NFB cap sizing routine is similar for most amplifier designs.
AndrewT has already given correct information for determining NFB cap size, and that is applicable to both of the amplifiers that you mention (and many more). The theory is solid and I can verify that his method also works optimally in practice on a wide variety of real amplifiers.
It is also true that ideal support circuits for each different model of power op-amp is different per each; so, just remember that the resistor series to the NFB cap also affects the capacitance value needed. Compensation options that directly govern harmonic tone are mostly sealed inside the chip--compare that discrete designers endlessly adjust compensations in simulators. But, with a chip-amp the support circuit options must be adjusted to a specific model chip so that there is not either overcomp dullness or undercomp harshness. Since we can't change the internal design, we push the chip-amp's support circuit around until eventually arriving at "just right" compensation that gives an appreciable level tone. For the same reason, op-amp rolling is mostly inappropriate and could only make improvement if the existing support circuit was off the mark or made for the wrong chip.
P.S.
Exact support circuit for an LM3886 is the MyRef.
OK, Fine. I know you are saying that, Mr. AndrewT's rule does'nt need to be implemented on LM1875, so delibrately. I agree.
But at least, i feel that we can keep his rule in mind while designing or changing any input stage capacitor value or NFB capacitor value, to achieve low distortion, so that no potential is developed across the NFB capacitor.
Then i request you to please consider LM3886, this time and come back to the post38 question.
What if we have both filters in our input? Which of these RC constants do we compare to find the value of NFB capacitor?
a) Input low pass filter capacitor's RC constant Vs NFB capacitor RC constant, or
b) Input high pass filter capacitor's RC constant Vs NFB capacitor RC constant.
This is what i am really confused about, in general for every amplifier's input design stage or selecting the value of NFB capacitor.
Meanwhile, i see your circlophone project thread but i could not read it for long due to a shortage of time. May be some other time. You are doing a lot of hard work to build that project. Well done! Just keep it up. I wish this project will give you excellent outputs and will be 100% true to your expectations.
Thanks.
But at least, i feel that we can keep his rule in mind while designing or changing any input stage capacitor value or NFB capacitor value, to achieve low distortion, so that no potential is developed across the NFB capacitor.
Then i request you to please consider LM3886, this time and come back to the post38 question.
What if we have both filters in our input? Which of these RC constants do we compare to find the value of NFB capacitor?
a) Input low pass filter capacitor's RC constant Vs NFB capacitor RC constant, or
b) Input high pass filter capacitor's RC constant Vs NFB capacitor RC constant.
This is what i am really confused about, in general for every amplifier's input design stage or selecting the value of NFB capacitor.
Meanwhile, i see your circlophone project thread but i could not read it for long due to a shortage of time. May be some other time. You are doing a lot of hard work to build that project. Well done! Just keep it up. I wish this project will give you excellent outputs and will be 100% true to your expectations.
Thanks.
Last edited:
OH, I see.
Series with the input is a cap for blocking DC, the Input Cap. Keep this an octave higher pitched than the NFB cap.
With a typical non-inverting amplifier, observe the Input cap transporting signal to IN+, and observe the NFB cap transporting signal to IN-.
Parallel with the input (as a load) is a little picofareds value cap for filtering off non-audio HF noise that could irritate the audio amplifier. We call it the RF blocker. Consider this a compensation part, not an audio signal cap. Observe the Picofareds size RF blocker defending the amplifier by transporting non-audio signal to ground.
It also has a partner that you haven't seen yet: Partner to the RF blocker is a cap or RC (also Picofareds) that goes parallel with the feedback resistor for compensation adjustment duties, such as slowing down the amplifier. It is sometimes needful to run the amplifier slower than the output devices max. Since these caps are non-audio sized, you'll want LTSpice simulator and/or a scope for adjusting the values.
All of the information has been compressed into post 1 for fast reading: Building Elvee's Circlophone: Documentation, Parts, Accessories, & beginner friendly
Last edited:
Neither you nor i. None of us is against Mr.AndrewT rules.
Nothing. I got a false impression that you don't want me to implement the RC constant comparison rules to determine the NFB capacitor values because you asked me to stick with LM3886 when i wanted to do this with LM1875.
Let me read and understand your next post........
Thanks.
Nothing. I got a false impression that you don't want me to implement the RC constant comparison rules to determine the NFB capacitor values because you asked me to stick with LM3886 when i wanted to do this with LM1875.
