D
Deleted member 148505
Oh I see, I haven't changed the value of R9, and I also added some trimpot resistance so that I can set the osc freq to 300kHz, I'll try to put the trimpot to 0R. Does that mean that lower operating frequency will result to lower THD, correct?
Regarding the gain, I just realized that I was tapping the wrong pin on the IC socket (pin 6 instead of 7), so it added 10K to the total input resistance
, maybe I got excited. I will remeasure again. Thank you for the hints and ideas! 
Regarding the gain, I just realized that I was tapping the wrong pin on the IC socket (pin 6 instead of 7), so it added 10K to the total input resistance
, maybe I got excited. I will remeasure again. Thank you for the hints and ideas! ...rubbish...
The simulation which suddenly showed much better results had 80V rails, not 53V.
Furtheron at 100W/8R +/-53V the modulator already has to modulate the duty cycle between 10% and 90%, not 25%-75%.
Now I also checked the older findings of the +/-40V version.
Published in posting #116 you can see distortion vs power.
There the +/-40V amp was powered from 43V rails.
When we translate ratios and rail voltages for your operating point of 100W/8R with 53V rails down to 43V rails then this would correspond to something like 65W into 8R - and there also the 40V version does not show low THD anymore.
When seeing this in the +/-40V version and also failing to finding a better sweet spot for the +/-80V version powered from +/-53V, then I tend to say that this is a general topic.
The LiteAmp has nice speaker control due to its post filter feedback, also nice low power THD, nicely low noise - but high power THD is not its strength.
..just for completion I am adding the simulation results of 100W/8R with +/-53V rails for R9+Pot=200 Ohms and R9+Pot=150 Ohms.
So for optimization it will be worth to spend efforts in the low power range and mid power range, but not the upper power levels.
Sorry that I had been too optimistic in this regard during the last days.
Attachments
Last edited:
D
Deleted member 148505
My LiteAmp setup is still here because I haven't cleaned my room yet, so I got excited again , I changed R6 to 1k but the THD only got worse so I just soldered 1.8K on top of it. R9 is 150R, there is a trimpot threshold where I cannot lower the oscillating frequency to 300kHz, so I just turned it to fully 0R. So default osc freq is around 300kHz.
With parallel 1K + 1.8K I got better mid power measurements, tradeoff is higher noise floor even when preamp is bypassed (maybe because the gain is now corrected), so the preamp is not the bottleneck now I think.
Mid power level THD values are now good, but low power THD is not as good as before but I think the present values are ok and acceptable. When I have extra time I'll gonna setup my +/- 86VDC linear supply and will get the THD values at 1000W. In comparison, with my other prefilter IRAUD implementations, when the power is from 200W to 1000W, THD values are sitting around 0.5%.
Good thing about cheap abletec smps supply is that it is regulated and has a stiff +/-53V DC so I can measure up to unclipped 150W at 8 ohms load.
I think the powerstage has plenty of effective deadtime because the heatsink only got warm when I am measuring THD burst 2 ohm 100W.
, I changed R6 to 1k but the THD only got worse so I just soldered 1.8K on top of it. R9 is 150R, there is a trimpot threshold where I cannot lower the oscillating frequency to 300kHz, so I just turned it to fully 0R. So default osc freq is around 300kHz. With parallel 1K + 1.8K I got better mid power measurements, tradeoff is higher noise floor even when preamp is bypassed (maybe because the gain is now corrected), so the preamp is not the bottleneck now I think.
Mid power level THD values are now good, but low power THD is not as good as before but I think the present values are ok and acceptable. When I have extra time I'll gonna setup my +/- 86VDC linear supply and will get the THD values at 1000W. In comparison, with my other prefilter IRAUD implementations, when the power is from 200W to 1000W, THD values are sitting around 0.5%.
Good thing about cheap abletec smps supply is that it is regulated and has a stiff +/-53V DC so I can measure up to unclipped 150W at 8 ohms load.
