PI Filter
Maybe AndrewT will appear and shed some light on the topic? I'm not good with explanations.
But here's a try. . .
The Pi filter removes that screeching diode noise.
Here's what a v+ rail looks like with pi filter:
rectifier > 4700uF > 2R > 10,000uf > interconnect cable >330uF > 100nF > LM3875
See the 2R?
Well, that might not be the right value for your application; however, I'd just like you to note that its in series with the power (aka, the power runs through it).
Point: RC Filter
Resistance separates the 4700uF cap from the 10,000uF, and a much smaller amount of resistance (the interconnect cable) separates the 10,000uF at the power supply from the 330uF at the amplifier board (in this example).
In this example, although the 2R figure is more sensational in creating an RC filter, please note that the resistance of the interconnect cable does exist, and it does create an additional RC filter, tiny but working.
Point: If the interconnect cable is missing, you might like to substitute a tiny value resistor (simulate the cable) rather than a solid trace (no filter).
Instead of the 2R in the example above, you can use anything in the range of 0.22R to 4.7R. The smaller figure isn't effective at much of anything. The largest figure is going to reduce power during clipping so that it isn't either as noticeable or as hazardous to speakers. This helps during big loud dynamics so that they won't also suddenly pipe up with loud clipping (you'd exchange that for quiet clipping).
So, both current limiting and RC filter. Yeah, its just a resistor, but man, isn't it nifty?
P.S.
Previous examples added 10,000uF caps that didn't make a difference because there was no RC filter. To make those caps work as intended, use PI filter. That will make those diodes shut up, decrease apparent midrange, decrease volume of clipping, and then your 10,000uF caps will do what you expect.
Maybe AndrewT will appear and shed some light on the topic? I'm not good with explanations.
But here's a try. . .
The Pi filter removes that screeching diode noise.
Here's what a v+ rail looks like with pi filter:
rectifier > 4700uF > 2R > 10,000uf > interconnect cable >330uF > 100nF > LM3875
See the 2R?
Well, that might not be the right value for your application; however, I'd just like you to note that its in series with the power (aka, the power runs through it).
Point: RC Filter
Resistance separates the 4700uF cap from the 10,000uF, and a much smaller amount of resistance (the interconnect cable) separates the 10,000uF at the power supply from the 330uF at the amplifier board (in this example).
In this example, although the 2R figure is more sensational in creating an RC filter, please note that the resistance of the interconnect cable does exist, and it does create an additional RC filter, tiny but working.
Point: If the interconnect cable is missing, you might like to substitute a tiny value resistor (simulate the cable) rather than a solid trace (no filter).
Instead of the 2R in the example above, you can use anything in the range of 0.22R to 4.7R. The smaller figure isn't effective at much of anything. The largest figure is going to reduce power during clipping so that it isn't either as noticeable or as hazardous to speakers. This helps during big loud dynamics so that they won't also suddenly pipe up with loud clipping (you'd exchange that for quiet clipping).
So, both current limiting and RC filter. Yeah, its just a resistor, but man, isn't it nifty?
P.S.
Previous examples added 10,000uF caps that didn't make a difference because there was no RC filter. To make those caps work as intended, use PI filter. That will make those diodes shut up, decrease apparent midrange, decrease volume of clipping, and then your 10,000uF caps will do what you expect.
AndrewT
Why not just parallel LM3886s to get the o/p current required for the load if a single one is struggling into less efficient or lower impedance speakers?
I have heard stories that this affects sound quality adversely but I can't for the life of me think why it should if the sharing resistors are carefully chosen. In fact it should help since there will be less thermal modulation of the input stage transistors by the output stage.
Any ideas?
Why not just parallel LM3886s to get the o/p current required for the load if a single one is struggling into less efficient or lower impedance speakers?
I have heard stories that this affects sound quality adversely but I can't for the life of me think why it should if the sharing resistors are carefully chosen. In fact it should help since there will be less thermal modulation of the input stage transistors by the output stage.
Any ideas?
Last edited:
Brilliant! Combine a bit extra pi filter and a bit extra emitter resistor and you have a soft clip amplifier.
