DIYA store "Wolverine" (Son of Badger) .... suggestions ??

So here is a more realistic test for comparing 0ohms vs 100ohms for the driver base resistor.

Ambient is swept from 25°C to 40°C

Driver is 5°C above Ambient, Q103 thermally coupled to the Driver

Outputs are 15° above Ambient, Q104 thermally coupled to the Outputs

Below are plots showing bias current variation over temperature through one of the outputs emitter resistors

Jeremy

To those who are concerned about the bias trending upwards and myself included, put your minds at ease. The problem is with the Cordell model for MJE340, well at least for temperature. Specifically in the MJE340_C in the KT-Cordell models.txt.

I tried Onsemi's spice model for MJE340, KSC3503, and TTC004b. All of them the bias is trending downward slightly.

Here is the updated sim with driver base resistor adjustments.

Jeremy
 

Attachments

  • Driver Base Resistor = 0ohms.png
    Driver Base Resistor = 0ohms.png
    75.3 KB · Views: 711
  • Driver Base Resistor = 100ohms.png
    Driver Base Resistor = 100ohms.png
    76.3 KB · Views: 637
  • Schematic.png
    Schematic.png
    112.2 KB · Views: 638
Last edited:
Hi Stuart

Did you verify the current on the LED's while the amp is clipping ?

It seems the current will be excessive .... Unless you use a huge value for R17... now you chose 27k .... did you verify the effect of such a high value resistor in the amp SR ?

In my build I found it affects SR if R17 is higher than 6k8

No issues here... slew rate is symmetrical 44V/us. I did a manual measurement so you could see.

No issues clipping... perfect

See pics below
-Left picture is slew rate measurement
-Middle picture is clipping 20KHz into 8ohms
-Right picture is clipping 20kHz into 4ohms

Jeremy
 

Attachments

  • Slew 56pF and 330pF Compensation.png
    Slew 56pF and 330pF Compensation.png
    85.5 KB · Views: 583
  • 20khz 8ohm clipping.png
    20khz 8ohm clipping.png
    79.3 KB · Views: 568
  • 20khz 4ohm clipping.png
    20khz 4ohm clipping.png
    79.2 KB · Views: 141
Last edited:
To those who are concerned about the bias trending upwards and myself included, put your minds at ease. The problem is with the Cordell model for MJE340, well at least for temperature. Specifically in the MJE340_C in the KT-Cordell models.txt.

I tried Onsemi's spice model for MJE340, KSC3503, and TTC004b. All of them the bias is trending downward slightly.

Here is the updated sim with driver base resistor adjustments.

Jeremy

On Post#1161

The left picture is Driver base resistor = 0ohms
The right picture is Driver base resistor = 100ohms

Jeremy
 
Sounds like a good idea to me. To be honest, to this date I have never heard a good reason to have R10. So far, I have heard:
- It provides an easy way to measure I_tail since the voltage across / 1000 would be the current. I guess that is fair since if R11 and R12 are variable, it is hard to tell what the current is.
- It Isolates the capacitance of the tail current source from the diff-pair. That is unlikely since if COBO of the transitor is ~5pF, the frequency where it would matter is ~32MHz.
- I reduces the voltage across the CCS transistor. Maybe, but a 5V or so reduction is not that much to matter.

Maybe the first one is the best reason.
 
Member
Joined 2006
Paid Member
Sounds like a good idea to me. To be honest, to this date I have never heard a good reason to have R10. So far, I have heard:
- It provides an easy way to measure I_tail since the voltage across / 1000 would be the current. I guess that is fair since if R11 and R12 are variable, it is hard to tell what the current is.
- It Isolates the capacitance of the tail current source from the diff-pair. That is unlikely since if COBO of the transitor is ~5pF, the frequency where it would matter is ~32MHz.
- I reduces the voltage across the CCS transistor. Maybe, but a 5V or so reduction is not that much to matter.

Maybe the first one is the best reason.

It also avoids turn on thump.
 
When discussing thermal compensation IIRC OS had remarks about running an amp outside in the snow and having no trouble with thermal stability, so the point seemed to be that his thermal design was tested good regardless of what simulation says. I'm hesitant to start redesigning everything without his input.
 
Nothing has been redesigned with the thermal compensation, only base resistors have been added to the drivers. To prove this a I simmed the two configurations and they both had the same downward slope in bias over temperature. The bias over temperature will be even flatter if you factor in some thermal attenuation between the heat sources at the vbe multipliers.

See post #1161 for plots

Jeremy
 
I explained more here:

diyAB Amp - The "Honey Badger"

Turn on thump would be a compelling reason to have the resistor, and for me it is also a good thing to have a way of measuring the CCS current.

The main question was not that series resistor R10 on the CCS was installed or not but if the buffer transistors Q7 base for the zener /cascode should be tied toward the CCS end or the Diff pair end. With the buffer transistor Q7 base is tied to the top side of R10 the cascode voltage will change depending on what current is going through R10. Moving the base of Q7 to the bottom of R10 the cascode voltage will be constant with tail current.

Jeremy
 
Last edited:
It might or might not. When the amp turns on, the resistor limits the LTP current until the CCS comes out of saturation, so it ramps up slowly until the resistor's operating voltage is reached. This means that the LTP can't overdrive the proceeding stages as much, so there is not such a large imbalance condition or stored energy to correct, and if there is latching behavior in the VAS it can prevent starting up with the output clamped to one rail.

How much of a problem this is depends on details of the circuit design, so even for the same topology you can't always assume that it will behave the same across all variations. Some circuits might need this resistor to be absent to start up properly.

When there is a need to power up the stages in a specific sequence the resistor is one thing to try.

When you have high rail voltage and a hot biased LTP, the resistor can help to spread out the heat generation or move it away from the board.

In amps with TPC the amp can be unstable at low LTP currents so it might be better to omit this resistor to prevent a burst of oscillation on startup.

This along with the stability compensation and thermal design are things that are much more quickly decided on after the first build, than trying to predict. And that is why we are haggling about them right at the end of the design phase. But they aren't any easier to determine now than they were before. If we had a build we could just try each available option to see what works best. That would not take weeks like what is happening here. Hence why I suggested to move on to building over a month ago...
 
Last edited:
Makes sense. Another approach would be build a cascoded current source for the CCS. The cascode transistor would take the voltage drop and hence the power consumption and also control the turn on of the CCS. But, it is 2 components over 1 (you need a transisor and proably a diode to bias) and you lose the ability to measure I_TAIL.

If you are a nut, there is a way to configure the cascode to null out the CCS capacitance... but now we are entering the overkill arena.
 
Member
Joined 2016
Paid Member
Project update.
1. Schematic has been updated and high resolution PDF created.
2. IPS and EF3 boards updated.
3. Basic LTspice file created with new values and simulation parameters.
4. BOM 95% complete. Harry has done a great job with the BOM and we all need to understand that unfortunately ATM part are going out off stock quickly so in a month or two some part maybe out of stock. So people may need need to use the parts specifications to source a suitable replacement.


5. Final checking is happening ATM.
Once that is completed I will send OS a full file set for his review.

I'm not sure what he is up to ATM but hopefully everything is ok with him.

Once his feedback is received we can start ordering boards.