Let me read and understand your next post........
Thanks.
Less complex than that. This thread is titled "Buffer/Gain stage suggestions for LM3886"
Off topic discussion is disrespectful. Therefore, here I want to talk about LM3886. And maybe a moosefet.
If you want to discuss LM1875, please select a thread titled LM1875.
1.It would be more clear and i will be more confident if i can understand an example.
Suppose we have this signal capacitor value as 10uf with || 22k(input impedance).
RC constant will be say: 220ms. This should be an octave higher pitched than the NFB capacitor. So the RC constant of the NFB capacitor should be 110ms or lower.Am i right?
2.Do this partner resides inside the chip?
LM3886 circuit has a 1.5nf as the RF blocker. NFB capacitor RC costant, should be at least half octave higher than the RC constant of the RF blocker, as stated by Mr.AndrewT. Am i right here too?
In short and in one line, we can say that the NFB capacitor should be half octave higher than the compensation capacitor but one octave lower than the DC signal blocking capacitor?
Thanks.
Sorry, but that's not quite right.
You need to ignore non-audio parts during NFB cap size calculation.
And that means. . .
You need to ignore the RF blocker during NFB cap size calculation.
P.S.
The schematic on this thread seems to show omitted input cap (bad design), meaning 0hz, and in this case, attempt to calculate the NFB cap to one octave below 0hz. Expect the size to be inconveniently large.
P.P.S.
Input and NFB caps are calculated for the bass.
RF blocker is calculated to higher than treble.
Last edited:
if you want 20Hz performance from the amplifier, then try setting the input filter to 2Hz.
This will give the generally accepted F-1dB ~ 4Hz.
That 2Hz is equivalent to 80ms.
The NFB should be set to >80ms*sqrt(2) i.e. >113ms
Let's suppose you use 1k0 & 120uF (or 1k2 & 100uF, or .......) for 120ms.
The PSU RC should be >120ms*sqrt(2).
Let's use 8ohms speaker. That requires +-21mF (8*0.021 ~= 0.120*1.414) for each 8ohms speaker. That 20mF is the value I have been recommending for some years since finding this correlation between the three time constants on this Forum.
Now back to my first statement and the 20Hz to 2Hz ratio.
You can choose any amplifier bandwidth, be it 1Hz or 10Hz or 1kHz (for a treble only speaker). Then all the rest follows from that first choice.
What you should not do is choose a 20Hz F-3dB for the amplifier input filter and then set the NFB & PSU to suit ....... and then change your mind and alter ONLY the input filter looking for more bass. That will completely upset the way the amplifier is trying to copy and reproduce the input signals.
This will give the generally accepted F-1dB ~ 4Hz.
That 2Hz is equivalent to 80ms.
The NFB should be set to >80ms*sqrt(2) i.e. >113ms
Let's suppose you use 1k0 & 120uF (or 1k2 & 100uF, or .......) for 120ms.
The PSU RC should be >120ms*sqrt(2).
Let's use 8ohms speaker. That requires +-21mF (8*0.021 ~= 0.120*1.414) for each 8ohms speaker. That 20mF is the value I have been recommending for some years since finding this correlation between the three time constants on this Forum.
Now back to my first statement and the 20Hz to 2Hz ratio.
You can choose any amplifier bandwidth, be it 1Hz or 10Hz or 1kHz (for a treble only speaker). Then all the rest follows from that first choice.
What you should not do is choose a 20Hz F-3dB for the amplifier input filter and then set the NFB & PSU to suit ....... and then change your mind and alter ONLY the input filter looking for more bass. That will completely upset the way the amplifier is trying to copy and reproduce the input signals.
Last edited:
RF attenuation
The amplifier must never be presented with a "too fast" rise time input signal. That could be at the conventional inputs (+In & -IN) but also the speaker output (OUT) and also the power pins (Vcc & Vee).
We sort the +IN & -IN by inserting an RF input filter that allows the audio signal that we require to pass and attenuates as much of the RF interference as possible. That passing audio and attenuating as much as possible is a BIG compromise.
Let's take 20kHz as the absolute max audio frequency that a hypothetical Member wants to pass. That requires F-3dB @ ~40kHz (F-1dB ~20kHz).
This will be provided by a RC of 4us.
