Stable clipping is difficult, so much more so if you have a complex compensation scheme. Somehow you must clip without disturbing the phase margin, and this requires strategic design so it clips in the right place. And you must figure out how to do this without leaking distortion into the signal path.
You're using a 5V output square wave. You may be stressing the amp a lot, producing more errata. I would see how much you can reduce the square voltage before the square edges stop looking nicer and become symmetrical. This will tell you what a small-signal square wave is for your amp. It is possible to ruin small-signal behavior by fixating on large-signal behavior and compensating only for that. Large-signal step response needs robust topology, small-signal needs phase linearity (tuned compensation).
You're using a 5V output square wave. You may be stressing the amp a lot, producing more errata. I would see how much you can reduce the square voltage before the square edges stop looking nicer and become symmetrical. This will tell you what a small-signal square wave is for your amp. It is possible to ruin small-signal behavior by fixating on large-signal behavior and compensating only for that. Large-signal step response needs robust topology, small-signal needs phase linearity (tuned compensation).
Almost by definition, a small-signal square wave does not cause smoke. 200KHz is totally appropriate if your levels are not too high. I've explained why I used 40KHz, but adding to that, I'd say that if an amp smokes at 40KHz then I'm apprehensive about its robustness at 20KHz.
What might be a problem is if the square wave is triangulated - this means the circuit never recovers from the transients. If the step response is asymmetric, the uneven recovery adds up until the amp's operating points are out of whack and then anything can happen. There's not much more to learn about an amp when it begins to triangulate except its RFI failure modes. Unless you're using too large a square wave though, the disturbance is usually not enough to cause worry.
Then again, the amp could be triangulating because of lead compensation and be totally fine.
This is something I have yet to get LTSpice to do for me. Simulate phase margins when the output voltage is near the rails.
Going to take this and run with it. Can see a possible way forward to solve this issue but need to spend some time simulating to make sure the solution is not going to cause an issue elsewhere which is usually the case. Then have a think about what all this really means.
Thank you,
Paul
P.S. Been optimising the compensation scheme for a smooth group delay plot. Never even considered this kind of test or amplifier behaviour. Especially like the use of the d(Vout) plot.
If you're paying attention to group delay you should already be doing well for small-signal behavior.
To simulate near the rails I add a voltage source between the output and the feedback divider, which tricks the amp into putting the output where you want it. This has worked best for me. If you have bootstraps you'll have to support those with voltage sources as well.
To simulate near the rails I add a voltage source between the output and the feedback divider, which tricks the amp into putting the output where you want it. This has worked best for me. If you have bootstraps you'll have to support those with voltage sources as well.
A stable amp while clipping is still a clipping amp and spewing out harmonics in abundance while clipping.
THx-RNMarsh
THx-RNMarsh
Yes, lessons learnt from my previous attempt. 😉
Thank you for this. Will give it a go.
Avoiding bootstraps for now. Cascodes have more appeal.
Quick messing about in sims. If I sacrifice some THD and slew rate, the behaviour in said test becomes much cleaner. Another lesson learnt. There's other stuff to consider beyond THD and slew rate.
Can see a good many hours of simulating and real world experimenting ahead.
The beauty of all this is that I have a stripboard based working prototype to try all these changes on.
Another question about the sine + square wave test. What is a good value to set the square wave T Rise / T Fall in LTSpice?
Trise and Tfall won't do much good if your timestep is larger. In my SPICE rig for amplifiers I have Trise and Tfall equal to the simulation timestep, and I have some B-sources indicating the timestep and Nyquist frequency for the simulation.
Good function generators tend to get around 10nS, which is usually more than enough for a power amp. You should check to see that your square wave risetime is much shorter than your amp risetime. I found that I preferred 200KHz square waves in simulation (as well as life) because that resulted in the highest simulation BW while still being usable and timely.
EDIT: Rig for reference.
Good function generators tend to get around 10nS, which is usually more than enough for a power amp. You should check to see that your square wave risetime is much shorter than your amp risetime. I found that I preferred 200KHz square waves in simulation (as well as life) because that resulted in the highest simulation BW while still being usable and timely.
EDIT: Rig for reference.
True! If you don't have the steadfast (and possibly unrewarding) dedication it takes to get pristine clipping behavior though, I would still run the dynamic step response test to make sure everything's right within normal operating ranges.
The last time I tried to hear 'clipping' behavior with a JC-1, the police were called by the neighbors. '-)
Mmmm, I thought it was another name but I may be incorrect on this. It is a great test!
Your neighbors are only a 2" x 4" and some sheet rock away from you and them. I had same problem when i lived in an apartment complex at only 85-90dB spl in my room. I had to turn it down to where I couldnt hear it over their TV. I moved and rented a house. No more problems.
-RM
Last edited:
When people complain or even have to seriously consider clipping, they should either design or purchase a bigger amp, OR get more efficient speakers.
I think clipping needs to be considered. How many would have a completely unprotected beta enhanced VAS? In my very limited experience it clips in a very ugly way.
Suppose if you design an amp just for DIY hobby purposes you could live with it but for a commercial product it may not be a good idea.
Suppose if you design an amp just for DIY hobby purposes you could live with it but for a commercial product it may not be a good idea.
CUOA -
😎🙂
Note - Factoid:
1. A 512 Watt amplifier coupled with an 86dB effecient speaker will be just as loud as an 8 watt amplifier with a 104dB speaker. Source - AudioMasterClass.
2. Nearfield Monitoring: 25W for 85dB SPL average (with 15dB peaks); 250W for 95dB average (with 15 dB peaks)
3. Home Stereo: 150W for 85dB average (with 15dB peaks); 1500W for 95dB spl average (with 15dB peaks). Source- Harmon/JBL/Crown
You can play with the numbers to your hearts content but most of you are listening to amps which are clipping and producing a lot of extranious harmonics. To Clean Up Our hi-end Act, lets design for the proper power levels (higher); Like, 150W-250W/8 for typical size home listening room.
THx-RNMarsh
😎🙂
Note - Factoid:
1. A 512 Watt amplifier coupled with an 86dB effecient speaker will be just as loud as an 8 watt amplifier with a 104dB speaker. Source - AudioMasterClass.
2. Nearfield Monitoring: 25W for 85dB SPL average (with 15dB peaks); 250W for 95dB average (with 15 dB peaks)
3. Home Stereo: 150W for 85dB average (with 15dB peaks); 1500W for 95dB spl average (with 15dB peaks). Source- Harmon/JBL/Crown
You can play with the numbers to your hearts content but most of you are listening to amps which are clipping and producing a lot of extranious harmonics. To Clean Up Our hi-end Act, lets design for the proper power levels (higher); Like, 150W-250W/8 for typical size home listening room.
THx-RNMarsh
Last edited:
I think those figures are highly suspect tbh. 1.5kW for 95db average. That's around 110vrms output if your talking resistive loads. I'm with John on that one... if you really really have a clipping issue then you need to look at more efficient speakers.
We had a whole thread devoted to how much voltage swing was needed for a users chosen maximum listening level.
http://www.diyaudio.com/forums/mult...much-voltage-power-do-your-speakers-need.html
And using pink noise to get 85db
http://www.diyaudio.com/forums/mult...wer-do-your-speakers-need-19.html#post2893167