Okay, My PCB panel is now in at the manufacturer to made.
Something interesting just happened with my prototype though, I accidently powered it up without the Function generator on (it provides my triangle wave)...
And it runs great! Weird!. However the frequency has jumped up to about 290kHz. It sounds great, nothing gets hot etc... However the maximum volume has been reduced. I am thinking that maybe i won't need to populate the triangle wave generating section of my pcbs once they arrive... What do other people think??
Regards
Peter.
Something interesting just happened with my prototype though, I accidently powered it up without the Function generator on (it provides my triangle wave)...
And it runs great! Weird!. However the frequency has jumped up to about 290kHz. It sounds great, nothing gets hot etc... However the maximum volume has been reduced. I am thinking that maybe i won't need to populate the triangle wave generating section of my pcbs once they arrive... What do other people think??
Regards
Peter.
LOL.... bonus! Self oscillating, much better. Would you dare to say it sounds even better?
That's almost a decent frequency for a full range..
It might be worth analysing your simulation of it to see if everything on it is OK and if it is hell yeh go for it. A component value or two might need tweaking but if it works it works, go with it.
Regards
That's almost a decent frequency for a full range..
It might be worth analysing your simulation of it to see if everything on it is OK and if it is hell yeh go for it. A component value or two might need tweaking but if it works it works, go with it.
Regards
LOL that was a quick response.
I don't know if i could say it sounds better, but it sounds equally as good, I can't pick any faults with my ears with either setup...
But i am pretty keen to keep it self-oscillating. It has no problems starting etc no pops or thumps, weird huh.. but it sounds good and is simpler!
It has no problems getting the output to clip, it clips pretty cleanly too. So the volume is loud enough..
Output ripple is about 1.5v peak - peak now....
What would you suggest i should look at tweaking?
Peter
I don't know if i could say it sounds better, but it sounds equally as good, I can't pick any faults with my ears with either setup...
But i am pretty keen to keep it self-oscillating. It has no problems starting etc no pops or thumps, weird huh.. but it sounds good and is simpler!
It has no problems getting the output to clip, it clips pretty cleanly too. So the volume is loud enough..
Output ripple is about 1.5v peak - peak now....
What would you suggest i should look at tweaking?
Peter
Yeh I'd just turned the computer on again.
Well then you're off to a great self oscillating start I think. It is weird, that it would work so good.
I'm just thinking you didn't design it to be self oscillating and so it stands to reason it should be optimised with this in mind.
Step 1 I guess is figure out why and how it is self oscillating (I'm guessing phase shift) and then go from there, you'll probably find a few things needed some tweaking. Things worth having a look at, loop gains (closed/open) ummmmm.... phase margins.. I don't know, I think Charles can be more help here
I think once you do any required adjustments it might sound even better! Keep us posted.
Regards
Chris
Well then you're off to a great self oscillating start I think. It is weird, that it would work so good.
I'm just thinking you didn't design it to be self oscillating and so it stands to reason it should be optimised with this in mind.
Step 1 I guess is figure out why and how it is self oscillating (I'm guessing phase shift) and then go from there, you'll probably find a few things needed some tweaking. Things worth having a look at, loop gains (closed/open) ummmmm.... phase margins.. I don't know, I think Charles can be more help here
I think once you do any required adjustments it might sound even better! Keep us posted.
Regards
Chris
If for example i wanted to slow the switching frequency down, would i just increase the intergrator capacitor value??
Actually this works way better than i initally thought, infact the fets run much cooler at high power. this i beleive is due to the switching actually slows down quite significantly for large swings at the output. But for small signals the switching frequency is higher giving higher fidelity. Hmmmm from what i am seeing so far i like!!.
Actually this works way better than i initally thought, infact the fets run much cooler at high power. this i beleive is due to the switching actually slows down quite significantly for large swings at the output. But for small signals the switching frequency is higher giving higher fidelity. Hmmmm from what i am seeing so far i like!!.
Yeh I think that would slow it, it will also affect your gain. Charles will be much more help in this area. A series resistor is another option.
