Patchwork Reloaded: Circuit Optimization and Board Layout.

[ferencz]

MJL1302A

[MJL21193]

Originally posted by homemodder
Nice MJL21193

After you have changed to 2sa1381 youll get even better performance. The 2sa1011 are excellent drivers, very good choice.
2sa1360 or 2sa1407 are aslo excellent Vas transistors. Change those 2n devices for something like 2sa970/2sc2240, 2sa992/2sc1845 or 2sa1016/2sc2362 and youll up performance some more.

Hi homemodder,
I don't have any of the 2SA970/C2240, but given the position of the 2N's in my amp - current mirror and cascode - do you think they will make that much of a difference? I think I could use a better device for the EF on the VAS maybe.

My order came from Digikey. Lots of goodies!

Look at my new outputs and VAS----

[MJL21193]

Originally posted by ferencz
I simulated your circuit in Microcap-9.
I'm getting ~0.1% THD at 20kHz with 1Vrms input. Mostly first order (40kHz) distortion.

Originally posted by megajocke
Are you sure the period of the signal, the length of the simulation, the length of the FFT window and all that stuff is setup right?

Try measuring distortion of the input signal to see if it is working right.

Hi ferencz,
Like mega said AND also check idle current. Best results at 50-60mA per device.
Also, as mentioned, models make a huge difference. I have known good models that I'm working with. I have updated quite a few in my library and download from reliable sources (like Fairchild).

Here's what I have at 20kHz. Input signal 500mV, output power ~14 watts. At full power (145 watts) I get .011%.
Since I can't hear much above ~12kHz anymore, it's not a big deal for me. Still, given the test results from the original amp, these figures are very close to actual, and truly as good as they need to be.

[MJL21193]

Originally posted by Nico Ras

I confirm your findings. At 1kHz it is 0.0084%. Also Q3 and Q4 worsens matters slightly and could be omitted.

Originally posted by megajocke
No, the LPT transistors are 45V types so you can not omit the cascode. One of those things most simulators don't simulate

Originally posted by Nico Ras
I am sorry 2SC2240 and 2SA970 is 120V so what now?? I thought those where the suggested changes.

Hi Nico,
I'm not going to change the BC550 in the LTP, so the cascode will need to remain. I don't see this as a major source of distortion in the present setup.
The 2N5400 and 2N5550 are not bad for this. Not high fashion, but get the job done.

NOW, the EF might be an area for consideration. Suggestions? Am I correct in my understanding that with 56V rails, this need only be the rail voltage, as it doesn't swing to the negative rail? So maybe a BC556 would look better here?

[darkfenriz]

Going below say 0.02% THD at 20kHz in a design like this is hardy doable. I don't know where these 0.0000x% figures came from.

[MJL21193]

Thanks for the models homemodder.

Originally posted by CBS240



Hi

Looking better. Still a little bit slow on the slew for 100KHz, but not bad for this topology.

I usually disconnect the Zobel and input filters during square wave tests. I have burned a Zobel resistor up before with fast signals (crappy little carbon resistor caught fire., now I use metal film).

Hi,
I left the input filter and the Zobel, just a quick look at the high frequency stability. When I get the final board layout together for testing, I'll do these runs without the input filter and Zobel.

Originally posted by MJL21193

Here's what I have at 20kHz. Input signal 500mV, output power ~14 watts. At full power (145 watts) I get .011%.

Not correct. I still had the 220pF miller cap (i was experimenting earlier with higher values) in the simulation and a few other things boggled.

More like this. This is at full output power. Fact or fiction, it's what the machine says.

[MJL21193]

Originally posted by Nico Ras


Hard to tell whose models are real, would you consider a manufacturer or an audio enthusiast. By the way these models push the THD up a tad over those supplied by on semi.

Hi Nico,
Andy_c seems to have a competent grip on the simulation process, and I have been using his MJL models since the original amp. These give much better performance than the ones in the Multisim library, or the ones from On Semi.

Originally posted by darkfenriz
Going below say 0.02% THD at 20kHz in a design like this is hardy doable. I don't know where these 0.0000x% figures came from.

Hi darkfenriz,
Here the original amp was tested on an HP339A. It pretty much verifies the distortion figures I was getting at the time. All of the changes in the current project have driven the speed up and the distortion down. I'm not looking for ultra-low distortion, but I'll take it if it comes with the effort here.

[homemodder]

Hi Nico, good to see some south africans here. Im ex SA.

AndyC s models are excellent in fact, he is much more than just a audio enthusiast. If you look at the graphs vbe/Ic he gets it is much more realistic, it must have taken him a lot of time and patience to get his results. Manufacturers models are terrible especially onsemi, worse ive seen. You got to keep in mind some of these audio enthusiasts are qualified engineers and could easily be working for these companies, if fact some that come to this forum do. If you get worse results with AndyC s models that is what youll get real time, its no use using bad models giving you false indications of better performance if you not gonna get them real life. Have a look at his webpage, he is defenitly qualified to make these models.

MJL21193,
the use of 2sa970 will definetly bring improvement, measured and sonically. The bc s you using arent so bad in ltp, but those 2n devices in current mirror ect are poor performers. Trannies in mirrors and current sources should have very large beta and minimum capacitance, have a look at walt jung s papers. Using 2sa970 in vas Ef will also bring improvement. There is a thread here by fotios in which he wanted to improve rise time of his amp, also a Lin topology. After some failed attempts he tried john curls advice and replaced 2n for the japanese devices, have a look at that thread and improvements he got, he took some pics of his scope too. A simple amp like this can sound very very good if one optimises component choices and working parameters.
Nice devices you got there, does digikey sell 2sa1011 and comp?? What are those output ones?? For these harder to obtain japanese devices you could try bdent, canadian company I think they have large selection and in most cases cheaper than digikey.