I think the powerstage has plenty of effective deadtime because the heatsink only got warm when I am measuring THD burst 2 ohm 100W.
Attachments
D
Deleted member 148505
My congratulation!
This is better than my simulation let's expect.
I agree, the 0.003% for 1W/4R is easily on a level which is fine.
The 0.0007% of my +/-40V version were the result of one month tweaking... and most likely were paid by worse high power performance.
Your result of 0.08% for 200W/4R is by far better than what I had achieved with the +/-40V version at comparable modulation levels.
Regarding R6:
Now you have 1k in parallel with 1k8, means resulting 642 Ohms?
That's perfectly fine, also for stability.
Dead time:
Yes, DT4 is pretty long. I agree, there you still have some options for tweaking - especially in combination with further options for gate drive impedances.
Snubbering:
I also have tried to adjust my simulation according to your layout and snubbering. As you already earlier have stated, your snubbers are pretty light weight, not strong enough to catch all parasitic resonances at high loads.
But this is definitely something which you have to examine and check
on the real proto. And obviously it is already now good enough to have the amp properly running with high performance.
Enjoy
This is better than my simulation let's expect.
I agree, the 0.003% for 1W/4R is easily on a level which is fine.
The 0.0007% of my +/-40V version were the result of one month tweaking... and most likely were paid by worse high power performance.
Your result of 0.08% for 200W/4R is by far better than what I had achieved with the +/-40V version at comparable modulation levels.
Regarding R6:
Now you have 1k in parallel with 1k8, means resulting 642 Ohms?
That's perfectly fine, also for stability.
Dead time:
Yes, DT4 is pretty long. I agree, there you still have some options for tweaking - especially in combination with further options for gate drive impedances.
Snubbering:
I also have tried to adjust my simulation according to your layout and snubbering. As you already earlier have stated, your snubbers are pretty light weight, not strong enough to catch all parasitic resonances at high loads.
But this is definitely something which you have to examine and check
on the real proto. And obviously it is already now good enough to have the amp properly running with high performance.
Enjoy
Last edited:
D
Deleted member 148505
Nice, good thing R6 is good for stability, yes it's around 640 ohms.
I think I'll settle for current values as I'm readying the other channel for stereo listening for sound check. Will use +/-86V supply and will make chassis for them, hopefully THD won't be many times worse on higher supply. Thanks again
I think I'll settle for current values as I'm readying the other channel for stereo listening for sound check. Will use +/-86V supply and will make chassis for them, hopefully THD won't be many times worse on higher supply. Thanks again
D
Deleted member 148505
Hmm I just noticed that the lower THD results do have higher odd harmonics, compared to the results that have high THD on 100W.
I'll try to bridge the amp and remeasure, maybe the odd harmonics will cancel out.
Attached my JLA-250D prefilter IRS2092, even order harmonics are more dominant when bridged.
I'll try to bridge the amp and remeasure, maybe the odd harmonics will cancel out.
Attached my JLA-250D prefilter IRS2092, even order harmonics are more dominant when bridged.
Attachments
I need a help
Hi ChocoHolic,
Your PM space is full so I ask you here.
I have read your thread Lite Amp and would like that you take a look this thread http://www.diyaudio.com/forums/solid-state/304722-200w-class-amp-high-efficency.html you will see that I am trying to design Class A amp in combination with Class D amp as variable power supply. I bought the IRS2092 kit and want to move the nfb point after output inductor but as I am no very familiar with Class D amp I need a help. What I need to change more to make that amp working?
Do you know where to find LTspice IRS2092 model?
Thank you in advance,
Damir
Hi ChocoHolic,
Your PM space is full so I ask you here.
I have read your thread Lite Amp and would like that you take a look this thread http://www.diyaudio.com/forums/solid-state/304722-200w-class-amp-high-efficency.html you will see that I am trying to design Class A amp in combination with Class D amp as variable power supply. I bought the IRS2092 kit and want to move the nfb point after output inductor but as I am no very familiar with Class D amp I need a help. What I need to change more to make that amp working?
