I've been sitting here for two days and I can't find the error in my simulation. My current consumption is 52A. Therefore I'm asking for help and I'm asking for understanding.
Just search on my name and ltspice in this thread.
You could also post your .asc so we can take a look and see if anything obvious stands out.
You have two faults both on the same component.
The P channel FET has a floating gate and is also in reversed so it presents as a forward biased diode (in real life it would anyway).
The P channel FET has a floating gate and is also in reversed so it presents as a forward biased diode (in real life it would anyway).
Given you seem to have some faults in your LTspice simulation circuit - see here:
... (which presumably cause the error you found) ... I'm fascinated to know why you felt you needed to simulate the circuit in LTspice first, before considering building it?
Yes, the posted .asc was like this.
I'm not considering building it 😉 In fact searching my LTspice files showed I had in fact simulated this one as far back as 2019.
Slightly different reference numbers on mine.
No LTSpice files not available.
The circuit however is open source so there is nothing to prevent you transcribing it yourself.
It is copyright for commercial purposes and is about to be manufactured by VeraFi in July this year.
Hugh R. Dean
The circuit however is open source so there is nothing to prevent you transcribing it yourself.
It is copyright for commercial purposes and is about to be manufactured by VeraFi in July this year.
Hugh R. Dean
The p channel mosfet is driven by the output of the n channel mosfet. I expect it would exhibit more even harmonics than the odd ones. The bad thing is it could not go beyond Class A into Class B if it runs out of current.
It does have that classic falling harmonic residual of Class A and should sound excellent imo.
This is 1 watt/8ohm
This is 1 watt/8ohm
Here's another option for the Alpha Nirvana
Hugh has designed two versions of the Alpha Nirvana, one optimized for 8R loads and another for 4R or lower.
The AN8R uses 28v rails with 0.22r source resistors
The AN4R uses 20v rails with 0.12r source resistors
The Alpha Nirvana that I built has 25v rails and asymmetric source resistors:
0.22r for the upper N-channel and 0.15r for the lower P-channel of the CCS.
I noticed that by lowering the source resistor of the P-channel from 0.22r to 0.15r I could get 20v negative voltage at 4R load,
with 0.22r the negative already bottomed out at 14v at 4R load.
So this version seems to be a more allround version, somewhere between the 4R and 8R version.
I did some simulations of the three different configurations to see what the maximum power is with a 4R and 8R load at 0.1% THD
Here are the results for the three versions for 4R and 8R loads, the new variant is labelled ANxR:
The detailed sim results for 4R and 8R loads, for the three different versions at 0.1% THD
AN8R
28v 0.22-0.22 4Rload
1.83A 102W
vpp: 28.9516
power_output: 26.3022
28v 0.22-0.22 8Rload
1.83A 102W
vpp: 53.2737
power_output: 44.5411
AN4R
20v 0.12-0.12 4Rload
3.08A 121W
vpp: 34.9866
power_output: 38.2513
20v 0.12-0.12 8Rload
3.08A 120W
vpp: 35.0406
power_output: 19.1838
ANxR
25v 0.22-0.15 4Rload
2.12A 107W
vpp: 40.7525
power_output: 52.1566
25v 0.22-0.15 8Rload
2.12A 107W
vpp: 47.5202
power_output: 35.3979
For the ANxR, R142 is changed to 0.15r:
And it also sounds superb 🙂
Hugh has designed two versions of the Alpha Nirvana, one optimized for 8R loads and another for 4R or lower.
The AN8R uses 28v rails with 0.22r source resistors
The AN4R uses 20v rails with 0.12r source resistors
The Alpha Nirvana that I built has 25v rails and asymmetric source resistors:
0.22r for the upper N-channel and 0.15r for the lower P-channel of the CCS.
I noticed that by lowering the source resistor of the P-channel from 0.22r to 0.15r I could get 20v negative voltage at 4R load,
with 0.22r the negative already bottomed out at 14v at 4R load.
So this version seems to be a more allround version, somewhere between the 4R and 8R version.
I did some simulations of the three different configurations to see what the maximum power is with a 4R and 8R load at 0.1% THD
Here are the results for the three versions for 4R and 8R loads, the new variant is labelled ANxR:
| Version | Power 4R load | Power 8R load |
| AN8R | 26w | 44w |
| AN4R | 38w | 19w |
| ANxR | 52w | 35w |
The detailed sim results for 4R and 8R loads, for the three different versions at 0.1% THD
AN8R
28v 0.22-0.22 4Rload
1.83A 102W
vpp: 28.9516
power_output: 26.3022
28v 0.22-0.22 8Rload
1.83A 102W
vpp: 53.2737
power_output: 44.5411
AN4R
20v 0.12-0.12 4Rload
3.08A 121W
vpp: 34.9866
power_output: 38.2513
20v 0.12-0.12 8Rload
3.08A 120W
vpp: 35.0406
power_output: 19.1838
ANxR
25v 0.22-0.15 4Rload
2.12A 107W
vpp: 40.7525
power_output: 52.1566
25v 0.22-0.15 8Rload
2.12A 107W
vpp: 47.5202
power_output: 35.3979
For the ANxR, R142 is changed to 0.15r:
And it also sounds superb 🙂
Last edited:
Hi Danny,
Ditto for Vunce - thank you very much for the R&D and letting us the detail.
