Hi,
it's the high bias in the VAS that was the potential problem.
but driver transistors do not suit this location. I said 5w to 7W, I should have added that they need to be low Cob ~3pF.
Keep To92 for the LTP, but go to higher hFE amd higher fT
In the real circuit match NPN to NPN and PNP to PNP to mimic what the sim assumes.
it's the high bias in the VAS that was the potential problem.
but driver transistors do not suit this location. I said 5w to 7W, I should have added that they need to be low Cob ~3pF.
Keep To92 for the LTP, but go to higher hFE amd higher fT
In the real circuit match NPN to NPN and PNP to PNP to mimic what the sim assumes.
It looks better now, I'm still using BCs but I lowered LTP bias to 3.4/3.5 mA (it was about 4.5 mA earlier). I also increased LTP resistors a bit. At maximum output power those BCs reach now about 220 mW. That definetely looks better, and I didn't change VAS bias at all, merely adjusted the component values so that it still is about 5 mA.
EDIT: I still dropped LTP bias to 2.8/2.9 mA, and also VAS bias to 3.5 mA. Now maximum power goes up to 190 mW and that means 65 degrees if Ta=25 degrees. Is this low enough now?
EDIT #2: Stupid mistake, there is no point doing what the previous EDIT says, the max temperature drops only about 4 degrees!
EDIT: I still dropped LTP bias to 2.8/2.9 mA, and also VAS bias to 3.5 mA. Now maximum power goes up to 190 mW and that means 65 degrees if Ta=25 degrees. Is this low enough now?
EDIT #2: Stupid mistake, there is no point doing what the previous EDIT says, the max temperature drops only about 4 degrees!
Hi,
How about using BD139/140 in CCS instead of small signal transistors? They can be heatsinked and therefore kept in relatively constant temperature. If I use 2N5401/5551, I may not be able to use heatsinks, and so they do not give constant current.
I'm still considering the cascode for LTP to reduce power loss (temperature rise) of a single transistor.
How about using BD139/140 in CCS instead of small signal transistors? They can be heatsinked and therefore kept in relatively constant temperature. If I use 2N5401/5551, I may not be able to use heatsinks, and so they do not give constant current.
I'm still considering the cascode for LTP to reduce power loss (temperature rise) of a single transistor.
Hi,
After a somewhat short pause I tried to tame this circuit with an approach already suggested. I decided to use cascode in the input stage just to reduce the power loss dissipating at the small signal transistors. But after simulating the bias the offset jumped up to few hundred mV.
I deliberately left those zeners unlabeled, since I found no great difference in using either 6.8 V, 12 V or 24 V zeners - at least I couldn't get rid off that dc offset. That led me to seek help from you guys, have I constructed the cascode wrong?
I'd really appreciate if someone could help me
After a somewhat short pause I tried to tame this circuit with an approach already suggested. I decided to use cascode in the input stage just to reduce the power loss dissipating at the small signal transistors. But after simulating the bias the offset jumped up to few hundred mV.
I deliberately left those zeners unlabeled, since I found no great difference in using either 6.8 V, 12 V or 24 V zeners - at least I couldn't get rid off that dc offset. That led me to seek help from you guys, have I constructed the cascode wrong?
I'd really appreciate if someone could help me
Hi,
You have the basic cascode properly implemented.
However, I cannot tell apart the various grounds which all have the same symbol.
The output base stopper is far too big, this value would better suit the driver base.
The current limit is poorly implemented. Constant current protection will limit the peaks too early but still allow excessive dissipation in the output devices.
It looks like you are running the LTP with a higher current than the VAS. Is this deliberate and for what reason?
I suspect the output offset problem is in the PCB or just as likely mismatch between the low gain devices chosen for the LTP and maybe the VAS. Measure the voltage drops across the four 68r emitter resistors.
You have the basic cascode properly implemented.
However, I cannot tell apart the various grounds which all have the same symbol.
The output base stopper is far too big, this value would better suit the driver base.
The current limit is poorly implemented. Constant current protection will limit the peaks too early but still allow excessive dissipation in the output devices.
It looks like you are running the LTP with a higher current than the VAS. Is this deliberate and for what reason?
I suspect the output offset problem is in the PCB or just as likely mismatch between the low gain devices chosen for the LTP and maybe the VAS. Measure the voltage drops across the four 68r emitter resistors.
