I am looking for about 220uF non polar cap for using in the global negative feedback. I have seen a lot of amps using 220uF for AC ground to prevent excessive DC offset. That is a HUGE and very expensive cap.
1) Why 220uF? If you calculate with 500ohm resistor, the break frequency is less than 2Hz!!! Seems like I can use 10Hz as the minimum, then the cap reduce to 33uF!!! That's still big, but at least it's smaller and cheaper.
2) If it has to be very high value cap, is film cap the only ones I can use? Can I use two electrolytic back to back. Then I can get two 470uF 25V back to back cheap and small. I just parallel with say 1uF and 0.1uF film cap.
3) Any suggestion I can get this cap cheaper? Digikey, Mouser and Newark only stock high voltage ones like over 100V. That make the size huge and expensive.
Thanks
1) Why 220uF? If you calculate with 500ohm resistor, the break frequency is less than 2Hz!!! Seems like I can use 10Hz as the minimum, then the cap reduce to 33uF!!! That's still big, but at least it's smaller and cheaper.
2) If it has to be very high value cap, is film cap the only ones I can use? Can I use two electrolytic back to back. Then I can get two 470uF 25V back to back cheap and small. I just parallel with say 1uF and 0.1uF film cap.
3) Any suggestion I can get this cap cheaper? Digikey, Mouser and Newark only stock high voltage ones like over 100V. That make the size huge and expensive.
Thanks
Last edited:
Hi,
The feedback capacitor is a quagmire of conflicting opinions,
DC servos are one way out for dual rail designs.
https://www.madisoundspeakerstore.com/surplus-capacitors/10.0-mfd-polypropylene-cap-10mfdp/tyee/
Are very cheap for 10uF film.
rgds, sreten.
The feedback capacitor is a quagmire of conflicting opinions,
DC servos are one way out for dual rail designs.
https://www.madisoundspeakerstore.com/surplus-capacitors/10.0-mfd-polypropylene-cap-10mfdp/tyee/
Are very cheap for 10uF film.
rgds, sreten.
there are two LF poles. its best to have some offset between them IDK remember why. define the 1st LF spec ~ 1 octave lower than 20Hz. so let the inverting input be 2 octaves lower. I think its best to avoid extreme additive phase shifts when the 2 poles coincide?
The working voltage is quite low, but it's when things go south (open loop) they can see one rail or the other. most amps ignore failure modes.
leakage current is probably key so higher V rating can help there if using GP electros.
The working volt polarity can be worked out
The working voltage is quite low, but it's when things go south (open loop) they can see one rail or the other. most amps ignore failure modes.
leakage current is probably key so higher V rating can help there if using GP electros.
The working volt polarity can be worked out
Last edited:
Thanks
Can I use say two 470uF electrolytic back to back. Then I use say 1 uF film cap in parallel with 0.1uF with the two 470uF electrolytic?
Another option is to increase proportionally the value of NFB resistors. E.g. if the mentioned 500 ohm resistor is replaced with 5k, and other NFB resistors are replaced with 10x higher value, a 22uF capacitor can be used with the same result.
Of course, the effect of higher resistor values should be evaluated in regard to input bias current (in case of BJT), noise (for low noise amps), HF pole created by the NFB resistors with transistor equivalent capacity, and possibly some other interactions.
Of course, the effect of higher resistor values should be evaluated in regard to input bias current (in case of BJT), noise (for low noise amps), HF pole created by the NFB resistors with transistor equivalent capacity, and possibly some other interactions.
set the amplifier passband with the input filters.
Then any additional roll offs inside the amplifier should be set to wider than the input filters.
If you want to exercise your speaker to 20Hz and accept that the signal is already rolled off to -3dB at that 20Hz then set the input filter to 20Hz.
Then any additional roll-offs should be lower than 20Hz.
The NFB roll off when set LOWER than the input filter will have near zero AC signal across it. It will also have near zero DC offset across it.
In this operating condition an electrolytic contributes virtually zero distortion to the audio signal passing through the amplifier.
This allows you to use a low voltage reformed polar electrolytic and you will not hear any degradation of the audio signal.
Now the warnings:
1.
if the amplifier misbehaves then it can apply a large DC offset to the NFB capacitor. To protect the capacitor from damage, we can fit a diode or two to limit the maximum peak voltage that can be applied to the electrolytic. We can even do the same in the reverse direction to protect the electrolytic from excessive reverse voltage. These diodes also limit the voltage that can be applied to the -IN base lead of the LTP transistor.
2.
if you decide you would like a wider passband than the 20Hz chosen earlier, then many Members will advise you change the input filter to let more of the low bas signals to pass through.
Now the NFB capacitor will see a significant AC signal. This has TWO consequences.
2a.
the AC voltage will create some electrolytic distortion of the very low frequencies.
2b.
the AC voltage will overcome the small DC offset and allow the capacitor to become reverse biased on alternate half waves of the signal. This will damage the electrolytic in the longer term. Audio performance will deteriorate.
If you WANT to use the NFB capacitor as your LF bandwidth limiter then you must use a low distortion capacitor and it must be non polar/bipolar.
