Same kind of thing you noted recently - DC offset when you have 20KHz at the output close to clipping.
Actually, here is what I think.
We're using relatively high time constant in the servo integrator (4.7 seconds).
Because of this, integrator does not follow the offset changes fast enough.
Let's try 1 second. You need to change 4.7uF capacitor to 1uF one for that.
Let's test the IPS standalone of the whole amp with dummy load until we see the offsets are not too high.
Is it useful to play loud music on a dummy load, monitor the d.c out and save this in a file for further analyze?
Also may be useful. DC can fluctuate within some limits, but should not reach the limit where protection trips, for sure.
I did some analysis in simulations - in fact, DC offsets are higher in the cases where the servo time constant is higher.
So, let's switch to 1uF integrator cap everywhere (giving us 1 second time constant).
So, let's switch to 1uF integrator cap everywhere (giving us 1 second time constant).
I managed to permanently damage the bass mid of an AE1 playing it fairly loud in a 3way active system and it never got as loud as audible distortion or audible clipping.
I think my problem was that the lower bass was removed by the active filtering and that reduced the cooling of the voice coil. Next morning one AE1 was locked up. The VC appeared to have distorted due to excessive heating during "normal music" reproduction.
I can see this being possible, but I've never actually seen it happen. I used to run ridiculous car audio systems and have had many friends running competition systems who never managed to damage a voice coil without an amplifier problem either. Maybe we were just lucky or listen to music more appealing to our speakers, but I can't imaging the noise they play during SPL competitions being speaker friendly!
I was looking at that cap wondering if it being that large might be slowing things down for the servo. Will lowering it's value cause ill effects to sound?
Last edited:
The servo is usually just a bypass around the normal Feedback. The normal feedback controls the output.
Only when the servo detects an "error" within it's pass band does the servo add in an extra "bypass" feedback.
With a non inverting servo using a filter on it's input and a capacitor feedback in it's own loop (integrator) there is no correction to bypass the normal feedback, unless there is an "error" detected inside it's passband.
If there is an error due to near clipping HF signal, then the filter in the servo input should attenuate it to have no correction and thus leaves the normal feedback as the sole control of that near clipping HF signal.
Only when the servo detects an "error" within it's pass band does the servo add in an extra "bypass" feedback.
With a non inverting servo using a filter on it's input and a capacitor feedback in it's own loop (integrator) there is no correction to bypass the normal feedback, unless there is an "error" detected inside it's passband.
If there is an error due to near clipping HF signal, then the filter in the servo input should attenuate it to have no correction and thus leaves the normal feedback as the sole control of that near clipping HF signal.
Last edited:
Well, I was using the higher value just to have it slower than required to be sure we don't influence distortion at the low end of the audio range (some 10...20Hz).
However, 1 second time constant is enough - simulation is showing a good combination of 1uF for the integrator and 1uF DC blocking cap at the input (assuming 27K input impedance, that we have in the most of our front-ends.).
Right, it just looks like we've got the servo reacting too slowly, so the error is already there and continues growing, but servo output is moving too slowly with 4.7 s time constant.
I never noticed that offset coming at higher frequencies close to clipping before, so my approach with regards to time constant was sort of "the more is the better".
Now it looks like 1...2 seconds is the best compromise.
Now it looks like 1...2 seconds is the best compromise.
I thought having Thimios do some test builds would bring out some valuable information. It looks like I was correct, but it's too bad he lost a speaker in the process.
I suspect Terry's Mini Modular builds will be equally valuable.
I suspect Terry's Mini Modular builds will be equally valuable.
Absolutely.
Although I'm still pretty much surprised the DC could be that high to damage the speaker.
I was abusing my Vertical + NS-OPS prototype like hell at different levels with different speakers and never experienced anything like this.
Let's see what the faster servo tests will show. We have to make it bullet proof.
We can also test VFA front-ends with no servo, adding 220uF cap to the feedback network (from 1K resistor to ground).
Zero it out manually and then run the sweep and all the other type of provocations.
The aim would be to see what setup will keep less offset at all times - servo or no servo.
Although I'm still pretty much surprised the DC could be that high to damage the speaker.
I was abusing my Vertical + NS-OPS prototype like hell at different levels with different speakers and never experienced anything like this.
Let's see what the faster servo tests will show. We have to make it bullet proof.
We can also test VFA front-ends with no servo, adding 220uF cap to the feedback network (from 1K resistor to ground).
Zero it out manually and then run the sweep and all the other type of provocations.
The aim would be to see what setup will keep less offset at all times - servo or no servo.
I'm still in the dark,
Pushing the amplifier hard i can't measure any d.c output higher than +/-500?mV using many voltmeters digital or electronic analog in 1v range 0V center...
When the amplifier was pushed and V range of voltmeter is used i read 0v ,when d.c mV range is used i read -500mv ,How this explained?
I wonder if this woofer was at the end his life....
Pushing the amplifier hard i can't measure any d.c output higher than +/-500?mV using many voltmeters digital or electronic analog in 1v range 0V center...
When the amplifier was pushed and V range of voltmeter is used i read 0v ,when d.c mV range is used i read -500mv ,How this explained?
I wonder if this woofer was at the end his life....
Last edited:
I wonder if this has to do with the sampling rate of your voltmeters. If you are using a higher end Fluke you can select faster sampling rates on boot up that will show different voltage readings than other meters on odd applications.
He he Fluke 289 show 8v d.c when peak mode was selected... 20kHz at clipping levels.