Notch filter needs just a bit of fine tuning to strike the 6.5K resonance. At the moment it’s closer to 7KHz.
These things- wandering breakup modes, or passive components needing slight adjustment in values to precisely hit the resonance is somewhat of a pain- and easy to miss in QC / EOL testing… if not measuring each driver’s response individually.
These things- wandering breakup modes, or passive components needing slight adjustment in values to precisely hit the resonance is somewhat of a pain- and easy to miss in QC / EOL testing… if not measuring each driver’s response individually.
Last edited:
Can you be more specific about the problems with IcePower amps? The biggest issue I see is the higher distortion at 6.67kHZ. I figure if HF the output is limited to 70W THD stays below .08% Maybe not perfect but inaudible.
Amir rated fairly well, but some disagree.
https://www.audiosciencereview.com/...power-125asx2-class-d-amplifier-review.10442/
Just in general Class-D amplifier can create issues.
I don't care that much about distortion, but some audio interfaces don't play nice with the switching noise.
As seen here, a lot of them are not load independent.
Even IcePower isn't (not only talking around 80-90kHz, but also the difference in lines)
The only ones that do this well, are Hypex and a couple of diy projects.
IcePower is also not full PFFB (I think it's some kind of hybrid approach? I can't remember)
The long term continues power of the IcePower amps is also not great.
Some Class-D amps are inverting as well (can't remember with IcePower).
But I think the better question to ask and answer to give, is why would you?
With something like a LM3886 etc, or any other non-inverted Class-AB amplifier, you know for sure that the freq resp is load independent.
Something like this for example;
https://www.thomann.de/intl/the_tamp_pm40c_endstufenmodul.htm
But even a good old 2nd hand stereo amp will do just fine.
There also some composite kit's available for around 70 bucks I believe.
Still planning to make a kit myself that is slightly more focused on measuring (and much more affordable)
Class-D amplifiers are nice for being very efficient, but because of their complexity I find them often problematic in test and measurement setup.
Based on 15 years of personal and professional experience with literally all kinds of test solutions.
They often just bring unknowns to the table.
The last thing you want during a measuring session.
Some work great, the majority won't mostly because they don't have (enough) PFFB.
Seriously complex loads are often even more problematic.
So why going that route if we don't need the efficiency for a test setup?
Thanks for the update. My question was more general, not specifically for measurements. I agree Hypex nCore (not UCD) or Purifi are superior class D amplifiers, but they cost substantially more. I am currently contemplating building a 6 channel power amp to replace an 8 channel SpeakerCraft class A/B amp. My goal is lower weight and higher efficiency and the IceMatch systems is my most likely candidate.
A good reason for using class D for measurements is weight. I sure don't want to haul a 25kG amp out into my backyard to make measurements.
Jim, sorry for the thread hijack.
A good reason for using class D for measurements is weight. I sure don't want to haul a 25kG amp out into my backyard to make measurements.
Jim, sorry for the thread hijack.
In that case; whatever makes you happy, suits the project and floats your boat.
Because some Class-D amps don't have proper PFFB (in fact most don't), you might need to tweak the top-end frequencies a bit.
Like shown here in this thread.
Why do you think you need a 500-1000W Class-AB amplifier to do measurements?
I used to work with professional active PA-speakers and even those amps weren't that heavy.
Although close..... and no I don't miss these days....
Well I miss the headroom as wel as long term power those amps had lol
Even Hypex isn't great with this (that is quite an understatement)
Pascal is a little better.
For just a "spinorama" 50W is absolutely plenty!
In fact 20-30W will probably do just fine.
@tktran303 and @wolf_teeth - - sharp eyes, nice catch. I posted the wrong filter ... I was playing with "what if's" and experimenting, and I saved an experimental sim over the baseline... Here is the one I intended to post.
... I was playing with "what if's" and experimenting, and I saved an experimental sim over the baseline... Here is the one I intended to post.
The above filter "alternate notch", is interesting because the midrange circuit has 5 reactive elements in the high frequency section (low pass and notch). This provides a lot of degrees of freedom to make adjustments.
However, this is not a filter topology I will be using. Instead I will be using this topology which uses a parallel notch in series with the midrange driver:
The response of the two filters is very similar, close enough that it is within my margin of error. But this Rev 20.8 filter has the advantage in being more easily adjustable for voicing. I can adjust the BSC and the overall slope from 100 - 10k with just 3 resistors.
The Rev 20.8 filter also has an advantage in the impedance seen by the midrange driver.
Here is the midrange circuit, viewed in isolation, with the driver replaced with a voltage source, and the amp replaced with a path to ground. First the "alternate notch" circuit
And now the Rev 20.8 filter
In the important region from 5k to 10k, the Rev 20.8 filter presents the driver with a much higher impedance. The back EMF from the driver will generate much less current when looking into the Rev 20.8 filter than would be the case with the "alternate notch" filter. Apologies if I am not using the correct electrical engineering terminology... The Purifi white paper talks about the importance of this.