Does that help reduce the spike circuit noises?
Do the emitter resistor's loss reduce compatibility with 2-way ported speakers (whereby the decreased dampening allows the woofer to flap more, shaking up the singer's voice even worse)?
Does that help reduce the spike circuit noises?
Do the emitter resistor's loss reduce compatibility with 2-way ported speakers (whereby the decreased dampening allows the woofer to flap more, shaking up the singer's voice even worse)?
Daniel you're doing it again. What does that mean?
The power supply PI filter has absolutely nothing to do with the clipping behaviour of an amp and what the emitter resistor's loss has to do with "reduced compatibility with 2-way ported speakers (whereby the decreased dampening allows the woofer to flap more, shaking up the singer's voice even worse)" God only knows.
What drugs are you on man????
The power supply PI filter has absolutely nothing to do with the clipping behaviour of an amp and what the emitter resistor's loss has to do with "reduced compatibility with 2-way ported speakers (whereby the decreased dampening allows the woofer to flap more, shaking up the singer's voice even worse)" God only knows.
What drugs are you on man????
Last edited:
How did you measure the noise and where (on the rectifier, caps, amp board etc)? Even at 50dB PSRR that 2R resistor won't matter. It only produces heat. What's the slope of the PI filter you described? Regarding current limitation: it limits the current to what value?
The vast majority around here knows the answers, what about you?
Okay. Not sure if I did this to your preferences or not, but the indicators were voltage of the rails, heat at the capacitors, and heat at the amplifier's heatsink (amplifier on and with a load). Fixing the noise decreases the voltage slightly and cuts the heat in half.
A side effect was a more level frequency response.
As for PI filter versus current limiting, I'm wasn't making a guitar amplifier but merely noted the possibility of partial tube simulation.
As for emitter resistor versus compatibility with 2-way ported speakers, see also: Electronic Dampening
Daniel, I asked for numbers. That's what measuring does, it tells a value of something. I was talking about the 2R you recommended in the Pi filter, not about the one placed at the output of the IC.
Anyway, how did you fix the noise without knowing the cause of it? I'm sure that people around are eager to find out. I can put my hand in fire that if I added another 200uF in your system without you knowing about it, you didn't notice. So please, stop misleading people who are trying to actually learn something.
Later edit: If your smoothing caps get hot, that's really not OK. But I'm sure you'll find out why when they'll blow.
Anyway, how did you fix the noise without knowing the cause of it? I'm sure that people around are eager to find out. I can put my hand in fire that if I added another 200uF in your system without you knowing about it, you didn't notice. So please, stop misleading people who are trying to actually learn something.
Later edit: If your smoothing caps get hot, that's really not OK. But I'm sure you'll find out why when they'll blow.
Last edited:
I never recommended the specific value of 2R, because you'd certainly want to know what noise you're filtering before choosing a filter or its specific values. The 2R was in an example that's simply not as complex as you're making it out to be.
2R, 10R, 5k, it was just your example, it doesn't matter. The point was if you can show how much "noise" will be gone after that filter. I really don't think you have any idea how hard is to design a good power supply.
You say you need to know what noise needs to be filtered. Well, now the questions which I hope to blast off your credibility:
1. How much noise there is? (measured in something)
2. How do you measure? (when, where)
3. How does the noise behave? (white noise, spikes, HF interference etc)
4. How do you eliminate something which you're not sure what is exactly?
The truth is that the pi filter is useless, in only produces heat. You and other people need to realize that electronic design involves a hell lot of math and physics. I'm not making it complex at all, this is the situation in reality. Actually it's more complex than you can think of. Only after optimizing the general design somebody can do fine tuning on his system, and this if he/she knows what to do. Saying that X cap sounds better than Y cap even though they're placed on the rails is called placebo.
It doesn't mean that something cannot be measured because somebody has no idea about the measuring procedure and tools. Luckily, the engineers who design nowadays chips do not use your approach. An engineer is somebody who can make something for $1 what any moron can do the same thing for $100.
