Hi Nico,
Do me a favour and look at the PSR as detailed in the posts above and see if you get the same results.
Thanks.
Do me a favour and look at the PSR as detailed in the posts above and see if you get the same results.
Thanks.
MJL21193 said:
John,
Okay I optimized two resistors, try changing R3 to 120 Ω and R4 to 33KΩ.
THD is now okay at 0.0028@1KHz.
PSR is 70 dB (I place only ripple on one power line so it is asymmetrical)
I am still not quite happy with the temperature stability but both NPN and PNP now tracks each other perfectly.
I will look see what can be done with temperature, it may be selecting another Vbe multiplier. I used MPSA06, which gave me the better results but it is difficult to mount on the heat sink.
Regards
Nico
Okay solved,
use MJE15030 as Vbe multiplier and the thermal problem is solved. Now current reduces with increased temperature.
Have fun,
Nico
use MJE15030 as Vbe multiplier and the thermal problem is solved. Now current reduces with increased temperature.
Have fun,
Nico
Nico Ras said:
Thank you Nico,
Did you check the idle current? For this amp, best THD performance seems to be in the 70-90mA range per output. I'm seeing 0.001% THD at 1k at low power and 0.005% at near full output power (100 watts).
So you are not seeing a huge decrease in PSR with the unbalanced LTP? Strange.
Nico Ras said:
My sim doesn't do temp, but the real amp does. I have rock steady idle current using a BD139 for the Vbe. Of my working modules I have running my speakers, I have the tweeter amps idling near 100mA per output and no signs of thermal runaway.
You might want to check the resistor values in the Vbe multiplier. I've had nothing but troubles with this area of simulation.
Nico Ras said:Okay John, time to design your next amplifier, this one is put to bed.
There are always things to try and improvements to make. In this round, I increased stability. eliminated overshoot and reduced the parts count slightly. It's ready to be put to bed...for now 🙂
You want an amp to thermally behave. If it does not then mechanical fixes are required which is much more expensive than electronic fixed.
Microcap has the ability to optimize several components at a time finding best match for worst case. It is very helpful, but takes time.
I am running a bias current of 58 mA producing the proverbial just less than 25mV across Re. This produces 3.1 watt dissipation per device which is fine, there is no cross-over distortion detected.
Now that I understand the mechanism of you current mirror, it seems to work quite alright.
All in all, this is quite a nice amp and another one that you can offer to the blokes on the block.
I do not see the huge PSR as you say. Remember changing those two components brought everything into equilibrium, LTPs are now passing equal currents and the mirrors are working as they should.
IMO you can remove the two 4.7 ohm resistors, they do nothing to improve or worsen matters.
I added a 1 K pot between the Re's and tail of the LTP that so you can adjust the final off-set to zero.
The 200K feedback into the LPT tail does nothing, I would remove it.
Finally there is ever so slight improvement in THD connecting the bootstrap capacitor to the usual output.
In summary remove R28, R1 and R35. This simplifies the amp a little. reconnect capacitor C12 to the output and change R 2 to 120 ohm and R3 to 33 K. Place a 1K 10 turn pot between the tail resistors R23 & R24 with the slider connected to the junction of R27 and you are done, a relatively simple but very nice amplifier.
Kind regards
Nico
Microcap has the ability to optimize several components at a time finding best match for worst case. It is very helpful, but takes time.
I am running a bias current of 58 mA producing the proverbial just less than 25mV across Re. This produces 3.1 watt dissipation per device which is fine, there is no cross-over distortion detected.
Now that I understand the mechanism of you current mirror, it seems to work quite alright.
All in all, this is quite a nice amp and another one that you can offer to the blokes on the block.
I do not see the huge PSR as you say. Remember changing those two components brought everything into equilibrium, LTPs are now passing equal currents and the mirrors are working as they should.
IMO you can remove the two 4.7 ohm resistors, they do nothing to improve or worsen matters.
I added a 1 K pot between the Re's and tail of the LTP that so you can adjust the final off-set to zero.
The 200K feedback into the LPT tail does nothing, I would remove it.