Do you know where to find LTspice IRS2092 model?
Thank you in advance,
Damir
...rubbish...
The simulation which suddenly showed much better results had 80V rails, not 53V.
Furtheron at 100W/8R +/-53V the modulator already has to modulate the duty cycle between 10% and 90%, not 25%-75%.
Now I also checked the older findings of the +/-40V version.
Published in posting #116 you can see distortion vs power.
There the +/-40V amp was powered from 43V rails.
When we translate ratios and rail voltages for your operating point of 100W/8R with 53V rails down to 43V rails then this would correspond to something like 65W into 8R - and there also the 40V version does not show low THD anymore.
When seeing this in the +/-40V version and also failing to finding a better sweet spot for the +/-80V version powered from +/-53V, then I tend to say that this is a general topic.
The LiteAmp has nice speaker control due to its post filter feedback, also nice low power THD, nicely low noise - but high power THD is not its strength.
..just for completion I am adding the simulation results of 100W/8R with +/-53V rails for R9+Pot=200 Ohms and R9+Pot=150 Ohms.
So for optimization it will be worth to spend efforts in the low power range and mid power range, but not the upper power levels.
Sorry that I had been too optimistic in this regard during the last days.
Can you hep me understand this post, I keep looking at the frequency domain screen grabs but I don't get the point.
If you want to change standard designs with IRS2092 into a postfilter design then the easiest way is to convert it into a copy of the LiteAmp. For +/-60V...+/-80V, I think the value setting which jlester is using will give good results.
(Alternatively you could invest some years in linear control theory and
some more in non linear control theory.)
Feel free to use my simulations published earlier in this thread.
Thank you for your help, I'll come back to you after some years when I master linear and non linear control theory.
When looking to an output signal of 40Vpk:
With +/-80V rails you need just a moderate modulation of the duty cycle,
but with +/-53V rails you need a massive modulation of the duty cycle in order to achieve 40Vpk.
As a consequence with +/-80V rails the modulator distortions at output signals of 40Vp are much less.
To a certain extend this is also corresponding to the frequency drop at high modulation levels.
IMHO it is hard to read such things from frequency domain.
Not sure if Bruno would agree with me - but to me frequency domain is only helpful for the part of linear control theory. Everything else, including the topic of non linear control theory, I prefer to examine in time domain.
...after you manage this, you won't need a dinosaur like me.
Usually I have the impression that control theory is mastering me, not the other way round.
...after you manage this, you won't need a dinosaur like me.
Usually I have the impression that control theory is mastering me, not the other way round.
That was a joke, I am 72 and I don't have to much time to spent on Class D, but what a hack, it looks I have to put some effort in that.
I haven't seen that you simulated Loop Gain to see what phase and gain margin are there?
..looks like you already have spend some years on control theory.
It seems like you changed a little bit the load and added an AC current source, but the resolution of the picture does not allow to read your values.
Not promising? You wish to have more phase margin?
The massive lag phase shift of the output filter is generally no fun inside the feedback loop.
And the restricted options of the IRS2092 where you cannot adjust any parameter independently, is not really comfortable.
I think that's why Eva simply categorizes the IRS2092 as not suited for postfilter designs.
Nevertheless the solution turned out to be surprisingly forgiving versus
component tolerances, even without load.
Don't let you confuse that without load the phase will cross 180 deg before the gain reaches 0dB. Phase comes back again and at the point were the gain reaches 0dB the phase is OK again. Stable.
Bode diagram is often confusing people here.
Niquist diagram (sometimes called Smith chart) with its funny loopings definitely is helpful in case of doubts.
Somewhere earlier in the thread I posted an evolution of post filter structures which are possible with the IRS2092.
I guess the version with just 1st order integrating gain will allow adjustments with more margin. But due to much less loop gain you will get higher THD figures.
Nice thing: Modding a standard design based on IRS2092 with prefilter feedback into a postfilter design with first order integrating gain should be possible with just some few flying components, while most should fit in the standard component provisions.