You say you need to know what noise needs to be filtered. Well, now the questions which I hope to blast off your credibility:
1. How much noise there is? (measured in something)
2. How do you measure? (when, where)
3. How does the noise behave? (white noise, spikes, HF interference etc)
4. How do you eliminate something which you're not sure what is exactly?
The truth is that the pi filter is useless, in only produces heat. You and other people need to realize that electronic design involves a hell lot of math and physics. I'm not making it complex at all, this is the situation in reality. Actually it's more complex than you can think of. Only after optimizing the general design somebody can do fine tuning on his system, and this if he/she knows what to do. Saying that X cap sounds better than Y cap even though they're placed on the rails is called placebo.
It doesn't mean that something cannot be measured because somebody has no idea about the measuring procedure and tools. Luckily, the engineers who design nowadays chips do not use your approach. An engineer is somebody who can make something for $1 what any moron can do the same thing for $100.
Well, if you're "gunning for" credibility in audio, I think you're going to run out of personnel quickly.
I'd like to try #3 in the context of this thread.
#3. The MUR860's emit RF interference. I say this because of attempting to use these in an integrated receiver amplifier, and then its radio reception range was cut in half (barely Bartlesville from the KS state line).
Installing Mark Houston's snubbed 1-piece rectifier example brought back the original performance (all the way to Tulsa from the KS state line). Its no surprise because the Mark's working example is exactly the same circuit as is found in a production receiver. Repairing the problem by replacing the noisy rectifier also made 4% less voltage, likewise less noise as well. Not surprisingly, the amplifier's heat sink ran consistently much cooler after the repair.
It seems as if you'd prefer that everyone in diy audio made the first step with an engineering degree. Well, as delightful as that might be, its not related to reality. I'd just be happier if we didn't put extra noisy diodes into audio amplifiers for marketing purposes. That wish is also not related to reality. Marketing rules both engineer and audio. Best of luck with that.
Now if you'd like to do something constructive, how would you go about getting the gainclone people better bass (via a flatter response) and less noise? Care to be constructive?
Daniel,
post31,
Did you find out what the reciever was doing before and after you modified it.
I suspect the cause and solution had nothing to do with what you are describing.
I suspect there was science in the cause and science in the solution.
Your version of the (non)science is too far into infinity for me to understand and emulate.
post31,
Did you find out what the reciever was doing before and after you modified it.
I suspect the cause and solution had nothing to do with what you are describing.
I suspect there was science in the cause and science in the solution.
Your version of the (non)science is too far into infinity for me to understand and emulate.
There is no one-fits-all solution for that. I suspect that many people who complain about weak bass have a suboptimal combination of speakers, amp and power supply.
For satisfying bass you need sufficient voltage, sufficient current and an amplifier that can deal with the reactive energy that comes back from speaker and crossover.
E. g. a 4 Ohm speaker with 2x1000µF power supply capacitance is likely to see too little voltage due to supply sagging every now and then. An 8 Ohm speaker combined with a 2x18 V transformer is likely to see too little voltage whatever the supply capacitance. A bass-reflex speaker with a complex crossover may require flyback diodes from the amplifier output to the rails to deal with back-EMF. A woofer with a heavy membrane that is optimised for low bass, will need more voltage and current below 100 Hz than a chipamp can provide.
Combine the knowledge you can gather from the closed-box.xls spreadsheet on the Linkwitz homepage with the information contained in the equal loudness curves to find out how much voltage and current your speakers need. Then check, whether your chipamp can deliver that. If it can, but still has weak bass, tweak the filters.
You will have to file that claim with NSC. Although it will be hard to justify that a chipamp should have better s/n ratios than 98 dB @ 1 W like e. g. the LM3875.
Thank you.
I changed only the rectifier, removing the MUR860's and replacing them with the rectifier pictured in this project: Synergy - LM3875 Gainclone Amplifier (Chip Amp)
Goodbye heat, hello reception!
Yes you're right. My example could be either diode noise and/or transformer noise.
But, I'm under the impression that I wanted DC and I'm sure that the MUR860's were "insufficiently helpful" at obtaining clean DC in a linear supply.