This is a great find; more power into a 4R load - usual these days for speakers - is a boon for the AN!
Now should be called the AN52!!
Ciao,
Hugh
Ditto for Vunce - thank you very much for the R&D and letting us the detail.
This is a great find; more power into a 4R load - usual these days for speakers - is a boon for the AN!
Now should be called the AN52!!
Ciao,
Hugh
Hi Danny,
I just run the sim with 0.22 for nmos and 0.15 for pmos and saw this: RED is the nmos, and BLUE is the pmos:
This is into an 8R load, but I'm confident it would look very similar with a 4R load.d
It tells us that the current range on the nmos is 1A to 2.9A, while with the pmos it is 0.6A to 3.3A.
Any mosfet changes its transconductance as it current changes, like most active devices. With any amp you need to minimise the variation of current to keep the transconductance as close as to a constant as you can. This is the reason a Class A, operating at a lesser range of current, sounds better as a linear transconductance curve creates much less higher harmonics, where the objectionable sounds arise.
Conclusion: With asymmetrical source resistors you should create less distortion because from 1A to 2.9A the transconductance is more consistent that with equal source resistors. This shows the gm/current curve for a IXTQ52N30P (a 52A 300V 400W device I found good data) and it shows close to a linear relationship:
It could be better, of course, a completely horizontal curve, but it's almost a straight line from 1A to 6A, where we are are operating it in audio, and a narrower current range is better than a large range. Ergo - less distortion, particularly of H2 and H3. It could actually sound better since it's more linear.
Thank you very much, Danny, a great find for all of us.......
Cheers,
Hugh
I just run the sim with 0.22 for nmos and 0.15 for pmos and saw this: RED is the nmos, and BLUE is the pmos:
This is into an 8R load, but I'm confident it would look very similar with a 4R load.d
It tells us that the current range on the nmos is 1A to 2.9A, while with the pmos it is 0.6A to 3.3A.
Any mosfet changes its transconductance as it current changes, like most active devices. With any amp you need to minimise the variation of current to keep the transconductance as close as to a constant as you can. This is the reason a Class A, operating at a lesser range of current, sounds better as a linear transconductance curve creates much less higher harmonics, where the objectionable sounds arise.
Conclusion: With asymmetrical source resistors you should create less distortion because from 1A to 2.9A the transconductance is more consistent that with equal source resistors. This shows the gm/current curve for a IXTQ52N30P (a 52A 300V 400W device I found good data) and it shows close to a linear relationship:
It could be better, of course, a completely horizontal curve, but it's almost a straight line from 1A to 6A, where we are are operating it in audio, and a narrower current range is better than a large range. Ergo - less distortion, particularly of H2 and H3. It could actually sound better since it's more linear.
Thank you very much, Danny, a great find for all of us.......
Cheers,
Hugh
Thanks for the analysis Hugh,
now I understand better the mechanism behind the asymmetric source resistors.
now I understand better the mechanism behind the asymmetric source resistors.
This is really neat information. Since I have put my AN39 aside for awhile due to life changes and to work on other projects, I will be following / possibly implementing this new concept into my build. Thank you Danny and Hugh for your insights.
MM
MM
Just watch out for the heat dissipation,
the original AN8R 0.22-0.22 dissipates 102W with 28v rails,
this new ANxR 0.22-0.15 dissipates 107W with 25v rails
the original AN8R 0.22-0.22 dissipates 102W with 28v rails,
this new ANxR 0.22-0.15 dissipates 107W with 25v rails
@AKSA @danny_66
Assuming that the P channel fet is FQA36P15, can we do a similar analysis (changing R142 to 0.15 ohms) if the Nchannel device is an SJEP Jfet as discussed previously?
My rails are at +/-25V and I can easily live with 107W dissipation per channel given the size of my heatsinks.
Thanks,
Anand.
Assuming that the P channel fet is FQA36P15, can we do a similar analysis (changing R142 to 0.15 ohms) if the Nchannel device is an SJEP Jfet as discussed previously?
My rails are at +/-25V and I can easily live with 107W dissipation per channel given the size of my heatsinks.
Thanks,
Anand.
Here it is, the SJEP120R100 behaves just the same:
with a 4R load and 0.15r for the CCS you can get max 20vp instead of 14vp 🙂
with a 4R load and 0.15r for the CCS you can get max 20vp instead of 14vp 🙂