GreenAmp ++ modulated Class G output

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
OK , back to the "stuff".


(Below 1) is my input-VAS stage.



It is a full complimentary LTP (leach - like) combined with the "special sauce"

of a Baxandall super-pair / hawksford cascode VAS.


- Easy PPM performer ! Have to load it down to 2R to break 10ppm. Typically ,
sub ppm from 10K down.
-Self clamping saturation at clip , I will show this later. At 1% OMG clipping ,
all even order (like a Valve amp).


I will combine this with a error correcting - (modulated) CLASS G output stage. I will
show this in all its fast modulated goodness in my next few entries.
80% efficiency , little heat ,, here we come ... could go lateral MOSFET , as well.
SMPS is also the recommended supply. :D:D:D


This will be a full amplifier in the final form , it may start as a modular in it's
test phase .

OS
 

Attachments

  • greenamp V1.jpg
    greenamp V1.jpg
    376.4 KB · Views: 4,194
clipping and anti saturation

(below 1) is gross overload , the IPS will run cooler , like a "greenamp" should.


(below 2) is what to expect with a better output stage. An error correcting
output stage has 1/10th the THD (for NFB) to correct. This greenamp IPS
is a "distortion eater" , nearly sub - ppm in isolation.


I ALMOST have the GEC = (G class error correcting) OP stage done.
The greenamp IPS should have no problem "eating" it's distortion.


OS
 

Attachments

  • greenamp V1 clip . sat.jpg
    greenamp V1 clip . sat.jpg
    129.2 KB · Views: 4,044
  • greenamp V1 IPS.jpg
    greenamp V1 IPS.jpg
    77.1 KB · Views: 3,768
Nice approach!

But, some questions:
1. Why noninverting?
2. What will be CMRR?
3. What will be PSRR?
4. Why IPS CCS (Q5/Q6) aren’t cascoded?
5. Where D3/D4 reference filters?
6. Why righ shoulder of the LTP’s aren’t used? Put mirror there.
7. Why so small active LTP shoulder load(R19/R21) are? Again put mirror.
8. Followers (Q7/Q11) - clearly and exactly!
9. Why VAS stage hybrid driven? Redraw as common base.
10. Why last VAS stage are supplied by low-voltage rails?
11. Where are VAS load resistance? What nominal of RC-circuit?
12. Where base stoppers at first stage of output follower?
13. 2SA1381 drive MJL4302 - mouse pick a trailer. Add at least KSA1220A inbetween.

To be cont’d...
 
Class g woes ....

I have simulated successfully , but more THD and "glitches" beyond 10k.


Here (below) , are the .ASC + models.


The greenampV1 is the bada$$ input stage running a triple OP.
The greenampV1-ops is the class G trial run.


BTW - the GOAL is to produce a 200W amp with 1/2 the heatsink requirements
running on a SMPS in a 1U enclosure.
AMP will be 4 devices (2 steps + 2 active) , I expect to at least come close to
EF3 performance.

PS - Does anybody have the models for the Exicon lateral MOSFET's ? Exicon Lateral MOSFETS - Our Product Range
With these , more class A bias before the class G stepping (pix below)

OS
 

Attachments

  • greenampV1.asc
    31.9 KB · Views: 228
  • greenampV1-ops.asc
    26.1 KB · Views: 165
  • KT-Cordell models.txt
    57.3 KB · Views: 210
  • greenampV1-OPS.jpg
    greenampV1-OPS.jpg
    259.4 KB · Views: 3,645
Hi,

try faster driver for the outer devices, such as 2SC4793/2SA1837. My solution has extreme small glitch in the simulator.

Sajti


I did try the C4793/A1837 pair.


If you run the simulation , you will see the negative step is not going symetrically
with the positive one. More load also makes thing worse.
Hitachi had a line of receivers and amps (sr-200x) that used 4 devices. 230W
@ 40 + 90V rails. They used the driver side to initiate the pos/neg step instead
of output local FB like this iteration.


The local FB of this class G stage adds another pole globally to the amp as a whole.
I did get my IPS to oscillate. Instability in the local OP FB.



I still have quite a few tricks left !!! :scratch2::scratch2:


Edit - Thanx much , Dadod !!

OS
 
Last edited:
Nice approach!

But, some questions:
1. Why noninverting?
2. What will be CMRR?
3. What will be PSRR?
4. Why IPS CCS (Q5/Q6) aren’t cascoded?
5. Where D3/D4 reference filters?
6. Why righ shoulder of the LTP’s aren’t used? Put mirror there.
7. Why so small active LTP shoulder load(R19/R21) are? Again put mirror.
8. Followers (Q7/Q11) - clearly and exactly!
9. Why VAS stage hybrid driven? Redraw as common base.
10. Why last VAS stage are supplied by low-voltage rails?
11. Where are VAS load resistance? What nominal of RC-circuit?
12. Where base stoppers at first stage of output follower?
13. 2SA1381 drive MJL4302 - mouse pick a trailer. Add at least KSA1220A inbetween.