Bonsai, if I use 12 V zeners (for example) at input stage, and insert 1 Vp-p sine in this circuit, the load voltage will be distorted sine wave approx. 200 mVp-p. Actual waveform looks more like a triangle wave rather than sine.
Now that I cannot get proper output signal, I haven't spent my time on the current limiter part, let alone simulating stability. But before adding cascode the load voltage behaved very nicely, and while having 470nF || 8 ohm load, only slight vibration (which quickly damped out) could be seen (at 50 kHz, 1 Vp-p output).
Conclusion is that I cannot answer that question
Do you see something that would cause instability?
Andrew, using small VAS bias wasn't deliberate, I'm having about 8 mA right now.
It would also help if could find other desings with similar input stages, but so far I haven't found any. I will keep searching.
Now that I cannot get proper output signal, I haven't spent my time on the current limiter part, let alone simulating stability. But before adding cascode the load voltage behaved very nicely, and while having 470nF || 8 ohm load, only slight vibration (which quickly damped out) could be seen (at 50 kHz, 1 Vp-p output).
Conclusion is that I cannot answer that question
Do you see something that would cause instability?Andrew, using small VAS bias wasn't deliberate, I'm having about 8 mA right now.
It would also help if could find other desings with similar input stages, but so far I haven't found any. I will keep searching.
Now there has been both my mistake and misunderstanding. That schematic ver 4 does not include trimmers I added in parallel with R11 & R14. I also raised R11 & R14 to 820 ohm, and now I can match the LTP emitter currents quite precisely. Also the values of R15 & R17 are now 82 ohm and R7-R10 are 470 ohm. With these I get VAS current approx 8 mA, LTP emitter currents 3 mA and the current sources for LTP give/sink about 6 mA.
My mistake with the schematic
My mistake with the schematic
Bootstrapper, wrt to the distortion problem you have:-
check that D9 and D10 are the right way around - you should have about 11.4 volts on the emitters of your cascode transistors.
check that the cascode transistors in the LTP are ok and also you have the correct pining
Check that with no signal in, you have across th e LTP load resistors is about 1.8V.
Check your VAS cascodes as well to make sure you have all the right voltages.
Looking at your circuit, R29 and R30 should be moved to the base of Q25 and Q26 - but lets get th e front end working properly first and then tackle this part of the design.
check that D9 and D10 are the right way around - you should have about 11.4 volts on the emitters of your cascode transistors.
check that the cascode transistors in the LTP are ok and also you have the correct pining
Check that with no signal in, you have across th e LTP load resistors is about 1.8V.
Check your VAS cascodes as well to make sure you have all the right voltages.
Looking at your circuit, R29 and R30 should be moved to the base of Q25 and Q26 - but lets get th e front end working properly first and then tackle this part of the design.
Bonsai, I haven't build this one yet, I'm just simulating it.
Perhaps you can continue with those base resistors you mentioned? I have understood that with these power transistors small base resistors (one lecturer of mine said that about 20 ohm would be appropriate) can be used to reduce their tendency to oscillate. Andrew suggested smaller values previously - I will compare the values when I get the rest of the circuit to work properly.
By the way, how about using BD139/140 as drivers instead of TIPs? They are medium power transistors and so they can withstand moderate amount of current and voltage. There's no need for me to change drivers, so I may keep on with those good old TIPs.
Perhaps you can continue with those base resistors you mentioned? I have understood that with these power transistors small base resistors (one lecturer of mine said that about 20 ohm would be appropriate) can be used to reduce their tendency to oscillate. Andrew suggested smaller values previously - I will compare the values when I get the rest of the circuit to work properly.
By the way, how about using BD139/140 as drivers instead of TIPs? They are medium power transistors and so they can withstand moderate amount of current and voltage. There's no need for me to change drivers, so I may keep on with those good old TIPs.
Bootstrapper,
A value of around 3.3 Ohms is probably sufficient - I would go with this value. In your simulation, if you drive the amp into 4 or 3 ohms, you will note that distortion with a 20O base stopper resistor is mush higher than with 0 Ohms, or a low value like 3 Ohms. For this reason, the lowest value you can get away with is always best - 20Ohm in my expereince is far too high in the output base.
You need to get the basic design up and running (so no gross distortion, excessive offsets etc) so you can move to the next step which is to fine tune the amplifier compensation for stability and best dynamic performance.
