I think thinks it's better to continue this in a separate thread:
As I figured out a resistor combination that gives an almost perfect stabilization of the bias over time with this circuit:
...I implemented this on the second channel. There it was highly overcompensated and not satisfactory at all.
A drawback of this circuit is also, that some current is bypassed through the voltage source. Another one is that the voltage divider or the voltage across R30 depends on current in R29.
As I figured out a resistor combination that gives an almost perfect stabilization of the bias over time with this circuit:
...I implemented this on the second channel. There it was highly overcompensated and not satisfactory at all.
A drawback of this circuit is also, that some current is bypassed through the voltage source. Another one is that the voltage divider or the voltage across R30 depends on current in R29.
Last edited:
Then I tried a different type of voltage source: A germanium and Schottky diode:
It turned out that the main source for the different compensation was that I used MJE340 on the one and 2N3503 on the other channel.
Both issues were solved, voltage is stable and no current is bypassed. It worked very well for first channel, then again the second showed a completely different behaviour.
After trying around, I finally made 8 measurements with different combinations of Q13, diode and both channels:
My conclusion:
BJTs vary in tempco slightly and this small variation is amplified by the Vbe multimplier and hence the choice of Q13 has a major influence on compensation. I also swapped the power devices between the channels and the difference was not as high as between different Q13.
The forward voltage of Schottky and Germanium is slightly different. In my case I can compensate the difference between 3503 and MJE340 with either a Germanium or a Schottky diode
It turned out that the main source for the different compensation was that I used MJE340 on the one and 2N3503 on the other channel.
Both issues were solved, voltage is stable and no current is bypassed. It worked very well for first channel, then again the second showed a completely different behaviour.
After trying around, I finally made 8 measurements with different combinations of Q13, diode and both channels:
My conclusion:
BJTs vary in tempco slightly and this small variation is amplified by the Vbe multimplier and hence the choice of Q13 has a major influence on compensation. I also swapped the power devices between the channels and the difference was not as high as between different Q13.
The forward voltage of Schottky and Germanium is slightly different. In my case I can compensate the difference between 3503 and MJE340 with either a Germanium or a Schottky diode
Last edited:
Here is a table which is easier to read:
You see the significant difference between 3503 and MJE340, the small influence of other BJTs involved and the significant difference between germanium and schottky.
You see the significant difference between 3503 and MJE340, the small influence of other BJTs involved and the significant difference between germanium and schottky.
Very interesting, especially to learn from different tempcos of rather similar transistors. Thanks!
Two questions left (at least for me): Did you fasten your auxiliary diodes to the heat sinks, too? And how could that different behaviour of both HB specimen be explained?
Best regards!
Two questions left (at least for me): Did you fasten your auxiliary diodes to the heat sinks, too? And how could that different behaviour of both HB specimen be explained?
Best regards!
No, diode is mounted on the PCB, hence the ambient temperature has an influence. But to what extend is indeed an interesting question, thanks for the hint. Could cause some issues when the amp is mounted in an enclosure. I gotta check this.
What do you mean with HB specimen?
I just noticed that I cut the last line out of the table. I will post the complete one later today.
What do you mean with HB specimen?
I just noticed that I cut the last line out of the table. I will post the complete one later today.
Allright, HB = Honey Badger 🙂
I swapped the power devices of the channels and this causes slight change in tempco. I did not try around with the drivers, because I think I have enough evidence to say that every BJT slightly vary in tempco. I will post the tempco of the channel with different power devices (same type but just from the other channel) also today evening.
I swapped the power devices of the channels and this causes slight change in tempco. I did not try around with the drivers, because I think I have enough evidence to say that every BJT slightly vary in tempco. I will post the tempco of the channel with different power devices (same type but just from the other channel) also today evening.
Maybe the slightly different tempcos are caused by somewhat other than the junctions' behaviour? Such as different fastening torques (as done usually), different amounts (thus different thickness) of thermal compound between the devices?
Best regards!
Best regards!
Last thing that is probably worth checking.
R31 is set to 15r in all models.
This resistor is discussed by D.Self. I think Hawksford is the other author that describes this correction action.
Few other implement this correction.
The 15r adjusts the output voltage variation as current through the multiplier changes.
If there is never a change in through current, then R31 can be set to zero ohms.
But all Vbe multipliers see a change in through currents as the amp warms up and as the drivers suck out more current. This is especially important when the VAS/TIS current is sourced via a resistor.
D.Self shows a nice plot (it is on his website as well as in his Power Amp design books) that shows the effect of different R31 values for different nominal through currents.