To be cont’d...


I will answer all these questions after I get the OP stage. This IPS will
work , already built "slewmaster" Slewmaster - CFA vs. VFA "Rumble"
search for "kypton C" and "spookyamp" , this IPS is the hybrid of the two.
OPS will have cap multipliers , the spooky had close to -100db PSRR
with the "slewmaster" output stage/cap multiplier combo.


PS - I simplified the Leach/Spooky LTP stage so more would build it.
The Spooky had over 100 components , I built it ... but most went for the far simpler
CFA IPS's.



OS
 
Last edited:
I don't think, that it is very tricky.

Sajti


Your Amp simulation need your model list. They must be OEM models, I use
the KT - Cordell models.

Edit - You chose a true blue low inpedance NFB CFA for your amp. Would
the CFA be the better choice (more stable , faster - versus a VFA) , for this
type OPS ??



PS- your input stage is cool. I used that "enhanced mirror" for my IC/CFA

that used an op-amp instead of the diamond.
In your OPS , both drivers and OP's "stepped". Where you inspired by a classic design or is it original ?



OS
 
Last edited:
Your Amp simulation need your model list. They must be OEM models, I use
the KT - Cordell models.

I get most of modells from here. You can use the Cordell models too.

In your OPS , both drivers and OP's "stepped". Where you inspired by a classic design or is it original ?

The OPS idea comes from Crest Audio. It is very simple, and the common emmitter outer stage means very small loss, almost like the mosfet switched version.

Sajti
 
I don't think, that it is very tricky.

Sajti


Your Amp simulation need your model list. They must be OEM models, I use
the KT - Cordell models.




PS- your input stage is cool. I used that "enhanced mirror" for my IC/CFA

that used an op-amp instead of the diamond.
In your OPS , both drivers and OP's "stepped". Where you inspired by a classic design or is it original ?



OS
 
AS I suspected.


.MODEL 2SA1837_A PNP ( IS=2.3937e-010 NF=1.304 BF=250 VAF=273 IKF=1.9914 NK=0.9117 ISE=2.1180e-011 NE=1.5815 BR=4 NR=1 VAR=20 IKR=1.05 RE=0 RB=1.702 RC=1.65 CJE=1.5564e-010

VJE=0.726406 MJE=1 CJC=7.2230e-011 VJC=0.75 MJC=0.33 TF=1.16651e-9 FC=0.966476 ITF=1270.37 XTF=10000 VTF=35 TR=1.38U Vceo=230 Icrating=1A mfg=Toshiba).MODEL 2SA1837 PNP (

IS=2.39372559E-10 NF=1.304015937 BF=300 VAF=273 IKF=2.087725944 NK=0.94719458 ISE=1.46829699E-11 NE=1.526663542 BR=4 NR=1 VAR=20 IKR=1.05 RE=0 RB=1.8 RC=1.65 CJE=4.7407E-10

VJE=1.1 MJE=0.5 CJC=8.6700E-11 VJC=0.3 MJC=0.3 TF=1.642191E-09 FC=0.5 ITF=1.076260106 XTF=5.868994022 TR=1.38U)


.MODEL 2SA1837 PNP ( IS=2.39372559E-10 NF=1.304015937 BF=300 VAF=273 IKF=2.087725944 NK=0.94719458 ISE=1.46829699E-11 NE=1.526663542 BR=4 NR=1 VAR=20 IKR=1.05 RE=0 RB=1.8

RC=1.65 CJE=4.7407E-10 VJE=1.1 MJE=0.5 CJC=8.6700E-11 VJC=0.3 MJC=0.3 TF=1.642191E-09 FC=0.5 ITF=1.076260106 XTF=5.868994022 TR=1.38U)
The "standard" model has much less info (saturation , early ... many others).


It is "trickier" with the Keantoken/Cordell models , I have found they model
the real thing (built/tested slewmasters) right down to the " glitches". :D

I did have one of my PNP emitter follower step drivers E-C reversed.

Nice now , clipping and THD (below 1/2).
Now to add some EC to the OPS !!



OS
 

Attachments

  • Greenamp-OPS-clip.jpg
    Greenamp-OPS-clip.jpg
    112.8 KB · Views: 982
  • greenamp-ops-good.jpg
    greenamp-ops-good.jpg
    372.2 KB · Views: 976
AS I suspected.
The "standard" model has much less info (saturation , early ... many others).


It is "trickier" with the Keantoken/Cordell models , I have found they model
the real thing (built/tested slewmasters) right down to the " glitches". :D

I did have one of my PNP emitter follower step drivers E-C reversed.

Nice now , clipping and THD (below 1/2).
Now to add some EC to the OPS !!

I use some models from Keantoken, especially the KSC1845/KSA992 and 2SC4883/2SA1859 for my V4 amplifier.

The CFP solution needs slow down with the Ccb on the drivers. This needs to be adjusted carefully in practice, to avoid instability. Some more phase margin for the feedback may necessary too.

Sajti
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.