A value of around 3.3 Ohms is probably sufficient - I would go with this value. In your simulation, if you drive the amp into 4 or 3 ohms, you will note that distortion with a 20O base stopper resistor is mush higher than with 0 Ohms, or a low value like 3 Ohms. For this reason, the lowest value you can get away with is always best - 20Ohm in my expereince is far too high in the output base.
You need to get the basic design up and running (so no gross distortion, excessive offsets etc) so you can move to the next step which is to fine tune the amplifier compensation for stability and best dynamic performance.
I'll keep posting my notices and the overall progress of this project even if no one would be interested in this
I'm reaching the solution soon, I simulated ac frequency response and it looked terrible! Previously I had -3dB points somewhere near 6 Hz and 180 kHz, but after all these modifications I made, now those points are at 2.3 Hz & 20.8 Hz!! So no wonder I cannot get anything rational at 1 kHz (where the simulated gain is approx -15.7 dB).
Something reduces the band and that I need to find out. It's not the filter at the input - that can be easily verified, and using those values shown at the schematic, the input filter reduces the input voltage by -1 dB. Naturally the next step is checking the feedback loop and also the Miller caps at VAS transistors. Sweeping their values over some decades showed no significant change.
As I'm writing this post, at the same time I'm using my left hand to simulate the circuit and that was really worth it: If I replace that 47u cap in the middle of VAS stage with two 15p caps so that their interconnecting point is connected to ground, the frequency sweep gets right up to where I want it to be! I'm still having that annoying 365 mV offset, so I'm focusing on that now.
I'm reaching the solution soon, I simulated ac frequency response and it looked terrible! Previously I had -3dB points somewhere near 6 Hz and 180 kHz, but after all these modifications I made, now those points are at 2.3 Hz & 20.8 Hz!! So no wonder I cannot get anything rational at 1 kHz (where the simulated gain is approx -15.7 dB).
Something reduces the band and that I need to find out. It's not the filter at the input - that can be easily verified, and using those values shown at the schematic, the input filter reduces the input voltage by -1 dB. Naturally the next step is checking the feedback loop and also the Miller caps at VAS transistors. Sweeping their values over some decades showed no significant change.
As I'm writing this post, at the same time I'm using my left hand to simulate the circuit and that was really worth it: If I replace that 47u cap in the middle of VAS stage with two 15p caps so that their interconnecting point is connected to ground, the frequency sweep gets right up to where I want it to be! I'm still having that annoying 365 mV offset, so I'm focusing on that now.
I'm on the brink of success now!
Using the circuit which I'll post later (due to minor technical difficulties), I managed to get following results:
-Offset: -7.5uV
-LTP bias: 2.984mA/3.024mA
-LTP current sources: 6mA each
-VAS bias: 9 mA
-Predriver's bias: 2.8 mA
-Drivers' bias: 16 mA
-Power Transistors: 240 mA (will make about 25 mW across 0R1)
-Worst-case LTP 5551's power loss: 40 mW -> 33 degrees C (8 + ambient)
-Total gain: +30dB
-THD @ 200 W / 8 ohm: 0.011 %
-THD @ 50 W / 8 ohm: 0.0031 %
Stability & protection incomplete, will focus on them from now on.
While calculating THD @ 200 W, I noticed that major part of distortion was 3rd harmonic, I will spend some time trying to reduce that. I may have to reconsider those driver bias currents, 15 mA sounds a bit too large and 2.8 mA a bit too small.
It's definetely looking better and better (but those results do not tell anything about the layout,wiring,PSU and so on)
Using the circuit which I'll post later (due to minor technical difficulties), I managed to get following results:
-Offset: -7.5uV
-LTP bias: 2.984mA/3.024mA
-LTP current sources: 6mA each
-VAS bias: 9 mA
-Predriver's bias: 2.8 mA
-Drivers' bias: 16 mA
-Power Transistors: 240 mA (will make about 25 mW across 0R1)
-Worst-case LTP 5551's power loss: 40 mW -> 33 degrees C (8 + ambient)
-Total gain: +30dB
-THD @ 200 W / 8 ohm: 0.011 %
-THD @ 50 W / 8 ohm: 0.0031 %
Stability & protection incomplete, will focus on them from now on.
While calculating THD @ 200 W, I noticed that major part of distortion was 3rd harmonic, I will spend some time trying to reduce that. I may have to reconsider those driver bias currents, 15 mA sounds a bit too large and 2.8 mA a bit too small.