Both are bulky and expensive.
It is far easier and far cheaper to use a much larger value NFB capacitor so that the INPUT filter sets the bandwidth you require.
set the input filter to ~10Hz and 20Hz signal will be ~ 1dB down.
set the input filter to ~2Hz and 20Hz signal will be ~ 0.1dB down.
This latter is my recommendation.
But it requires the NFB roll off to be even lower, meaning an even larger value of low voltage polar electrolytic capacitor to avoid 2a and 2b
Then any additional roll offs inside the amplifier should be set to wider than the input filters.
If you want to exercise your speaker to 20Hz and accept that the signal is already rolled off to -3dB at that 20Hz then set the input filter to 20Hz.
Then any additional roll-offs should be lower than 20Hz.
The NFB roll off when set LOWER than the input filter will have near zero AC signal across it. It will also have near zero DC offset across it.
In this operating condition an electrolytic contributes virtually zero distortion to the audio signal passing through the amplifier.
This allows you to use a low voltage reformed polar electrolytic and you will not hear any degradation of the audio signal.
Now the warnings:
1.
if the amplifier misbehaves then it can apply a large DC offset to the NFB capacitor. To protect the capacitor from damage, we can fit a diode or two to limit the maximum peak voltage that can be applied to the electrolytic. We can even do the same in the reverse direction to protect the electrolytic from excessive reverse voltage. These diodes also limit the voltage that can be applied to the -IN base lead of the LTP transistor.
2.
if you decide you would like a wider passband than the 20Hz chosen earlier, then many Members will advise you change the input filter to let more of the low bas signals to pass through.
Now the NFB capacitor will see a significant AC signal. This has TWO consequences.
2a.
the AC voltage will create some electrolytic distortion of the very low frequencies.
2b.
the AC voltage will overcome the small DC offset and allow the capacitor to become reverse biased on alternate half waves of the signal. This will damage the electrolytic in the longer term. Audio performance will deteriorate.
If you WANT to use the NFB capacitor as your LF bandwidth limiter then you must use a low distortion capacitor and it must be non polar/bipolar.
Both are bulky and expensive.
It is far easier and far cheaper to use a much larger value NFB capacitor so that the INPUT filter sets the bandwidth you require.
set the input filter to ~10Hz and 20Hz signal will be ~ 1dB down.
set the input filter to ~2Hz and 20Hz signal will be ~ 0.1dB down.
This latter is my recommendation.
But it requires the NFB roll off to be even lower, meaning an even larger value of low voltage polar electrolytic capacitor to avoid 2a and 2b
Last edited:
A feedback capacitor has a major imact upon sound quality. To miminise the impact you need high quality capacitor; that is low esr and idealy low esl.
As Sreten says there are a many views on feedback capacitors. However the lowest esr and lowest esl I have found are Elna Silmic 2 and I normally use 220uf 35v. These are not that large and three months age cost me $1.50 each from a major supplier.
Don
As Sreten says there are a many views on feedback capacitors. However the lowest esr and lowest esl I have found are Elna Silmic 2 and I normally use 220uf 35v. These are not that large and three months age cost me $1.50 each from a major supplier.
Don
It does not, if you ensure that the LF bandwidth is set by the INPUT FILTER.
An electrolytic with zero AC signal across it contributes zero distortion to the passing signal.
It is filtering capacitors that add distortion, if they are not of the low distortion type.
Polypropylene (FKP/MKP) is one low distortion type.
Polyterephthalate (FKT/MKT/MKS/FKS) are not low distortion type.
Neither are non polar, nor bi-polar, electrolytics.
however
a pair of back to back bi-polar electrolytics have very low distortion and are nearly upto plastic film distortion performance.
Similarly back to back polar electrolytics are much lower distortion than single polarised and may exceed the distortion performance of a single non-polar electrolytic.
a pair of back to back bi-polar electrolytics have very low distortion and are nearly upto plastic film distortion performance.
Similarly back to back polar electrolytics are much lower distortion than single polarised and may exceed the distortion performance of a single non-polar electrolytic.
By all means use 2 back to back bi-polar electrolitics but remember when you do this the capacitance is reduced by 50% so you will need 2 470uf caps to = 235uf . Some people who have done this find that a small bias voltage applied to the two back to back cap junction reduces distortion further. I did this once using a 9v battery through a 10m ohm resistor.
There are a lot of problems using higher value. The feedback resistor is going to go from 10K to 100K, DC offset current of the input transistor comes into play, input capacitance can create a pole before the 0dB crossover that affect stability.
What is bipolar lythic?
Thanks
Last edited:
I checked the Elna Silmic 2, they seems to be electrolytic that has polarity. You put them back to back to get bipolar operation?
bipolar = non-polar electrolytic capacitors are designed for use in AC circuits, can be connected either orientation (also, according to Bateman's measurements distort less than polar types).
I would add, if input diff pair is well matched and has low drift, considering moderate Av=21, maybe you can replace C with short. Let's say, if BJT's are matched to 0.5mV Vbediff, overall DC offset would be 10.5mV, possibly tolerable for power amp.
Last edited:
- Status
- Not open for further replies.