You see, a DC coupled amplifier with its four 1500uF caps getting dirty DC just seems like a mistake even though it happens to work. So, its just amazing that the extent of the problems were disproportionate frequency response and extra heat.
I'm also under the impression that installing larger caps (or assembling effectively larger caps) cannot substitute for a small filter when the point is removing high pitched noises.
For added noise blocking performance, don't we use/stack an additional filter?
That added filter seems better than trying to stop HF noise by adding 10,000uF caps which only expand the bandwidth, much like trying to catch a mosquito with a really big butterfly net. Of course, the 10,000uF do other useful things than just adding extraneous bandwidth to the pre-existing filter.
And, I'm under the impression that gainclone design is already making as much bass as its going to do, so that the way to hear more bass is to alter and clean the reference in order to level the response. . . like dirty DC power of a DC coupled amplifier wasn't a good start if we want a level response, so that needs patched.
Last edited:
So very elegant!
As for the input load directly at (not isolated via cap) the LM3875 gainclone, what do you think of these three options (with the input cap being located upon the RCA jack):
10k load at the amp board, no potentiometer (10k)
22k load at the amp board, 20k potentiometer (11k)
22k load at the amp board, no potentiometer (22k)
And how would these options affect bass in typical use situations?
As for the input load half indirectly coupled with the LM3875 in this option:
22k load at the amp board, 4.7uF, 20k potentiometer (22k DC, 20k AC)
How does it adversely affect the bass in typical use situations?
Here's another popular option:
22k load at the amp board, 4.7uF, 50k potentiometer (22k DC, 50k AC)
Why is it a popular option despite worsened mismatch?
Thanks!
100uF at the amp board may get you a lot more mids and highs, clean sure, but quite loud.
Options:
The 1500uF that comes with the audiosector board
470uF
100uF//100uF//100uF
330uF (either stereo or monobloc)
220uF (stereo models)
100uF//100uF (stereo models)
Amp board caps for non-inverting LM3875: As in 300uf minimum for monobloc/dual mono -or- 200uF minimum for stereo models.
Or choose your power circuit based on amp design:
Smaller goes brighter and non-inverting LM3875 gainclone already has loud mids, so use 330uF~1500uF sizes.
Larger goes warmer, and inverting LM3875 decibel dungeon design is already warm, so use 100uF~330uF sizes.
EDIT:
330uF isn't exactly a "can't miss" size because the caps do vary in quality (especially when used directly pin to pin with a sensitive semiconductor device); however, many of the modern switching amps use 100uF//100uF//100uF per rail at the amplifier board because its simply easier to find and use good quality 100uF caps even with a random selection of capacitor.
Last edited:
The last option will lead to a roll-off one octave below the other two. Depending on the input cap size you may or may not notice any difference.
That is 22k DC and 10k5 AC.
Your question includes two fuzzy concepts. For one thing everybody has a different idea of what good bass is, hence a different perception of what is an adverse effect on it. The second thing is that typical use is also very different. Most people will use their equipment most of the time to create some constant background stream. But that does not make it the typical use for everybody.
What is good bass for you? That extremely clean and well-defined but not punchy bass of OB speakers? Or that powerful bass you get from PA and horn speakers that kicks you in the chest and robs you of your breath or affects your heart rate? Or rather that boomy sound you sometimes get from bass-reflex speakers or boom cars?
Filters have several effects.
One is the frequency roll-off. In the case of a high-pass filter you want to push that so low that the audible frequencies are not affected. At the same time you want is as high as possible to filter any unnecessary disturbances and to not make the amplifier work to reproduce tones you won't be able to hear anyway.
A side effect are phase-changes. We usually don't want them to happen at all, but they cannot be avoided. To keep them as low as possible you want as low a roll-off as possible.
Another side effect are distortions that come from imperfections in the filter components, especially capacitors. That is why some people prefer DC-coupling and why others play around with different capacitor types from different manufacturers.
Because people don't know better and copy from somebody whom they believe to be competent.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.