Compensation in this topology is critical, I dont understand why you would be increasing the miller instead of reducing it, as long as amp is stable.

[MJL21193]

Originally posted by homemodder

A simple amp like this can sound very very good if one optimises component choices and working parameters.
Nice devices you got there, does digikey sell 2sa1011 and comp?? What are those output ones?? For these harder to obtain japanese devices you could try bdent, canadian company I think they have large selection and in most cases cheaper than digikey.

Compensation in this topology is critical, I dont understand why you would be increasing the miller instead of reducing it, as long as amp is stable.

Hi homemodder,
That is the overall objective - to improve performance, but within reason. I'm picking up bits and pieces as I go through this and become more and more informed. I will run the simulation with higher gain devices and see how the results look. Like I said before, the original amp sounded pretty darn good and measured very well, this had the 2N's in circuit.

I have registered at B&D and found the 970/2240. When I get the chance, I will order some of these excellent looking units.

No, Digikey doesn't sell the 2SA1011, I bought those last year from AmpsLAB.
I did get the 2SB1011 (Panasonic) 2SA914 (Panasonic) and the new ones - 2SA1381 (Fairchild) at Digikey. The outputs are Fairchild 2SC5200/1943. 25 of each.

As for the 220pF for the miller, I was trying a few different things in the simulation, worst case stuff. Just as an exercise, not to implement in the amp. I will be trying lower miller caps as I go along. The present one is 100pF and the amp is dead steady stable.

I changed the outputs and the VAS in the prototype. Stability is awesome. Base stoppers for the outputs are still shorted, so I'll be getting rid of those in the final design / board layout.
I won't show how it looks handling a 100kHz squarewave, because I promised...

But, I do have a clipping pic to share. This is clipping at +/-30V rails, my lab supply maximum.

[MJL21193]

Originally posted by homemodder
Bad models probably, try these

.MODEL SA1943 PNP (
+ IS =3.5476E-11
+ BF =159.9
+ NF =1.0
+ BR =25.75
+ NR =1.011
+ ISE =2.5119E-10
+ NE =2
+ ISC =7.9433E-11
+ NC =1.37
+ VAF =60.0
+ VAR =11.07
+ IKF =2.8370
+ IKR =0.3548
+ RB =2.74
+ RBM =0.0381
+ IRB =3.6308E-3
+ RE =0.06
+ RC =0.01
+ CJE =4.1783E-9
+ VJE =0.6354
+ MJE =0.3374
+ FC =0.5
+ CJC =1.1383E-9
+ VJC =0.5
+ MJC =0.3699
+ XCJC =0.7624
+ XTB =1.5306
+ EG =1.1751
+ XTI =3.0 )


MODEL 2SC5200 NPN (
+ IS =4.3031E-12
+ BF =152.1
+ NF =1.0
+ BR =6.155
+ NR =1.028
+ ISE =1.3924E-11
+ NE =1.5
+ ISC =2.7542E-11
+ NC =1.95
+ VAF =60.0
+ VAR =6.51
+ IKF =10.8637
+ IKR =0.1585
+ RB =2.47
+ RBM =0.02
+ IRB =0.08
+ RE =0.04
+ RC =0.015
+ CJE =5.8111E-9
+ VJE =0.6506
+ MJE =0.3357
+ FC =0.5
+ CJC =6.4394E-10
+ VJC =0.5
+ MJC =0.3966
+ XCJC =0.7624
+ XTB =1.0445
+ EG =1.1663
+ XTI =3.0 )

Those are different from the ones I have.
The ones I have were downloaded from Fairchild and are excellent models, giving performance equal to or better than andy_c's models.
TRY them

model F5200 NPN
+ IS = 3.0463E-11 BF = 96.20 VAF = 100
+ IKF = 15.04256 ISE = 5.6190E-11 NE = 2.0
+ BR = 4.849 IKR = 1.05012 VAR = 100
+ ISC = 7.18E-8 NC = 1.5 RE = 0.0025
+ RB = 20.18 RBM = 0.0014 IRB = 1.0E-7
+ RC = 0.01137 CJE = 4.5000E-10 CJC = 8.4915E-10
+ VJC = 0.68977 MJC = 0.54081 TF = 6.8583E-10
+ XTF = 9.5721 VTF = 10.425 ITF = 6.8697E-2
+ TR = 1.000E-8 XTB = 1.45 EG = 0.82
+ FC = 0.5

.model F1943 PNP
+ IS=1.30E-10 BF=91.42 VAF=100
+ IKF=4.480 ISE=1.02E-10 NE=2.0
+ VAR=100 ISC=5.0900E-9 NC=1.5
+ BR=0.882 IKR=2.9015 RE=0.0011
+ RC=0.0553 RB=140.05 RBM=0.0041
+ IRB=8.5e-9 CJE=2.00E-10 FC=0.5
+ CJC=9.45E-10 VJC=0.48 MJC=0.28
+ TF=9.250E-10 XTF=10 VTF=10
+ ITF=1 TR=1.00E-8 EG=0.76
+ XTB=2.68