It does make sense for it to run cooler by switching slower at higher outputs, but you don't want Fs to vary by a great deal either.
If it works good at 290Khz you may consider leaving it at that anyway, less residual ripple on the output and everything, it's easier to filter. Also as you said, it does vary somewhat and having it higher helps ensure it can't go too low. Mosfets should easily handle this speed.
Are you just leaving the non inverting input of your comparator floating or have you tried grounding it yet?
I notice you're taking feedback pre filter, there's a good tweak for you to attempt, try taking it after the filter, you may as well take full benefit from the self oscillating scheme. No doubt that will force a few values around your integrator to change. Should be safe enough for a simulator to perform anyway.
Regards
It does make sense for it to run cooler by switching slower at higher outputs, but you don't want Fs to vary by a great deal either.
If it works good at 290Khz you may consider leaving it at that anyway, less residual ripple on the output and everything, it's easier to filter. Also as you said, it does vary somewhat and having it higher helps ensure it can't go too low. Mosfets should easily handle this speed.
Are you just leaving the non inverting input of your comparator floating or have you tried grounding it yet?
I notice you're taking feedback pre filter, there's a good tweak for you to attempt, try taking it after the filter, you may as well take full benefit from the self oscillating scheme. No doubt that will force a few values around your integrator to change. Should be safe enough for a simulator to perform anyway.
Regards
Hello Charles,
The schematic i posted with my pcb files is current. I was wondering that as well, but i assure you it is running very well.
Maybe the frequency is down because i am only using a TL072 op-amp, which is reasonably slow.... It could also be set by the 1st order lag which is set up by R16 & C18 in the divider network..... What do you think?
I injected a sine wave into the input, such that the output was 80v peak 56v RMS into a 2.3ohm load for 1 hour.....Thats over 1300W RMS. It actually runs significantly cooler than when modulated by an external Triangle Wave from my fuction generator...
All i have done is tie the Triangle Wave input to ground....
The schematic is at http://tesla.reidconsulting.com.au/Fixed.pdf
Regards
Peter
The schematic i posted with my pcb files is current. I was wondering that as well, but i assure you it is running very well.
Maybe the frequency is down because i am only using a TL072 op-amp, which is reasonably slow.... It could also be set by the 1st order lag which is set up by R16 & C18 in the divider network..... What do you think?
I injected a sine wave into the input, such that the output was 80v peak 56v RMS into a 2.3ohm load for 1 hour.....Thats over 1300W RMS. It actually runs significantly cooler than when modulated by an external Triangle Wave from my fuction generator...
All i have done is tie the Triangle Wave input to ground....
The schematic is at http://tesla.reidconsulting.com.au/Fixed.pdf
Regards
Peter
My guess is both, along with all the other delays, and I guess the time constant of your integrator as well.
Did the frequency change in any way once you grounded the input of the comparator?
Also, how much does your frequency tend to vary, and what was it originally, 100Khz??
You know another option you have available is to add some hysteresis from the output to the comparator
Regards
The lowest frequency i have measured so far is about 150kHz. Initially it was running at 250kHz from the triangle wave.
Yes the frequency increased when i disconnected the input from my function gen ( 50ohm impendance) and just connected it straight to ground, I think about 5kHz higher by memory.
290khz - > 295kHz. I could add hysteresis, but i can't really see the point.
Regards
Peter
Yes the frequency increased when i disconnected the input from my function gen ( 50ohm impendance) and just connected it straight to ground, I think about 5kHz higher by memory.
290khz - > 295kHz. I could add hysteresis, but i can't really see the point.
Regards
Peter
Hi,
295Khz from 250Khz isn't that much of a jump when you consider it's working by a completly different mechanism now.
I'd be more concerned with the drop to 150Khz. Whatever the cure will be for that (I don't know offhand) I doubt it will be a big deal.
Hysteresis will add extra gain in your feedback loop I believe, and some overall complication. I wouldn't worry about it now but maybe one day when you're bored again.. It's just an option you have with it, that's all. Your circuit seems kind of bare bones (which is good) and as such is easily morphed into working differently as we've found. I like it.