Alternatively one may want to pass 90kHz from 192kS/s digital equipment and looking for F-0.1dB = 90kHz. That would require F-3db out at about 2MHz, requiring an RC of 80ns. Two extremes in the range of audio signal passing requirements.
You can see that the RF attenuation of a 80ns filter is much, much lower than a 4us filter.
The commonly adopted range of RC for the RF attenuation of an audio power amplifier is generally 1.5us to 220ns.
You choose where in that range you want to be. 1.5us removes some audio and more of the the RF. 220ns removes virtually none of the audio and in return removes much less RF. Listen and test and decide.
The amplifier must never be presented with a "too fast" rise time input signal. That could be at the conventional inputs (+In & -IN) but also the speaker output (OUT) and also the power pins (Vcc & Vee).
We sort the +IN & -IN by inserting an RF input filter that allows the audio signal that we require to pass and attenuates as much of the RF interference as possible. That passing audio and attenuating as much as possible is a BIG compromise.
Let's take 20kHz as the absolute max audio frequency that a hypothetical Member wants to pass. That requires F-3dB @ ~40kHz (F-1dB ~20kHz).
This will be provided by a RC of 4us.
Alternatively one may want to pass 90kHz from 192kS/s digital equipment and looking for F-0.1dB = 90kHz. That would require F-3db out at about 2MHz, requiring an RC of 80ns. Two extremes in the range of audio signal passing requirements.
You can see that the RF attenuation of a 80ns filter is much, much lower than a 4us filter.
The commonly adopted range of RC for the RF attenuation of an audio power amplifier is generally 1.5us to 220ns.
You choose where in that range you want to be. 1.5us removes some audio and more of the the RF. 220ns removes virtually none of the audio and in return removes much less RF. Listen and test and decide.
Last edited:
Please, see post 22 and post 24. I am sure he is talking about the input capacitor filter(high pass filter) and not the RF Blocker(low pass filter)? or, is it the other way?
Now see post 26 by Mr. Zakman35..............
He has changed the component values of the RF Filter (Low pass filter, 1k & 1.5n, Fc=106Khz(not 106Hz) and RC constant = .0015ms(not 1.500ms)) to 9.1K and 10uf, which now still acts as a Low pass filter but no more as a RF Blocker. He has selected the NFB capacitor now according to this RF Blocker capacitor's RC constant, which he has changed from 1.5nf to 10uf.
He has compared 91ms (new RF blocker's RC Constant) to choose the value of 151ms for NFB capacitor.
This confuses me still.
Thanks.
Last edited:
Thanks, Mr.Andrew,
Yes! i can calculate all the values easily. However, before executing any build, i will post the final, complete circuit diagram here on chipamp forum to get the approval of it with all you technicallly experienced guys here.
So, don't worry about me doing things right way because i can never go wrong with all you out there, to help and support me with each and every step. Actually, I trust your knowledge and experience achieved with years of hard work and struggle, more than anything here on diyaudio. I can never go wrong, until you all are here with me.
And, thanks for my concern about being able to calculate correct values or not. I felt something really so nice about it, inside.
Thanks.
Yes! i can calculate all the values easily. However, before executing any build, i will post the final, complete circuit diagram here on chipamp forum to get the approval of it with all you technicallly experienced guys here.
So, don't worry about me doing things right way because i can never go wrong with all you out there, to help and support me with each and every step. Actually, I trust your knowledge and experience achieved with years of hard work and struggle, more than anything here on diyaudio. I can never go wrong, until you all are here with me.
And, thanks for my concern about being able to calculate correct values or not. I felt something really so nice about it, inside.
Thanks.
Last edited:
the PSU behaves like a filter as well.
It's RF is the Capacitance times the speaker impedance.
It too should exceed the amplifier RC, i.e NFB RC * sqrt(2) = 120 * 1.414 and that must be less than or equal to 8ohms * smoothing capacitance.
Therefore smoothing capacitance >= 120*1.414/9 >=21mF on each supply rail since only one supply rail is active during any half cycle of the waveform.
It's RF is the Capacitance times the speaker impedance.
It too should exceed the amplifier RC, i.e NFB RC * sqrt(2) = 120 * 1.414 and that must be less than or equal to 8ohms * smoothing capacitance.
Therefore smoothing capacitance >= 120*1.414/9 >=21mF on each supply rail since only one supply rail is active during any half cycle of the waveform.