Finally there is ever so slight improvement in THD connecting the bootstrap capacitor to the usual output.
In summary remove R28, R1 and R35. This simplifies the amp a little. reconnect capacitor C12 to the output and change R 2 to 120 ohm and R3 to 33 K. Place a 1K 10 turn pot between the tail resistors R23 & R24 with the slider connected to the junction of R27 and you are done, a relatively simple but very nice amplifier.
Kind regards
Nico
BTW square wave response is absolutely clean and slew is calculated at about 22V/uS.
So now you have a better amp than the previous patchwork. Are you going to bin the old one?
N
So now you have a better amp than the previous patchwork. Are you going to bin the old one?
N
Hi Nico, how did you come to the values of R2 and R3 ?? Is it to balance the LTP?? Is the ltp unbalanced with the values MJL21193 used??
Nico Ras said:
Hi Nico,
Multisim has a similar utility, though I haven't used it. With that said, some of your recommendations go against my own findings and the advice I've receive during the design process.
For instance:
Nico Ras said:
These are part of an RC network on the rails of the front end to improve the PSRR of the amp. My intention was to actually increase these rail resistors to 22 ohm. They also provide some isolation for the front end, discouraging large current from flowing from the output stage into the front end.
And:
Nico Ras said:
The "bootstrap" resistor has demonstrated a positive effect on THD in my sim, lowering it at higher frequency. It does very little at low frequency but does not have a negative effect. Any reduction in high frequency THD is to be welcomed though.
Again, I found the complete opposite to be true regarding the placement of the bootstrap cap. When it was suggested by jcx, I immediately implemented it and it brought about a fairly large reduction in THD. I'm not sure in what part of the spectrum it was. I'll run the sim to see.
As for offset, I don't have a problem in the real amp. Measured offset is less than 10mV in all of the modules I've built so far (mostly with unmatched input pairs). It could prove to be a problem with the revised current mirror, but I have decided to abandon that. Further simulations have shown the benefits are not as big as I initially thought.
Thanks for taking the time to run this. It is good to have some verification from another simulator for comparison.
homemodder said:
Hi homemodder,
I'm not Nico, but the purpose of that mirror was for it to be unbalanced. If you look at the schematic from which it came, you'll see I configured it the same way. It was this imbalance that seemed to give the benefits I was seeing (lower THD at high output).
I have, as I've mentioned above, gone back to the standard Widlar mirror to balance the LTP.
Hi John, with this improvement, how close now is the Patchwork to your Abomination (I think you should have called it Abominable, sounds funkier, like Abominable Snowman)
🙂
🙂
MJL21193 said:
Hmm, the schematic above doesn't reflect the return to the balanced state. I'm confused 🙂.
Edit: Hey, you changed it back! Haha.
KLe said:
Hi KL,
They are actually not that far apart, raw (simulated) spec wise. The Abomination looks more formidable and is definitely faster, but the trade off if higher complexity. I like simple things best.
Anyway, Abomination is a noun and abominable is an adjective 😀
andy_c said:
Yes, I wised up. It is now a pretty good, simple, low distortion amp with some sprinkles of magic to make it sound Golden. 🙂
MJL21193 said:
It can't possibly sound good if it doesn't have a French name, like "Le Monstre". And where are the boutique capacitors, boutique resistors and so on? Does the chassis have boutique screws? 🙂
andy_c said:
I considered renaming it after myself - L'un Misérable. Doesn't 'ave a good ring to et! 🙂
With the proper French title, we could step up to high end bolts and such. One must use the caps and cheap resistors I have chosen, as it is the "voice" of the amp. To replace any of the parts that I have personally selected is to run the risk of harshness and fatigue from listening!
I have been meaning to ask you Andy about the loop gain and such.
Looking at the plots for this amp, I have the open loop gain where it should be - 0db at -100 degrees:
And I remember reading somewhere that the closed loop gain should look like this:
That is the phase should not swing 180 degrees before the gain hits 0db. is this correct?