You have to check you have the correct value for R31 to suit the through current that you want to pass. Fog 6.17 2nd ed and fig11.19 in 6th ed.
You can check this on a plug in breadboard.
Set up an output voltage (say 2.4V for a two stage EF) and pass a constant current through the multiplier.
Now adjust the constant current to lower and higher values. You want a nearly constant output voltage for changes in through current. If the output voltage (Bias Voltage) increases as through current is increased, then R31 needs to be made larger (try the next E12 value). Conversely if the bias voltage decreases as you increase the through current then decrease the value of R31. Try to find an R31 value where BV holds steady for a through current of variation of +-1mA
R31 is set to 15r in all models.
This resistor is discussed by D.Self. I think Hawksford is the other author that describes this correction action.
Few other implement this correction.
The 15r adjusts the output voltage variation as current through the multiplier changes.
If there is never a change in through current, then R31 can be set to zero ohms.
But all Vbe multipliers see a change in through currents as the amp warms up and as the drivers suck out more current. This is especially important when the VAS/TIS current is sourced via a resistor.
D.Self shows a nice plot (it is on his website as well as in his Power Amp design books) that shows the effect of different R31 values for different nominal through currents.
You have to check you have the correct value for R31 to suit the through current that you want to pass. Fog 6.17 2nd ed and fig11.19 in 6th ed.
You can check this on a plug in breadboard.
Set up an output voltage (say 2.4V for a two stage EF) and pass a constant current through the multiplier.
Now adjust the constant current to lower and higher values. You want a nearly constant output voltage for changes in through current. If the output voltage (Bias Voltage) increases as through current is increased, then R31 needs to be made larger (try the next E12 value). Conversely if the bias voltage decreases as you increase the through current then decrease the value of R31. Try to find an R31 value where BV holds steady for a through current of variation of +-1mA
the BJT tempco does not change with devices.
If you change the current through a device you will find the tempco changes for that device.
If you then compare a few small devices which have a low die area you will find they all have the same tempco at similar currents.
If you compare medium power devices in To126 packages and vary the currents for each of these you will find that devices with similar die areas have the same tempco for their range of device current.
Now if you want to compare tempco between a low power device with a medium power device you need to look at similar current densities for the devices.
A low power To92 has a range of tempco for low currents. You will find that you get a similar range of tempco for a bigger device if you test it at MUCH higher currents where the current density roughly matches that of the To92.
At first, I need to correct myself: Instead of 2N3503 I meant 2SC3503.
With regard to Andrew's post, I checked the power rating of both, MJE340 and 2SC3503. They have different power ratings. MJE340 = 20W, 2SC3503 = 7W. So according to Andrew's post, this could cause the different tempco, since current is all the same, correct?
If there is no tempco variation between the power devices, the different bias tempco could be caused by different isolation materials. I used silicone for the one channel and mica for the other.
I am wondering if a variation on bias current of lets say 20mA to 100mA per device could cause measurable tempco variation. Andrew, do you have any charts showing the dependency of current density to tempco?
With regard to Andrew's post, I checked the power rating of both, MJE340 and 2SC3503. They have different power ratings. MJE340 = 20W, 2SC3503 = 7W. So according to Andrew's post, this could cause the different tempco, since current is all the same, correct?
If there is no tempco variation between the power devices, the different bias tempco could be caused by different isolation materials. I used silicone for the one channel and mica for the other.
I am wondering if a variation on bias current of lets say 20mA to 100mA per device could cause measurable tempco variation. Andrew, do you have any charts showing the dependency of current density to tempco?
Note the significant change in tempco from 1mA to 10mA for that one device.
The actual temperature of the sensor junction will be a lot lower than the Tj of the device being monitored.
If the sensor changes by half of what the Tj changes then the tempco will have only half the effect.
This makes the installation quite different for different mounting methods.
The actual temperature of the sensor junction will be a lot lower than the Tj of the device being monitored.
If the sensor changes by half of what the Tj changes then the tempco will have only half the effect.
This makes the installation quite different for different mounting methods.
Last edited:
This demands for a small device (TO-92 or similar), connected to the PCB via thin leads, as thermal conduction through the leads is the only way it's die can have a lower temperature than it's case.
Best regards!
The heat has to flow from the device junction to the case, then across the case thickness, across the thermal interface into the heatsink, along the heatsink to the sensor thermal interface, across the thermal interface into the sensor package across the sensor case, to reach the sensor junction.
Each of those 7 routes has a thermal resistance. Each of those 7 routes will have a temperature drop due to the heat flowing.
The sensor junction must be at a lower temperature than the device junction.
- Status
- Not open for further replies.