I'd expect distortion is much higher at 20KHz.
A couple of thoughts:
C2 and C3 are in parallel now, giving a total of 245pF of Miller compensation. That really is very high. The problem is that the Miller cap has the full signal voltage across it and all the signal current through that cap has to be supplied by the input stage. The bigger the cap, the harder the input stage has to work at high frequencies. A smaller cap gives lower distortion and faster slew rate.
The emitter resistors in the input stage aren't providing much degeneration. The dynamic impedance of the emitters is about 20R anyway, so the 10R resistors are only reducing the transconductance by about 30%. Increasing the resistors to say 100R would give a 5-to-1 degeneration. That not only directly reduces the distortion of the input stage, it also allows you to use a much smaller Miller cap.
btw, Doesn't R8 need to be adjusted a bit to balance the currents in the LTP?
A couple of thoughts:
C2 and C3 are in parallel now, giving a total of 245pF of Miller compensation. That really is very high. The problem is that the Miller cap has the full signal voltage across it and all the signal current through that cap has to be supplied by the input stage. The bigger the cap, the harder the input stage has to work at high frequencies. A smaller cap gives lower distortion and faster slew rate.
The emitter resistors in the input stage aren't providing much degeneration. The dynamic impedance of the emitters is about 20R anyway, so the 10R resistors are only reducing the transconductance by about 30%. Increasing the resistors to say 100R would give a 5-to-1 degeneration. That not only directly reduces the distortion of the input stage, it also allows you to use a much smaller Miller cap.
btw, Doesn't R8 need to be adjusted a bit to balance the currents in the LTP?
Thanks.
Distortion @ 20kHz is 0.001%, a little higher.
I tried connecting C3 to the output of the VAS, but the bandwidth had a nasty resonant peak around 10MHz. Distortion was mostly unchanged. I put it back and it's fine now. Adding a bypass capacitor in the feedback network increases distortion.
When ever I increased the degeneration resistors, the distortion became worse, so I left them at 10ohms.
I adjusted R8 to 1.55k ohms. The currents are virtually identical in the degeneration resistors.
I know this stuff doesn't seem logical, but I'm just using the Fourier analyzer and the other tools in Tina to guide me as to how to make this circuit better. I use 0.005 seconds delay in the Fourier analyzer.
This is all assuming that the simulation is reasonably accurate.
Last edited:
Try this circuit in fig 2 on this site. The Class-A Amplifier Site - JLH Headphone Amplifiers
Excellent performance and can drive low impedances, if youre looking at very low THD replace the outputs with CFP stage. It has no harmonics beyond 2nd and thirds if designed properly with about 15ma on outputstage.
Excellent performance and can drive low impedances, if youre looking at very low THD replace the outputs with CFP stage. It has no harmonics beyond 2nd and thirds if designed properly with about 15ma on outputstage.
If only darlingtons had a fourth lead, we could get some healthy current in the first device. Integrated power darlingtons usually have a resistor across the second device's B-E, and would turn off verrrrry slowly without it.
Anyway, I will I think retire for the moment from active duty in this thread, particularly as it's getting attention from many good people, but check from time to time to see what finally comes when dirk puts something together! I might post a highly non-minimal (what would be the rough opposite of Zen? Vajrayana?
) circuit just to occasion scorn.Brad Wood
If only darlingtons had a fourth lead, we could get some healthy current in the first device. Integrated power darlingtons usually have a resistor across the second device's B-E, and would turn off verrrrry slowly without it.
Anyway, I will I think retire for the moment from active duty in this thread, particularly as it's getting attention from many good people, but check from time to time to see what finally comes when dirk puts something together! I might post a highly non-minimal (what would be the rough opposite of Zen? Vajrayana?
Brad Wood
OK, thanks very much for your much valued contributions. I've learned more in the last few weeks on this subject than I have over the last 10 years.
Our sim results often seem to be totally different. 
But back to basics...
If the currents are balanced, then the voltage across R8 is 1.5mA * 1.55K = 2.325V for the circuit on post 140. For the circuit in post 147, it's 1.75mA * 1.33K = 2.3275V.
That seems too low. According to the MPSA27 datasheet, the voltage between it's base and emitter should be about 1.15V, meaning there's only about 1.2V across the LED. With any normal LED, the voltage drop would be much higher.
In your simulation, what is the voltage across LED3, and between the base and emitter of T3?
btw, what is CQX35A? I can't find a datasheet for it and it isn't stocked by Mouser, Digikey or RS components.
P.S. How much current is flowing in the VAS and the output stage now?
But back to basics...
Something seems to be wrong here.
If the currents are balanced, then the voltage across R8 is 1.5mA * 1.55K = 2.325V for the circuit on post 140. For the circuit in post 147, it's 1.75mA * 1.33K = 2.3275V.
That seems too low. According to the MPSA27 datasheet, the voltage between it's base and emitter should be about 1.15V, meaning there's only about 1.2V across the LED. With any normal LED, the voltage drop would be much higher.
In your simulation, what is the voltage across LED3, and between the base and emitter of T3?
btw, what is CQX35A? I can't find a datasheet for it and it isn't stocked by Mouser, Digikey or RS components.
P.S. How much current is flowing in the VAS and the output stage now?
Last edited:
Our sim results often seem to be totally different.
But back to basics...
Something seems to be wrong here.
If the currents are balanced, then the voltage across R8 is 1.5mA * 1.55K = 2.325V for the circuit on post 140. For the circuit in post 147, it's 1.75mA * 1.33K = 2.3275V.