Looking at the plots for this amp, I have the open loop gain where it should be - 0db at -100 degrees:
And I remember reading somewhere that the closed loop gain should look like this:
That is the phase should not swing 180 degrees before the gain hits 0db. is this correct?
Just a bit of pedantry here. You have the open-loop gain AOL, which is the ratio of the output voltage to the difference-mode input voltage (between inverting and non-inverting inputs of the diff amp). Then there is the feedback factor B, which is a resistive divider R26/(R26 + R22) - assuming C11 is a short and neglecting R25. Then there is the loop gain AOLB. It's the magnitude and phase properties of the loop gain AOLB that determine small-signal stability. The loop gain AOLB is what you're (approximately) plotting with that funky floating AC voltage source in the feedback loop. So the "loop gain" and the "open-loop gain" are different things (by a factor of B).
Okay, the phase margin is the phase shift of the loop gain at the unity loop gain frequency minus a negative 180 deg. So with a phase shift of -100 deg at the ULG freq, you have a phase margin of 80 deg (-100 -(-180)). Phase margins less than approximately 78 deg give overshoot and/or ringing in the small-signal square wave response (assuming the input LPF is disabled). So with an 80 deg phase margin, you are fine here.
Don't worry about the phase shift of the closed-loop gain. One useful approximation is that if the phase margin is near 90 degrees, the -3 dB frequency of the closed-loop gain (with input filter disabled) is approximately equal to the unity loop gain frequency.
Oh, and one other thing. There's a spec called the gain margin. To find it, find the frequency at which the phase shift of the loop gain AOLB is -180 deg. Then find the magnitude of the loop gain at that freq. The negative of that value is the gain margin. It should be, say, 10 dB or more. In your case, it looks like the phase shift of the loop gain is -180 deg at 7 MHz. The magnitude of the loop gain looks like it's about -20 dB there. So you have a 20 dB gain margin. That's about as good as it gets. The gain margin is only a concern if you're aggressively trying to make the ULG freq as high as possible for maximum feedback. If the magnitude has some sort of peaking at very high frequencies, you can run into the case where the phase margin is fine but the gain margin sucks. But again, that's only a problem with aggressive high-feedback designs.
Okay, the phase margin is the phase shift of the loop gain at the unity loop gain frequency minus a negative 180 deg. So with a phase shift of -100 deg at the ULG freq, you have a phase margin of 80 deg (-100 -(-180)). Phase margins less than approximately 78 deg give overshoot and/or ringing in the small-signal square wave response (assuming the input LPF is disabled). So with an 80 deg phase margin, you are fine here.
Don't worry about the phase shift of the closed-loop gain. One useful approximation is that if the phase margin is near 90 degrees, the -3 dB frequency of the closed-loop gain (with input filter disabled) is approximately equal to the unity loop gain frequency.
Oh, and one other thing. There's a spec called the gain margin. To find it, find the frequency at which the phase shift of the loop gain AOLB is -180 deg. Then find the magnitude of the loop gain at that freq. The negative of that value is the gain margin. It should be, say, 10 dB or more. In your case, it looks like the phase shift of the loop gain is -180 deg at 7 MHz. The magnitude of the loop gain looks like it's about -20 dB there. So you have a 20 dB gain margin. That's about as good as it gets. The gain margin is only a concern if you're aggressively trying to make the ULG freq as high as possible for maximum feedback. If the magnitude has some sort of peaking at very high frequencies, you can run into the case where the phase margin is fine but the gain margin sucks. But again, that's only a problem with aggressive high-feedback designs.
Thanks Andy,
You have made a real difference by taking the time to explain these thing to us. I appreciate it. 🙂
I need to collect these nuggets of information and save them somewhere for a quick handy reference.
You have made a real difference by taking the time to explain these thing to us. I appreciate it. 🙂
I need to collect these nuggets of information and save them somewhere for a quick handy reference.
MJL21193 said:
I'm not homemodder, but I do have a question. Did the lower THD at high output feature work in practice in any way that affected the amplifier's presentation or increase the likelihood that it would perhaps be used at high output?
- Status
- Not open for further replies.