It's definetely looking better and better
(but those results do not tell anything about the layout,wiring,PSU and so on)
Hi AndrewT!
(sorry Bootstrapper because of my post)
Which bjt can I use for Input stage? I'm also using a symmetrical amplifier and I intend to modifiy it
I'm using 2N5551/2N5401. I'm no sure that I'm using genuine devices but I can easily find many devices whose beta is match.
Can you understand me? a kind of BJT which has high beta but it's also easy to find matched-beta device, and better.
Thank you!
(sorry Bootstrapper because of my post)
Which bjt can I use for Input stage? I'm also using a symmetrical amplifier and I intend to modifiy it
I'm using 2N5551/2N5401. I'm no sure that I'm using genuine devices but I can easily find many devices whose beta is match.
Can you understand me? a kind of BJT which has high beta but it's also easy to find matched-beta device, and better.
Thank you!
You're welcome, Thanh
There's one mistake in the last circuit of mine, if I use +/- 55 volts, the TIPs will not work. Spice says the Vce voltage for them will reach nearly 100 volts (@ 40 Vp-p output = 200 W / 8 ohm) which is the absolute maximum for TIP41C/42C, so I replaced them with MJE340/350.
There's one mistake in the last circuit of mine, if I use +/- 55 volts, the TIPs will not work. Spice says the Vce voltage for them will reach nearly 100 volts (@ 40 Vp-p output = 200 W / 8 ohm) which is the absolute maximum for TIP41C/42C, so I replaced them with MJE340/350.
Hi Bootstrapper,
Few comments on your circuit if you don't mind
- Input lowpass RC forms 1Mhz -3dB filter, feedback 15k/68p is 150kHz -3db. That's not right! Change the cap value in the input to say 470p and 68p in the feedback to 33p.
- 2sc3281/2sa1302 is obsolete now, every 3281/1302 device in the market are FAKE. I'd use MJL21193/94 or 2sc5200/2sa1943 (2 pairs) with mje15030/31 drivers. mje340/350's SOA are not really enough when used as drivers.
- a cap across the Vbe multiplier would be a good idea. I prefer 100nF foil cap here. Same for the rails, 100nF foil bypass caps from rails to ground.
- your short-circuit RMS current would be 3.5A/rail. Thats way too much (3.5A×50V=175W - too much for a 150W/Tc=25°C device!). 0.22r...0.27r...0.33r as emmiter resistors would solve the problem. The bias stability of the amp would be improved too.
Best regards,
Donk
Few comments on your circuit if you don't mind
- Input lowpass RC forms 1Mhz -3dB filter, feedback 15k/68p is 150kHz -3db. That's not right! Change the cap value in the input to say 470p and 68p in the feedback to 33p.
- 2sc3281/2sa1302 is obsolete now, every 3281/1302 device in the market are FAKE. I'd use MJL21193/94 or 2sc5200/2sa1943 (2 pairs) with mje15030/31 drivers. mje340/350's SOA are not really enough when used as drivers.
- a cap across the Vbe multiplier would be a good idea. I prefer 100nF foil cap here. Same for the rails, 100nF foil bypass caps from rails to ground.
- your short-circuit RMS current would be 3.5A/rail. Thats way too much (3.5A×50V=175W - too much for a 150W/Tc=25°C device!). 0.22r...0.27r...0.33r as emmiter resistors would solve the problem. The bias stability of the amp would be improved too.
Best regards,
Donk
Update #7.
After consulting Slone's books I modified the protection circuit by adding those diodes and also 100-ohm resistors to reduce the current when protection triggers.
Plus I replaced the drivers. What comes to SC3281/SA1302 pair, I can get those genuine parts from two distributors, so I found no reason to change them.
The circuit behaves quite nicely, what I need to find out is the correct values for base resistors at the output stage. I've seen several designs which did not include them at all - I fear this is again a matter of faith.
If someone has clear vision of those base resistors, I would like to hear that! All comments are welcome
After consulting Slone's books I modified the protection circuit by adding those diodes and also 100-ohm resistors to reduce the current when protection triggers.
Plus I replaced the drivers. What comes to SC3281/SA1302 pair, I can get those genuine parts from two distributors, so I found no reason to change them.
The circuit behaves quite nicely, what I need to find out is the correct values for base resistors at the output stage. I've seen several designs which did not include them at all - I fear this is again a matter of faith.
If someone has clear vision of those base resistors, I would like to hear that! All comments are welcome
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.