GainWire-NGNFB-classB-PowerAmp

[vzaichenko]

Good point. The EC will see that variation as error and will compensate, even after the servo, so no problem. But will make a simulation to test that scenario, and let you know .

Cool

 

[smms73]

I'm just not sure how critical it is for error correction circuit, if the spreader shoulders are slightly unbalanced. If not critical - then no problem, I agree.

In simulation there is not a problem for the EC, but is better that the difference between the upper part and the lower to be less than 300mv.
This can be an alternative, as the triple emitter follower stage has a very high input impedance there is no significant voltage drop at the 2k resistors even at 20Hz.

 

[dadod]

In simulation there is not a problem for the EC, but is better that the difference between the upper part and the lower to be less than 300mv.
This can be an alternative, as the triple emitter follower stage has a very high input impedance there is no significant voltage drop at the 2k resistors even at 20Hz.

Do we need to increase the CCS current for the bias spreader as 2k2 resistor will take 0.9 mA from the spreader? In the spreader you suggest one transistor spreader is the option. What is optimum spreader current in your opinion with this EF3 OPS?

 

[smms73]

Damir I think that having 10mA or 9mA in the bias spreader will not make any impact at all .

This is the one transistor option. The R2 serves to regulate the bias.

...
Here is the amp with better CCS suggested by Sergio. I put it only in use in the bias spreader, I think it's not needed in the EC part, what do you think Sergio? ...
BR Damir

The CCS from the EC stage are also important to be more stable in term of current variation, as the EC loop does not correct for this variation, a difference of 20º of temperature between the upper and lower CCS will increase the DC output in 180mV, this variations have to be compensated by the DC servo alone. Not a great difference, but as the led ccs does not increase the complexity of the circuit, I think it should be use in this position also.

 

[vzaichenko]

Excellent - looks very elegant and easy to control...

 

[dadod]

I have test your bias-spreader today and it worked very well, 2 diodes is enough to thermal compensate the output stage. using the 4 diodes can cause overcompensation.
It is only need to use two thermaltrak transistors (NJL3281D/NJL1302D) the other pair can be a (MJL3281A/MJL1302A) as it is the same transistor, but without the sense diode, and it costs half the price. of-course it can be use 4 thermaltrak transistors but one pair will have the diodes without use.

this is a image of the bias spreader. R1,R3,R5 have to be static resistors, R2 and R5 can be in serie with trim-pots to set the output Bias current.
With the values in the picture the tempco is 4,35mv/cº for compensation of the output transistors.

The two transistors will have to share a heat-sink with the prédriver and driver transistor for better thermal compensation.

Sergio, I think that your tempco calculation is wrong according to B. Cordell book.
It should be 1+(R2+R4)/R3. What resistor/s could be best used as a trimmer for setting the bias?

ps. and that is valid for your one transistor spreader too.

 

[smms73]

Damir, that is a mistake in the book, I guess that Cordell as all of us also make some mistakes . It is easy to see that R2 and R4 does not influence the multiplication of the diodes tempco. It is also easy to test by altering only the temperature in the diodes and then measure the correspond voltage variation at the spreader terminals.

 

[smms73]

Other thing that is also wrong is the Cordell model for the thermaltrak diode.
In first image is the variation of the forward voltage of the sense diode, at 10mA we have 570mv at 25º and 440mv at 100º and that gives a tempco of 1.7 mV/Cº.

The simulation with the Cordell model the Fv is 580mV at 25º and 380mV at 100º , and that is a 2.64 mV/Cº.

 

[smms73]

Actually the model that can be found on ONsemi site are much closer to reality.

.MODEL Dnjl3281dg_diode d
+IS=4.26979e-09 RS=0.0893177 N=1.52404 EG=1.10197
+XTI=0.5 BV=210 IBV=0.0001 CJO=4.69098e-11
+VJ=1.5 M=0.49835 FC=0.5 TT=5.10525e-08
+KF=0 AF=1

the simulation of the temperature variation from 25º to 100º gives 580mv 446mv or a 1.8mV/Cº .

I wonder why Cordell fell the need for changing the ONsemi model

 

[MiiB]

Thanks... good knowledge.

 

[smms73]

This is the test circuit that I use, it has the Cordell model and the ONsemi model. Verify for yourself´s as I can also be skipping something .
to change the temperature of diodes alone change the param T , as it is, make a temperature variation from 25 to 26º on the diodes and the transistor stays in 25º.
This is the circuit that is on Cordell book.

Ps: try also to change R1, and see if there is any variation on spreader tempco.

 

[dadod]

Damir, that is a mistake in the book, I guess that Cordell as all of us also make some mistakes . It is easy to see that R2 and R4 does not influence the multiplication of the diodes tempco. It is also easy to test by altering only the temperature in the diodes and then measure the correspond voltage variation at the spreader terminals.

Sergio, you are right again. In the Cordell book page 306 and 307, from six figure, tree are with quite wrong Sttd (fig 14.18 c and 14.19 a and c).
Problem coming from that is that one choose wrong resistor to be used to adjust the bias current.
I have nice explanation somewhere, how to calculate a tempco for different spreader types, but I can't find it now.

