage and VAS section are rather certain for layout.
JFET bootstrapped input section is rather certain for layout
The All BJT Input stage has to be checked and is subject to change.

For a soon layout I suggest the following procedure for the input stage:
combined layout with
- JFET bootstrapped version and
- All BJT Input stage with the less complex proposal from Hans
- provision of Solder pads / connector pins to connect a daughter board for experimenting with a more complex solutions


When I perform a quare wave test (Ua=2/3 Umax) with my test amp I still have overshots/post ringing with a tendency to heavy oscillation (some hundred KHz, no HF oscillation).
With an input filter of at least R=1K C=470p to increase output rise time >2us the amp is very stable and shows a perfect slightly rounded 20KHz quare wave.
This behaviour is present with simplest output configuration (no cascoding, no boostraping) and is indepandant from VAS / EF-VAS,
CCS-VAS / Bootstrapped VAS

@Max: did you watch this behavior at your test amp

@ Hans: I never saw an overshot in all the simulations with the pulse signal. This pulse signal generates a positive quare pulse at the amp input that then can be watched at the amp out.
My Square wave generator generates a square wave signal symmetrical to GND (pos and neg. half wave) like in case of a sine wave signal and no over shots were measured at the input of my test amp

Could this explain the difference between simulation and a real measurement.

Cheers
  Reply With Quote
Old 11th February 2019, 02:49 PM   #487
Hans Polak is online now Hans Polak  Netherlands
diyAudio Member
 
Join Date: Mar 2005
Location: Blaricum
Quote:
Originally Posted by JOSI1 View Post
Hi Max and Hans,

@ Hans: what was the reason to use 2N3906, 2N5550, 2N5401
(higher Uce voltage??)
Several reasons, the most important one stability.
I always test without input filter cap with 0.4Vin and 0.2usec rise and fall time. Especially the CFP transistor has to be very fast to prevent oscillations, there I used the 5550 or even better the 5551 with a higher Hfe.
I used the 2N3904 for the input, because there is a cheap dual in SOT package giving large benefits in temp distortion.
And I used the 5401 because of the much higher voltage margin. Don't use transistors at their edge. especially when switching on or off they may die.
Because Fets are usually so much faster, you wont have as much sttability problems, but matching will be another problem and of course the restricted voltage reach.
Quote:
@ Hans: I never saw an overshot in all the simulations with the pulse signal. This pulse signal generates a positive quare pulse at the amp input that then can be watched at the amp out.
My Square wave generator generates a square wave signal symmetrical to GND (pos and neg. half wave) like in case of a sine wave signal and no over shots were measured at the input of my test amp

Could this explain the difference between simulation and a real measurement.

Cheers
What are those differences ? symmetrical or non symmetrical should not make a big difference, but be careful in real life to keep the amplitude low, not to blow your zobel network and preferably loaded with an 8 Ohm resistor.


Hans
  Reply With Quote
Old 11th February 2019, 07:56 PM   #488
maxlorenz is offline maxlorenz  Chile
diyAudio Member
 
maxlorenz's Avatar
 
Join Date: Oct 2003
Location: osorno , Chile
Thank you guys, wholeheartedly, to come to my rescue.

Dear Symon,

The first arrangement was tested at some point but it did not go well...
The second one is very interesting but first simulation showed poorer results...

Dear JOSI1,

Excellent summary of the "State Of The Project" (SOTP; sorry... I must stop watching political videos )
I concur with your views.

Output section: is a good starting (or end) point, and could be built progressively from simpler to more complex, if you wish so.

VAS: today I plan to re-assault the EF VAS with BC546-->N222X cascoded with "SC4793, which gave good performance in the attached simulation.
Since the original LTMD schematic from Peufeu was simple cascoded common emitter, I wonder if the putative LMTD is jeopardized...we'll see...or hear better.

Input section: my JFET version is sounding good an is stable with J74 and J103 with 32V supplies. I am inclined to think that, when buying a good amount of J103 one will find several units whose gain is more near 4mS...I have somewhere a practical test circuit.

My ALLBJT version with activelly bootstraped CFP is sounding good. This version, which is a modification from Symon's clever idea, is very dynamic. At least we should have connecting points for daughter-boards at the critical points, those being the emitters and the common node before the degeneration resistor, which is my present option.

Dear Hans,

Regarding Q11-Q12, my previous simulation showed 100mV of Vce and the present one increased to 200mV with different LEDs, which seems enough for LTspice: that reinforces the notion that searching for the optimal (THD; stability and soundwise) operating points is crucial, hence the recommendation of using variable resistors at first to set those points, in case one want to use ALLBJT, letting for a second time fitting LED alone or in combination or whatever units you prefer.

Regarding the optional transistor types, I let you inform us the better ones and it is up to JOSI1 to provide for the most versatile PCB possible to accomodate those.

Regarding tests: I am usually too excited with a good sounding mode to test the square wave in real life I may remember today.

Soooo, I am ready if you are.
Be prepared to face the unknown, though; be brave and do not despair!
(don't Panic!)

Best wishes,
M.
Attached Files
File Type: asc AMNESIS ALLBJT CASCODED INPUT (1).asc (18.7 KB, 13 views)

Last edited by maxlorenz; 11th February 2019 at 08:04 PM.
  Reply With Quote
Old 13th February 2019, 10:29 AM   #489
maxlorenz is offline maxlorenz  Chile
diyAudio Member
 
maxlorenz's Avatar
 
Join Date: Oct 2003
Location: osorno , Chile
OK. Had a little time. Tried the EF VAS, with MPSA18 instead of 2N2222 which has not arrived yet...
One try: +/- 500mV oscillation high MHz.
Since I was not on the mood, I did not try to debug it. Maybe with 2N2222 and the new PCB it will be OK...

Good part is that now I have the little MPSA18 (instead of BD139) as common emitter VAS (cascoded with 2SC4793) and it is stable. Listening test ongoing.
I think I found a good .model of the said BJT...

Cheers,
M.
  Reply With Quote
Old 13th February 2019, 11:21 AM   #490
Symon is offline Symon  United Kingdom
diyAudio Member
 
Join Date: Mar 2002
Location: Herts, UK
Hi Max,
Just an observation, MPSA18 is a very good device, low noise and high gain at low current. It also has low Cob.
As such it is usually chosen as input device, although how well suited it is to this circuit in positions Q3 and Q5 would need to be tested, and probably feedback resistor values modified.

The high gain of this device in such a complex circuit can cause oscilation, so increasing values of feedback resistors, and adjusting bias levels may be required in any location to get hte best from it.

Regards,
Symon
  Reply With Quote

Reply


The AMNESIS amp: a good amplifier, like a gentleman, has no memory.Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
new memory for PC AndrewT Everything Else 36 24th October 2015 11:09 AM
down the memory hole? tomtt Everything Else 16 9th September 2014 02:27 PM
My memory is failing me I think?? Moondog55 Analogue Source 3 31st January 2011 04:57 AM
LCR and of course Memory Resistor impsick Parts 2 3rd May 2008 09:30 AM
Notebook Memory eag Swap Meet 4 24th April 2006 05:20 AM


New To Site? Need Help?

All times are GMT. The time now is 11:24 AM.


Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 14.29%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
Copyright ©1999-2019 diyAudio
Wiki