1.Sorry, but which smoothing capacitor are you really talking about?
a) Those 100n, right at the chipamp pins.
b) Those little bigger ones close to the chipamp pins(100mf, 220mf, 330mf, 470mf, 1000mf, or 1500mf).
c) Those bigger and bulky electrolytes in the power supply board(certainly not)., or
d) Those tiny capacitors(10n or 100n) which are used as the bypass capacitors around bigger electrolytes on the PS board.
20uf or 21uf capacitor is suggested in place of which of the above, a, b, c (certainly not) or d?
2. Does ONLY changing the value of NFB capacitor will increase or decrease the chipamp gain in any way? Offcourse, it will change the F-3db cut of point.
Thanks.
Last edited:
This question is not clear. That looks impractical.
For mf, millifared, here is a helpful conversion chart.
1 milli-Farad = 0.001 F
1 milli-Farad = 1mF
1 milli-Farad = 1000 uF
1 milli-Farad = 1000000 nF
1 milli-Farad = 1000000000 pF
P.S.
Shortcut for writing 470 micro-Fareds is: 470u
And, assuming a perfectly stable amp with good layout, that 470u is the figure recommended at close to the LM3886's pins by National Semiconductor's application note #1192 (AN1192.pdf). In their test conditions, it is probable that their umbilical from power board was nearby 4cm in length. If this cable/trace is longer, the needed cap size may increase. If zero length, the needed cap size may decrease. However, less than 330u may be unpleasant. For example 100u = very clear but with the risk of a disturbing/corrupted frequency response. A 22uF figure is not in the ballpark; however, 22mF is just a bit too big for the amplifier board albeit functional.
When comparing 2 caps that are big enough to pass audio band intact, the answer is: The gain doesn't change.
Last edited:
Thanks danielwritesbac, for clearing my confusion about micro farads and milli farads.
The local electronic guys here are very lazy to perform conversions and differentiate between different units of measurements.
They sometimes even, call microfarads as MFD in short.Just because they have hardly used millifarads and farad values ever. SO, most of them are not familiar with other uncommon values like millifarad and farad. They are only exposed to micro farad, pico and nano farads (practical units).
When i asked them what for MFD stands.They always told me, that MFD represents "microfarads" along with no arguments as they hear it from others. SO, i could too never differentiate between milli and micro farads, until now. I considered ('mf' and 'mfd') or ('m' and 'u') both as same. However, i am familiar with other conversions like pico, nano, micro and farad.
Theoretically, you seems correct, and i want to agree with your answer, but two points are still disturbing me a very long last.
Sorry! its a littlebit offtopic.
1. Gain bandwidth product is a constant value for every chipamp. So, if we increase the bandwidth, gain will decrease or if we decrease the bandwidth, gain will increase, and vice-versa.
Now, By changing or increasing the value of the NFB Capacitor, did'nt we increasing the Bandwidth (which in turn leads to decrease in gain)?
2.Its a practical and totally offtopic point and not related to this thread in any way, so i will start a new thread for it, maybe some other time. It's related to my experiment and experiences with boombox chipamps.
Thanks.
The local electronic guys here are very lazy to perform conversions and differentiate between different units of measurements.
They sometimes even, call microfarads as MFD in short.Just because they have hardly used millifarads and farad values ever. SO, most of them are not familiar with other uncommon values like millifarad and farad. They are only exposed to micro farad, pico and nano farads (practical units).
When i asked them what for MFD stands.They always told me, that MFD represents "microfarads" along with no arguments as they hear it from others. SO, i could too never differentiate between milli and micro farads, until now. I considered ('mf' and 'mfd') or ('m' and 'u') both as same. However, i am familiar with other conversions like pico, nano, micro and farad.
Theoretically, you seems correct, and i want to agree with your answer, but two points are still disturbing me a very long last.
Sorry! its a littlebit offtopic.
1. Gain bandwidth product is a constant value for every chipamp. So, if we increase the bandwidth, gain will decrease or if we decrease the bandwidth, gain will increase, and vice-versa.
Now, By changing or increasing the value of the NFB Capacitor, did'nt we increasing the Bandwidth (which in turn leads to decrease in gain)?
2.Its a practical and totally offtopic point and not related to this thread in any way, so i will start a new thread for it, maybe some other time. It's related to my experiment and experiences with boombox chipamps.
Thanks.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.