Tuning the filter
That is the problem with the Class D amps that have the filter outside the feedback loop. The frequency response is very load dependent at the top. I would try a 12mH coils and if it it still too bright keep increasing until the sound is just on the edge to maintain as much ripple and loss (good for the sound even if it is inefficient) as possible. You can then add a Zobel to tame the last of the edginess. (R and C in series, the whole in parallel between the speaker terminals). Start with .47uf and 10 pieces of 70 ohm 1/4 watt resistors twisted together in parallel to make 7 ohms. Measure and listen as you cut one resistor at a time to raise the value until you get just enough peak remaining to bring out the detail but not so much that it still sounds too bright. Probably around 10-15 ohms. Once you get an idea of the value, you can replace the parallel resistor mess with something nicer in a 5 or 10 watt resistor. It is a hassle to do this and the results are only good for one particular speaker but it is worth it in order to get the best resolution from the Tripath chips. I am sure you will find that a text book filter with no peak will sound veiled so measure AND listen as you go. This is probably of more interest to the readers of the mod thread so I will also copy my reply over there.
.
http://www.diyaudio.com/forums/clas...path-board-tc2000-tp2050-179.html#post2286520
.
That is the problem with the Class D amps that have the filter outside the feedback loop. The frequency response is very load dependent at the top. I would try a 12mH coils and if it it still too bright keep increasing until the sound is just on the edge to maintain as much ripple and loss (good for the sound even if it is inefficient) as possible. You can then add a Zobel to tame the last of the edginess. (R and C in series, the whole in parallel between the speaker terminals). Start with .47uf and 10 pieces of 70 ohm 1/4 watt resistors twisted together in parallel to make 7 ohms. Measure and listen as you cut one resistor at a time to raise the value until you get just enough peak remaining to bring out the detail but not so much that it still sounds too bright. Probably around 10-15 ohms. Once you get an idea of the value, you can replace the parallel resistor mess with something nicer in a 5 or 10 watt resistor. It is a hassle to do this and the results are only good for one particular speaker but it is worth it in order to get the best resolution from the Tripath chips. I am sure you will find that a text book filter with no peak will sound veiled so measure AND listen as you go. This is probably of more interest to the readers of the mod thread so I will also copy my reply over there.
.
http://www.diyaudio.com/forums/clas...path-board-tc2000-tp2050-179.html#post2286520
.
Thanks. Did you measure this with your real speakers ? or with a resistive load only ? In all cases the shape of you curve is typically what I see if I plot my measurements, but the peak is really worse
If you loaded only with a resistor, I suggest you try with a speaker to see how bad it goes due tothe variable impedance VS frequency.
Regards
Hi Sendler
always nice to read you
I fully agree : getting th ebest of thse Tripath chips requires a case by case tuning. Not any DIY guy has a scope and a generator to do so. So that is a clear weakness of this technology.I wonder if the circuit could be modified to take the feedback after the filter, have you ever considered trying ? I have two boards, one that I use, one in spare just in case, I could try to have a look if there is a chance if can work
regards
Gaël
in loop filter
I think at that point you would be looking to start out with a different amp.Hi Sendler
always nice to read you
I wonder if the circuit could be modified to take the feedback after the filter, have you ever considered trying ? I have two boards, one that I use, one in spare just in case, I could try to have a look if there is a chance if can work
regards
Gaël
Thanks. Did you measure this with your real speakers ? or with a resistive load only ? In all cases the shape of you curve is typically what I see if I plot my measurements, but the peak is really worse
If you loaded only with a resistor, I suggest you try with a speaker to see how bad it goes due tothe variable impedance VS frequency.
Regards
sure, with speaker the result is worse.
I wanted to say that some DIY 'only change the intuttanza and simply listen. this is obviously wrong.
the best solution is two feedback or post filter feedback, this is clear.
So I think we all agree that those TRIPATH are OK for a sub amplifier but not as full range audio band, unless the user is capable to fine tune the output to match his speakers and prevent this high frequency level increase affecting the listening comfort. At least, for me I consider I can not use this amplifier as my main amplifier, despite its other advantages.
thanks
Other amp?
Have you measured your previous amp that you are comparing the Tripath's sound to the same way? I wonder if it is down at 20KHz?
But the facts are there. It sounded too bright on my speakers that are reputed for being real flat.
I bought a scope and a generator last week to check, and what I measured confirms what my ear heard. For a while in audio, measurements confirm listening impressions. The increase begins as early as 15-16 KHz so no doubt it is linked
Hi scott,
I will do it tonight to confirm. Will keep you posted.
Just a thought : the recommended 2n order filter is for EMI reasons, right ? But who still listens to AM radio at 600 or 700 KHz ?