So at this point, I think I am finalized on a filter design. Next step is to order the parts.
However, this is not a filter topology I will be using. Instead I will be using this topology which uses a parallel notch in series with the midrange driver:
The response of the two filters is very similar, close enough that it is within my margin of error. But this Rev 20.8 filter has the advantage in being more easily adjustable for voicing. I can adjust the BSC and the overall slope from 100 - 10k with just 3 resistors.
The Rev 20.8 filter also has an advantage in the impedance seen by the midrange driver.
Here is the midrange circuit, viewed in isolation, with the driver replaced with a voltage source, and the amp replaced with a path to ground. First the "alternate notch" circuit
And now the Rev 20.8 filter
In the important region from 5k to 10k, the Rev 20.8 filter presents the driver with a much higher impedance. The back EMF from the driver will generate much less current when looking into the Rev 20.8 filter than would be the case with the "alternate notch" filter. Apologies if I am not using the correct electrical engineering terminology... The Purifi white paper talks about the importance of this.
So at this point, I think I am finalized on a filter design. Next step is to order the parts.
we were typing at the same time...
Yes I am aware, but only because smart people on this forum pointed the way... I am a structures engineer, so when it comes to EE stuff, I play "follow the leader"...
You can always unwind coils a couple turns to achieve curves.
Multiple resistor values to either side of the modeled values can help dial it in.
I recommend a set of 20, 25, 30, 33, and 40 ohm resistors as well in case you need to reduce the output a smidge and suppress Fs buzzing in the mids and tweeters. Placing these across the drivers can reduce and suppress. Play James Taylor, and you'll here it if there. It adds a gravel/nasality to his smoothness that shouldn't be there.
I usually try a next smaller value cap on the tweeter highpass to see if it audibly reduces sibilance. Sometimes this small xover Q adjustment helps the spectral balance.
Have fun!
Multiple resistor values to either side of the modeled values can help dial it in.
I recommend a set of 20, 25, 30, 33, and 40 ohm resistors as well in case you need to reduce the output a smidge and suppress Fs buzzing in the mids and tweeters. Placing these across the drivers can reduce and suppress. Play James Taylor, and you'll here it if there. It adds a gravel/nasality to his smoothness that shouldn't be there.
I usually try a next smaller value cap on the tweeter highpass to see if it audibly reduces sibilance. Sometimes this small xover Q adjustment helps the spectral balance.
Have fun!
I know the paper by Purifi states that reducing HD comes from adding damping in the range the parallel notch operated, lowering current flow in the range required.
What is not easily seen is that the tank portion also does this. The 1.2mH and the 0.47uF also add resistance to impede flow. The filter I suggested also takes it a step further by sending the offending ranges down the LCR across the driver. This is a trifecta, adding resistance, also shorting it out, and further reducing the offending range to below audibility. Breakup and energy storage are to be eliminated.
You can't just show the entire circuit like you did here and say one is better than the other when both are doing the job described. In fact, the alternate adds MORE resistance through the tank than the 56 ohms in the parallel notch. The difference here is that the load seen by the amplifier is generally less variant in magnitude and the impedance phase is more stable.
I look at it like this; the parallel notch is like a bouncer at the door of a club. Just says no.
The alternate also says no, but locks the secondary entrances, and places trap doors just inside the perimeter of the building specific to those that were told no.
What is not easily seen is that the tank portion also does this. The 1.2mH and the 0.47uF also add resistance to impede flow. The filter I suggested also takes it a step further by sending the offending ranges down the LCR across the driver. This is a trifecta, adding resistance, also shorting it out, and further reducing the offending range to below audibility. Breakup and energy storage are to be eliminated.
You can't just show the entire circuit like you did here and say one is better than the other when both are doing the job described. In fact, the alternate adds MORE resistance through the tank than the 56 ohms in the parallel notch. The difference here is that the load seen by the amplifier is generally less variant in magnitude and the impedance phase is more stable.
I look at it like this; the parallel notch is like a bouncer at the door of a club. Just says no.
The alternate also says no, but locks the secondary entrances, and places trap doors just inside the perimeter of the building specific to those that were told no.
This has been talked about and showed already in very great lengths.
I even showed so direct results myself as well.
Any impedance in series will lower this distortion.
Parallel circuits do it as well, just not as much.
So I guess there is a fair point to be made here, that probably the series inductor already takes out a big chunk.
Anyway, it takes a couple of minutes just to measure this
You may be right (probably are right)... but I am not knowledgeable enough in analog circuit analysis to grasp all the intricacies and make an informed decision...
I would not presume to declare one better than the other... I simply like this one better.
By the way, thank for all the comments. The back and forth discussion helps stimulate my mind and keeps me from falling into blind alleys or getting stuck in a rut.
both circuits acts as notches, reduce the voltage and current for the driver. However, the circuits have different impedances seen from the driver and this affects how much current the back EMF converts into. This back EMF is nonlinear due to the magnetic hysteresis in the motor.
cheers
Lars