Regards
chris
295Khz from 250Khz isn't that much of a jump when you consider it's working by a completly different mechanism now.
I'd be more concerned with the drop to 150Khz. Whatever the cure will be for that (I don't know offhand) I doubt it will be a big deal.
Hysteresis will add extra gain in your feedback loop I believe, and some overall complication. I wouldn't worry about it now but maybe one day when you're bored again.. It's just an option you have with it, that's all. Your circuit seems kind of bare bones (which is good) and as such is easily morphed into working differently as we've found. I like it.
Regards
chris
Have you taken measurements of quality (THD, noise) with both working modes in order to compare them?
With selfosc designs you can have wide freq. deviations, that I don't like at all, when it gets very low you have much more ripple. And when you connect several amplifiers together, fun is guaranteed!!!
Best regards
Pierre
With selfosc designs you can have wide freq. deviations, that I don't like at all, when it gets very low you have much more ripple. And when you connect several amplifiers together, fun is guaranteed!!!
Best regards
Pierre
Pierre,
No i havn't yet, although i will do this with the modules when i get the pcb's back.......
I agree, things may get interesting when i bridge two modules. But it may work okay too.. Worst comes to worst all i have to do is populate the triangle wave generator section of the pcb and its back to carrier based, so there really isn't anything to lose.
Regards
Peter.
No i havn't yet, although i will do this with the modules when i get the pcb's back.......
I agree, things may get interesting when i bridge two modules. But it may work okay too.. Worst comes to worst all i have to do is populate the triangle wave generator section of the pcb and its back to carrier based, so there really isn't anything to lose.
Regards
Peter.
Hi Peter,
Yep, makes good sense, that's when the amplitude of the feedback loop stops and it loses control, possibly oscillating by circuit delays alone?
Seems like there's nothing to fix so far, you got yourself a nice phase shift oscillator, neat
You're next Pierre... come to the dark side
Regards,
Chris
Yep, makes good sense, that's when the amplitude of the feedback loop stops and it loses control, possibly oscillating by circuit delays alone?
Seems like there's nothing to fix so far, you got yourself a nice phase shift oscillator, neat
You're next Pierre... come to the dark side
Regards,
Chris
Pierre said:
I just ran the amp though RMAA,
these are the figures i have, not the amplifier is nearly clipping during part of the test. So its running a fair amount of power.
Freq Response +2.1,-7.8 <---- that -7.8 is because i am not running the correct output filter. that will be fixed when i build up the pcbs.
Noise Level: - 77dB
Dynamic Range: 77.0
THD %: 0.445
Intermodulation : 0.33%
IMD Swept Freq: 0.48%
and the load impedance for the test was 2.3ohms.....
Other than the first the others aren't too bad, considering the power level, and the fact that there are 4 home made single sided pcb's wired together....
Regards
Peter.
It depends on where this -7.8 dB occurs. If it is at half a Megahertz then you'd have quite a wideband amp!!!!
It is not the fact that your amp is self oscillating that bugs me since all carrier-based class-d amps would do so due to the (theoretically infinite) gain of the comparator as soon as the triangle is removed. What really astonishes me is the low frequency it oscillates on because the delays involved are not that extremely high.
If you like self oscillating anyway and if it is not for sale then why not make it into a UcD ?
Regards
Charles
That results are nice. Peter, have you previously run a loopback test to see how much distortion does your soundcard produce itself?
What I was talking about several modules was not only bridging. You can also have intermodulation products between different channels that tend to produce annoying "whistles". That's a defect of every self-oscillating Class-D amplifier.
Anyway, as you said, you have nothing to loose. When you perform RMAA comparisons between "with" and "without" clock setups, you will have an objective reason to decide.
Best regards,
Pierre
What I was talking about several modules was not only bridging. You can also have intermodulation products between different channels that tend to produce annoying "whistles". That's a defect of every self-oscillating Class-D amplifier.
Anyway, as you said, you have nothing to loose. When you perform RMAA comparisons between "with" and "without" clock setups, you will have an objective reason to decide.
Best regards,
Pierre
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.