That seems too low. According to the MPSA27 datasheet, the voltage between it's base and emitter should be about 1.15V, meaning there's only about 1.2V across the LED. With any normal LED, the voltage drop would be much higher.
In your simulation, what is the voltage across LED3, and between the base and emitter of T3?
btw, what is CQX35A? I can't find a datasheet for it and it isn't stocked by Mouser, Digikey or RS components.
P.S. How much current is flowing in the VAS and the output stage now?
That's not encouraging that our sim results are totally different.
Voltages:
across LED3: 1.56V
base of T3: -29.66V
output of VAS: 351.83mV
current in R5: 1.76mA
current in R10: 1.77mA
current in R6: 12.43mA
Current in R1: 12.47mA
current in R2: 12.37mA
across LED1, LED2, LED4: 1.47V
make sure your constant current sources are set to 3.5mA
That LED is just a standard one, nothing special about it. I looked it up.
I have a question about the coupling caps. If I do the math, I only need about 4.7uF, but the Fourier analyzer gives much higher distortion if I use those values and the distortion goes down if I increase them to something large. I'm picking parts now and I was wondering if I'm supposed to believe the Fourier analyzer or not when it comes to this.
Attachments
So with 32V supply rails, that means there's only 0.78V between the base and emitter of T3? For a Darlington? Tina must be pretending MPSA27 is just a normal single transistor. Spice models are often bad, but that's ridiculous.
Any links? Did you find a datasheet?
Small caps are fine, that's just a problem with the simulation.
The datasheet for the MPSA27 I think shows this is fine.
The LED is the same as the Vishay TLUR5100, apparently. I looked at the datasheet and it's just a normal red LED. I've selected this one:
http://www.mouser.com/Search/ProductDetail.aspx?R=HLMPD155Avirtualkey63810000virtualkey638-HLMPD155A
OK, thanks on the caps. I'm trying to fit them on the board.
Last edited:
I don't understand the parameters either, but it seems to be using the same model for MPSA27 as for e.g. TIP31C and MPS3704, which are ordinary single transistors, so nevermind the parameter values, it looks like it's not even using the right model.
First pic below is from the MPSA27 datasheet, indicating Vbe should be nearly 1.2V, not 0.78V.
Meh, I couldn't find TLUR5100 either, but looked up the HLMPD155A you're using. Second pic below is from it's datasheet. Looks like it will have about 1.75V across it @12mA.
Bad Tina, Bad!
Attachments
That makes no sense. You could leave the caps out altogether in the sim, so there's one less excuse. Increasing the delay might help too, maybe to 100mS, i.e. a few periods of the waveform at 20Hz.
I'm starting to think you might be better off dumping Tina and getting different software. LTSpice is good, but horribly difficult to use. Simetrix is good and easy to use.
I started out with the "student version" of Circuitmaker, but eventually realized that some of the answers it was giving me were total science fiction. Presumably "student version" really meant "buggy prerelease beta version".
After that, I tried LTSpice for a while - no fun at all, the user interface really is a nightmare. Finally I found Simetrix/Simplis, and have been using the free version of that ever since.
Simming is fun again and I get sensible answers.
The graph you posted shows the collector-emitter voltage when the transistor is saturated. When that happens the collector voltage is lower than the base voltage, so even when the collector-emitter voltage is below 0.7V, the base-emitter voltage will be above 1V. Actually the base-emitter voltage will be a bit higher than when the transistor's not saturated.
All that's academic though. In your circuit none of the transistors get anywhere near saturation.
I looked at the models of the BC560C in Tina and in SIMetrix and they are different. I mean the values are different. They are in the same ball park, but I don't know if that's good. I couldn't find the model of the MPSA27 in SIMetrix. I like the interface of Tina better than SIMetrix, but I suppose I can get used to it.
If this simulator is really far off the mark, then I've just wasted a week of experimenting. I'm not happy.
If this simulator is really far off the mark, then I've just wasted a week of experimenting. I'm not happy.
Last edited:
@ Bonsai:
Well, my experience was something like:
Step 1: Place a component on the work area.
Step 2: Waste 10 minutes digging through the help to find out how to drag it to a different position.
Step 3: At least half an hour to figure out how to rotate a component.
I think it was days before I figured out that you have to do something daft like "CTRL ALT right click" to change the properties of a component.
Then when you finally get the circuit laid out, you have to type in SPICE commands, in plain text, straight onto the GUI, to get the simulator to actually do anything.
I actually did use it for a while with some success, but then didn't touch it for a month or so. When I went back, I'd forgotten everything, really didn't want to go through the whole learning curve again, and went looking for other software instead.
Below is a screenshot of Simetrix in action - that's the kind of user interface I can cope with!
Well, my experience was something like:
Step 1: Place a component on the work area.
Step 2: Waste 10 minutes digging through the help to find out how to drag it to a different position.
Step 3: At least half an hour to figure out how to rotate a component.
I think it was days before I figured out that you have to do something daft like "CTRL ALT right click" to change the properties of a component.
Then when you finally get the circuit laid out, you have to type in SPICE commands, in plain text, straight onto the GUI, to get the simulator to actually do anything.
I actually did use it for a while with some success, but then didn't touch it for a month or so. When I went back, I'd forgotten everything, really didn't want to go through the whole learning curve again, and went looking for other software instead.
Below is a screenshot of Simetrix in action - that's the kind of user interface I can cope with!
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.