It seems like you changed a little bit the load and added an AC current source, but the resolution of the picture does not allow to read your values.
Not promising? You wish to have more phase margin?
The massive lag phase shift of the output filter is generally no fun inside the feedback loop.
And the restricted options of the IRS2092 where you cannot adjust any parameter independently, is not really comfortable.
I think that's why Eva simply categorizes the IRS2092 as not suited for postfilter designs.
Nevertheless the solution turned out to be surprisingly forgiving versus
component tolerances, even without load.
Don't let you confuse that without load the phase will cross 180 deg before the gain reaches 0dB. Phase comes back again and at the point were the gain reaches 0dB the phase is OK again. Stable.
Bode diagram is often confusing people here.
Niquist diagram (sometimes called Smith chart) with its funny loopings definitely is helpful in case of doubts.
Somewhere earlier in the thread I posted an evolution of post filter structures which are possible with the IRS2092.
I guess the version with just 1st order integrating gain will allow adjustments with more margin. But due to much less loop gain you will get higher THD figures.
Nice thing: Modding a standard design based on IRS2092 with prefilter feedback into a postfilter design with first order integrating gain should be possible with just some few flying components, while most should fit in the standard component provisions.
One hint:
I am always using the combination of transient analysis and AC analysis.
Transient analysis will show the size of the carrier signal at the comparator input, which in conjunction with the supply rails tells the gain from comparator input to halfbridge output.
Transient analysis furtheron allows to analyze step response, non linear effects like clipping - and of course a rough THD prediction.
AC analysis allows to double check nyquist criteria and get hints in which
direction parameters should be shifted in for optimisation of phase margin and gain margin.
...it's an iterative design flow...
I am always using the combination of transient analysis and AC analysis.
Transient analysis will show the size of the carrier signal at the comparator input, which in conjunction with the supply rails tells the gain from comparator input to halfbridge output.
Transient analysis furtheron allows to analyze step response, non linear effects like clipping - and of course a rough THD prediction.
AC analysis allows to double check nyquist criteria and get hints in which
direction parameters should be shifted in for optimisation of phase margin and gain margin.
...it's an iterative design flow...
...oohps, my evolution of post filter structures was in another thread.
http://www.diyaudio.com/forums/class-d/284973-tips-suggestions-irs2092-implementation-8.html
Posting #79.
http://www.diyaudio.com/forums/class-d/284973-tips-suggestions-irs2092-implementation-8.html
Posting #79.
This is LG simulation with 22uH inductor (I have this one in the kit I brought). I used the compensation values you suggested for 22uH inductor. The Closed Loop Gain does not look so good.
THD is not mine main concern, as I am going to use Class D amp as OPS variable power supply for my Class A linear amplifier, but gain and phase variation.
I am going to proceed with this discussion in my thread http://www.diyaudio.com/forums/solid-state/304722-200w-class-amp-high-efficency.html and will be helpful if you visit it from time to time.
THD is not mine main concern, as I am going to use Class D amp as OPS variable power supply for my Class A linear amplifier, but gain and phase variation.
I am going to proceed with this discussion in my thread http://www.diyaudio.com/forums/solid-state/304722-200w-class-amp-high-efficency.html and will be helpful if you visit it from time to time.
Attachments
To me a phase margin of 56deg and gain margin of 14db looks more than comfortable. Which values do you target?
Two further questions:
1.
Your combination of 100k||100pF in the global feedback will lead to an unpleasant slow amp. Is this on purpose a low bandwidth version?
2.
Did you readjust the gain of E3 in your simulation?
Your margins are looking unrealistic large.
When I am doing the simulation with your low bandwidth version
I am getting values as attached.
Two further questions:
1.
Your combination of 100k||100pF in the global feedback will lead to an unpleasant slow amp. Is this on purpose a low bandwidth version?
2.
Did you readjust the gain of E3 in your simulation?
Your margins are looking unrealistic large.
When I am doing the simulation with your low bandwidth version
I am getting values as attached.
