Hi AAK,
Cool. looking good.
Hi darkfenriz,
If you remove that transistor, one side will have 2X the voltage on average as the other. It will therefore dissipate just under 2X the power and be much warmer than the other transistor.
Look again. It isn't really being used as a cascode. It's main purpose is to equalize the average voltages. The cascode effect is unavoidable.
-Chris
Cool. looking good.
Hi darkfenriz,
If you remove that transistor, one side will have 2X the voltage on average as the other. It will therefore dissipate just under 2X the power and be much warmer than the other transistor.
Look again. It isn't really being used as a cascode. It's main purpose is to equalize the average voltages. The cascode effect is unavoidable.
-Chris
Hi Clem,
One always hopes those are taken across the load. You have a big PM headed your way.
AAK,
Depending on your outputs, the compensation cap must be reduced. I'm thinking you might try 5 ~ 7 pF. Mica. A trim cap would allow you to "dial" it right in. Also try to throw some capacitance across your resistor and check again.
-Chris
One always hopes those are taken across the load. You have a big PM headed your way.
AAK,
Depending on your outputs, the compensation cap must be reduced. I'm thinking you might try 5 ~ 7 pF. Mica. A trim cap would allow you to "dial" it right in. Also try to throw some capacitance across your resistor and check again.
-Chris
anatech said:
Hi Chris,
True - but if it's scoped at the load side, it could simply be showing a ring as a result of the inductor driving a slightly capacitive load, and it could look perfectly square on the amplifier side - i.e. perfectly stable.
Got the pm - waiting for it to complete - many thanks!!!
Cheers
Clem
AAK said:
Hi AAK, nice work, but where is the input filter ? I mean the 2k in series with the input. As i can see, you took the outputstage from the AKSA, it might be needed to evaluate that change. I had only 22ohms basestoppers to the drivers because of speed issues.
Very compact layout !
Tim__x said:
Yes, that's a cascoded cascode, i kept that as Chris said, for thermal issues. As these devices are only to92, they run hot quickly, and having double the dissipation, you run quickly into limits. I just don't like having devices run too hot to touch, that's all.
For now, i'd like to keep the smallish to92, they are easy to get, cheap and fast.
Mike
You might want to try something like 2sa1538 or 2sa1360 in TO126 caseing instead of 2n5401.
These Japanese BJTs can dissipate more heat and have no worse specifications, wonderfully low capacitances and high ft.
...or leave the cascode or swap the casode with zener diode for loewr dissipation, that's your amp, your choice.
Good luck, nice to see the amp develop.
These Japanese BJTs can dissipate more heat and have no worse specifications, wonderfully low capacitances and high ft.
...or leave the cascode or swap the casode with zener diode for loewr dissipation, that's your amp, your choice.
Good luck, nice to see the amp develop.
Hi darkfenriz,
Using a higher power device still allows one to run almost twice as hot as the other. That is my central point. The TO-92 devices would be a cost accounting decision in production anyway.
This transistor is easy to replace with a jumper. Why not build one? Measure the DC offset, then remove the transistor and jump the location. Measure the DC offset again. Also, monitor the offset as the amplifier warms for for, say 15 min. with each version. Allow the ampo to completly cool between runs. Use the 2SA1538 or 2SA1360 in your version.
There are many commercial consumer amplifiers that configure that circuit the same way. To my way of thinking, there just has to be a good reason for this. Who knows. Could just be production costs.
Hi Clem,
That ring should show up eveywhere. If you ran a frequency response test way up, I'll bet you get a rising response at some frequency. I had this problem on another design of mine. The output had a burst of oscillation on one polarity only at a certian voltage on the dummy load. In my case it was around 360 KHz. It turned out to be one transistor dynamically current starved. How about that? Increasing the operating current solved the problem.
-Chris
Using a higher power device still allows one to run almost twice as hot as the other. That is my central point. The TO-92 devices would be a cost accounting decision in production anyway.
This transistor is easy to replace with a jumper. Why not build one? Measure the DC offset, then remove the transistor and jump the location. Measure the DC offset again. Also, monitor the offset as the amplifier warms for for, say 15 min. with each version. Allow the ampo to completly cool between runs. Use the 2SA1538 or 2SA1360 in your version.
There are many commercial consumer amplifiers that configure that circuit the same way. To my way of thinking, there just has to be a good reason for this. Who knows. Could just be production costs.
Hi Clem,
That ring should show up eveywhere. If you ran a frequency response test way up, I'll bet you get a rising response at some frequency. I had this problem on another design of mine. The output had a burst of oscillation on one polarity only at a certian voltage on the dummy load. In my case it was around 360 KHz. It turned out to be one transistor dynamically current starved. How about that? Increasing the operating current solved the problem.
-Chris
Hi Mike,
I was wondering about the input filter. Why did you use 2.2K and 100pF, 3db at 725KHz? In sims it didn't seem to make much of difference in or out with a 2N5551 at the input. Being as compact as the board is (amp section measures 2.5" x 3") I left it out. I figured the cap across the feedback resistor would provide enough filtering. From the square wave pic you can see a small overshoot but overall I think it's Ok. The amps 3db down at 1watt into 8ohm is about 550Khz so an LP on the front end may still be necessary..
Al
I was wondering about the input filter. Why did you use 2.2K and 100pF, 3db at 725KHz? In sims it didn't seem to make much of difference in or out with a 2N5551 at the input. Being as compact as the board is (amp section measures 2.5" x 3") I left it out. I figured the cap across the feedback resistor would provide enough filtering. From the square wave pic you can see a small overshoot but overall I think it's Ok. The amps 3db down at 1watt into 8ohm is about 550Khz so an LP on the front end may still be necessary..
Al
Leaving out the 2k at the input was not a good idea, if the amp gets fed by low impedance source, the filter does not work at all. This can even lead to stability problems.
The filter is mainly needed to keep hf out (several mhz), otherwise you risk SID. (Slew induced distortion)
I choosed 100pf only to keep input impedance nearly flat in audioband.
BTW, sims do not tell all... The complete input filter is mandatory. Please add it somehow, it's also part of the feedback compensation.
Mike
The filter is mainly needed to keep hf out (several mhz), otherwise you risk SID. (Slew induced distortion)
I choosed 100pf only to keep input impedance nearly flat in audioband.
BTW, sims do not tell all... The complete input filter is mandatory. Please add it somehow, it's also part of the feedback compensation.
Mike
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.