Why not just keep the inductors to be on a 1st order low pass filter ?
Indeed, 1st order, means no more Q factor, so no overshoot, so no hassle with speaker matching ! Ever tried ? (but I plan to keep the common mode capacitor)
Hi guys
JOB DONE : my SURE TRIPATH 2050 now officially runs WITHOUT second order output filter !
Just like my measurements confirmed the listening when using the recommended second order filter (overshoot in trebles, too bright sound), now that I simply run in first order filter, the bright sound disappearead like magic.
The overshoot is gone. My system now sound absolutely flat and listening simply confirms. First order just gives less carrier attenuation, ok, but impossible to have frequency response overshoot with a first order.
I have not made may listenings yet, but I tested with as many different music types I could in a short time: Nina Simone, The Clash, Bob Marley, Snoop Dog, electro, Lenny Kravitz, Massive Attack. In ALL cases the sound is better. More details, more natural soundstage, cymbals level is now right.
Please DO TEST on your system and let me know. No risk, you can easily switch back later to second order if you like. But please tell me how it sound in your system.
PS : my coils are 10 µH / WURTH HIGH CURRENT
JOB DONE : my SURE TRIPATH 2050 now officially runs WITHOUT second order output filter !
Just like my measurements confirmed the listening when using the recommended second order filter (overshoot in trebles, too bright sound), now that I simply run in first order filter, the bright sound disappearead like magic.
The overshoot is gone. My system now sound absolutely flat and listening simply confirms. First order just gives less carrier attenuation, ok, but impossible to have frequency response overshoot with a first order.
I have not made may listenings yet, but I tested with as many different music types I could in a short time: Nina Simone, The Clash, Bob Marley, Snoop Dog, electro, Lenny Kravitz, Massive Attack. In ALL cases the sound is better. More details, more natural soundstage, cymbals level is now right.
Please DO TEST on your system and let me know. No risk, you can easily switch back later to second order if you like. But please tell me how it sound in your system.
PS : my coils are 10 µH / WURTH HIGH CURRENT
The increase begins as early as 15-16 KHz so no doubt it is linked
Did you measure the acoustic output of the tweeters?
I'm driving a 4 ohm nominal speaker, but I do not believe this has any importance.
The output filter of a class D amp is more made to comply with the radiation figures of the american FCC and european CE organizations. Consumer products obey to different rules from DIY stuff
Do not consider this filter like a signal filter , for example a speaker crossover.
The speaker crossover is not approximate because it filters the signal. It must be accurate or you will not hear what you expect.
The output filter here is a cleaning filter, removing the noise generated in th 600 / 700 KHz range according to Tripath.
I was suprised to see on my oscilloscope how noisy the output signal is,while a linear class AB amp has a nice output signal. If you input a clean sinewave, you see a clean sinewave. If you input (like I did) a clean sinewave (thin trace on the scope), you get a noisy output sinewave (the trace is thick on the scope). That is because of all the noise inherent to the way class D amp work. SO your ear basically integrate and you hear the enveloppe, not the noise which is too "far" away in the hundreds of Khz.
This can be quite embarassing at first when you see such measures. On a linear amp, such a signal would probably mean poor listening, but not on class D. And the evidence is what I hear which sounds 1000 times better than how bad it looks on a scope.
Basically, if your neighbour is a grandma still istening to AM radio, she will probably hate you, but nowadays, FM rules so who cares of the 600 / 700 KHz carrier is spread by the antenna former by long speaker cables. In fact, the absolute best would be to make active speaker, ie put the amp behind each speaker to have like 10 cm wire only. It will not only be an advantage for noise radiated, but also better for bass response as no loss in cables. SO why not switch to active speakers with Tripath chips ? I think it is the ultimate solution.
The output filter of a class D amp is more made to comply with the radiation figures of the american FCC and european CE organizations. Consumer products obey to different rules from DIY stuff
Do not consider this filter like a signal filter , for example a speaker crossover.
The speaker crossover is not approximate because it filters the signal. It must be accurate or you will not hear what you expect.
The output filter here is a cleaning filter, removing the noise generated in th 600 / 700 KHz range according to Tripath.
I was suprised to see on my oscilloscope how noisy the output signal is,while a linear class AB amp has a nice output signal. If you input a clean sinewave, you see a clean sinewave. If you input (like I did) a clean sinewave (thin trace on the scope), you get a noisy output sinewave (the trace is thick on the scope). That is because of all the noise inherent to the way class D amp work. SO your ear basically integrate and you hear the enveloppe, not the noise which is too "far" away in the hundreds of Khz.
This can be quite embarassing at first when you see such measures. On a linear amp, such a signal would probably mean poor listening, but not on class D. And the evidence is what I hear which sounds 1000 times better than how bad it looks on a scope.