 

[dadod]

I simulated thermal compensation of output TT transistor with simple bias spreader using tree sets of TT models (output transistors and diodes).
Models for syn08 give closest to expected result. Stdd is about 1.3 (1+2k/6k8) for all tree tests.

Syn08 models (if I remember correctly from hum thy came from)

.MODEL Qnjl0281_bjt npn
+IS=7.35019e-14 BF=113.063 NF=1.21951 VAF=10.158\par
+IKF=8.51404 ISE=8.83106e-12 NE=2.0497 BR=4.9071\par
+NR=1.25326 VAR=8.60208 IKR=6.78731 ISC=3.59743e-13\par
+NC=3.99215 RB=0.526348 IRB=0.1 RBM=0.526348\par
+RE=0.000612798 RC=0.0302655 XTB=0.1 XTI=1\par
+EG=1.05 CJE=5.60463e-09 VJE=0.600913 MJE=0.848202\par
+TF=2.07647e-09 XTF=1000 VTF=4157.4 ITF=217.206\par
+CJC=4.77948e-10 VJC=0.528497 MJC=0.23 XCJC=1\par
+FC=0.430232 CJS=0 VJS=0.75 MJS=0.5\par
+TR=1e-07 PTF=0 KF=0 AF=1\par

.MODEL Dnjl0281_diode d
+IS=5.17122e-08 RS=0.0324668 N=1.78236 EG=1.12011\par
+XTI=0.5 BV=200 IBV=2e-06 CJO=5.59229e-11\par
+VJ=0.4 M=0.346657 FC=0.5 TT=2.59685e-08\par
+KF=0 AF=1\par

.MODEL Qnjl0302_bjt pnp
+IS=1.68444e-12 BF=84.9671 NF=0.992223 VAF=10.01\par
+IKF=7.49667 ISE=1e-08 NE=2.87955 BR=8.49671\par
+NR=0.990995 VAR=99.9986 IKR=0.382704 ISC=3.81183e-16\par
+NC=1 RB=4.23352 IRB=0.1 RBM=4.23351\par
+RE=0.00779693 RC=0.0390712 XTB=1.60824 XTI=3.9999\par
+EG=1.20458 CJE=1.56322e-08 VJE=0.99 MJE=0.23\par
+TF=1.97297e-09 XTF=1000 VTF=817.29 ITF=453.232\par
+CJC=5e-10 VJC=0.95 MJC=0.23 XCJC=0.756601\par
+FC=0.8 CJS=0 VJS=0.75 MJS=0.5\par
+TR=1e-07 PTF=0 KF=0 AF=1\par

.MODEL Dnjl0302_diode d
+IS=5.17122e-08 RS=0.0324668 N=1.78236 EG=1.12011\par
+XTI=0.5 BV=200 IBV=2e-06 CJO=5.59229e-11\par
+VJ=0.4 M=0.346657 FC=0.5 TT=2.59685e-08\par
+KF=0 AF=1\par

BR Damir

 

[smms73]

Cool.
Damir i have been use R12 with 2k2 R13 with 10k . Can you try those values, please ?

 

[dadod]

Cool.
Damir i have been use R12 with 2k2 R13 with 10k . Can you try those values, please ?

As this? Still quite overcompensated.

 

[smms73]

Damir, is overcompensated because the tempco of the diodes models are higher than reality, if you want your simulation to look good you have to change the 10k resistor to 13k2, but in reality that value is to high. The tempco of the onsemi models are 1.8mv per c and the transistor 2.1mv so the multiplier factor should be 1.16 not the 1.3 that you use or the 1.22 that i use , but using those values in reality will make the amplifier undercompensated.

 

[dadod]

Damir, is overcompensated because the tempco of the diodes models are higher than reality, if you want your simulation to look good you have to change the 10k resistor to 13k2, but in reality that value is to high. The tempco of the onsemi models are 1.8mv per c and the transistor 2.1mv so the multiplier factor should be 1.16 not the 1.3 that you use or the 1.22 that i use , but using those values in reality will make the amplifier undercompensated.

Bob said that the TT diode tempco is 1.7mV and that gives multiplication factor od 1.235 but he suggested 1.3.
Anyhow I have to finish my TT amps (six channels for Orion loudspeakers) and I will check with it. All boards are ready, only need to be put in the boxes and adjusted.
What I noticed with my two channels TT amp (playing for more then two years) is that TT diodes tempco could differ from one to another TT transistor.

 

[smms73]

I also think that 1.3 is better. Is safer to have a bit overcompensated than undercompensated.

 

[dadod]

Non global negative feedback showed here used EC in the OPS and only local NFB in the IPS.
I wanted to tray an CFA IPS and for that I needed different buffer as the one with jfet has limitation using CCS and could not drive Current Feedback low impedance circuit.
Here is the BJT buffer with current on demand.
It has quite low distortion, only 3 ppm at 20 kHz and 80 Vpp.
Damir

 

[dadod]

Here is the same amp with non GNFB, only it uses the BJT buffer presented.
Damir