Basically, if your neighbour is a grandma still istening to AM radio, she will probably hate you, but nowadays, FM rules so who cares of the 600 / 700 KHz carrier is spread by the antenna former by long speaker cables. In fact, the absolute best would be to make active speaker, ie put the amp behind each speaker to have like 10 cm wire only. It will not only be an advantage for noise radiated, but also better for bass response as no loss in cables. SO why not switch to active speakers with Tripath chips ? I think it is the ultimate solution.
Bought a signal generator and as scope last week, but not a soundmeter yet ! I do not think it would bring me anything anyway as it sounds right, and this is what matters. I trust my manufacturer measures, and they are damn flat originally. What is the tought behind this measurement you ask ?
Coil only
Now that is a radical filter. How does the scope across the speaker terminals look now? Can you see a big signal that would correspond to the switching frequency? Do you have a portable AM radio so that you can take a survey of the noise to see if it radiates outside of your house.
Just want to say thanks to MusicSeeker for his experiment, it restored my fun with my modded Sure board. I too had problems in the highs, I think I've tried over 10 different filter combinations (different both in values and types of caps and coils...) and a lot of simulations in PSpice (though I'm just have som basic electronic knowledge). But I dont now how to simulate speakers with high impedance peaks in certain higher freqs... (my speakers don't drop below 5 ohms though)
Anyway, got rid of all but the coils and my problems in the treble are completely gone! I live in the contry side, dont think I will cause any problems with RF. Have just my digital reciever in the other end of my house, but could'nt find anything on the AM between 500-800 khz (playing the Sure quite loud). One should use a analog AM radio to test though, think the digital "squelch" might filter the RF as interference... (which it is )
Anyway, got rid of all but the coils and my problems in the treble are completely gone! I live in the contry side, dont think I will cause any problems with RF. Have just my digital reciever in the other end of my house, but could'nt find anything on the AM between 500-800 khz (playing the Sure quite loud). One should use a analog AM radio to test though, think the digital "squelch" might filter the RF as interference... (which it is
)Just want to say thanks to MusicSeeker for his experiment, it restored my fun with my modded Sure board. I too had problems in the highs, I think I've tried over 10 different filter combinations (different both in values and types of caps and coils...) and a lot of simulations in PSpice (though I'm just have som basic electronic knowledge). But I dont now how to simulate speakers with high impedance peaks in certain higher freqs... (my speakers don't drop below 5 ohms though)
Anyway, got rid of all but the coils and my problems in the treble are completely gone! I live in the contry side, dont think I will cause any problems with RF. Have just my digital reciever in the other end of my house, but could'nt find anything on the AM between 500-800 khz (playing the Sure quite loud). One should use a analog AM radio to test though, think the digital "squelch" might filter the RF as interference... (which it is
Thanks for the "thank you"
and glad you found improvementI posted a reply two days ago but for some reason it is not here, so here we go again :
I connected my scope at the output of the amplifier loaded by my speakers
I noticed a high level of switching frequency noise (around 2V peak to peak). That is a lot.
I also recently rebuilt an old tube AM/FM receiver. It is only at 3 meters from my speakers.
- switched off amp
- switched on radio to AM
- tuned all across the band, only found one station at 800 KHz (the rest is strong noise, not like FM noise when no station)
- switched on my amp and could still listen to this station at 800 Khz
As the switching frequency is in the 650 to 700 KHz range, and it most probably generates a lot of spurious and harmonics, I was expecting this station to be totally polluted, but it is not. I see only two options to explain
- either the switching frequency only radiates at 650-700 KHz, which I doubt
- either, for some reason, the particular length of my cables, terminated by the speaker impedance, makes a poor antenna and therefore doesn't disturb that much (???)
Investing in a scope is enough for now, I don't plan to by a spectrum analyzer to check this ;
BUT : 2Vpp at 650 Khz is some kind of energy still. This must go to the speakers. At first I thought that it should slightly heat them even without music. But the tweeter coild impedance at this frequency must be damn high, like various hundreds of ohms. So basically, I think that the current really going through the tweeter is ridiculous if not neglectable.
If people really cause troubles to neighbours in 1st order filter, I would recommend to try a notch filter at the amp ouput terminal, (LC and C in serie, tuned at the right resonant frequency) in parallel. This should not affect the sound. Another way is to shield the speaker cables. But the absolute best to me is to use the Sure modules as close as possible to the speakers, like making active speakers. I do have 2 of those Sure stereo amps, I'm planing to use only one channel per speaker.
Any other member experienced the 1st order low pass ? I'd be deligthed to know if you found the benefits I found.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.