You can hear phase smearing in bass range caused by capacitive coupling if done wrong. My preference in my "beater" amps is also capacitive coupling for safety. But there is an approach that minimizes the influence of the output cap; and this is to drop the corner freq low enough; preferably a decade below the bass freq. of interest (eg. 4700uf @ 8 ohm bookshelf speakers) and perform the subsonic filtering at the input (with a film cap). This will also improve power-losses in the coupling cap a tiny little bit and the bigger cap will usually tolerate larger ripple current.
A trickyer to stabilize, but also workable is to wire the gnfb around the output cap to further linearize the system.
Last edited:
CPC their other outlet only needs a £5 order for UPS delivery included. Sometimes CPC is much cheaper than the parent Farnell. For example. 400V caps stand a chance of beating expensive Audiofool caps. The recipe for sound is big is better. What avoids high voltage brakedown also is the recipe for best sound or optimium high frequency performance into the radio frequencies.
http://cpc.farnell.com/on-semicondu...5&ddkey=http:en-CPC/CPC_United_Kingdom/search
http://cpc.farnell.com/magnatec/2n3...h&ddkey=http:en-CPC/CPC_United_Kingdom/search
http://cpc.farnell.com/vishay/bfc246828225/468-polyester-cap-400v-2-2uf/dp/CA05400?st=2.2 uf
http://cpc.farnell.com/panasonic-el...or-220uf-63v/dp/CA05189?st=Panasonic fc 220uF
Last edited:
My brother thought if all the time constants in the amplifer were similar a worse sound would be had. Even when no output capacitor this rule should be respected. Do it by ear and never assume bigger is better. Naim Audio were told by a reviewer the Naim Nait was wrongly designed. So they built him one to the " right " recipe. It sounded awful. Mostly it sounded slow when the " right " version. It could be Urban Myth, I think it's true as when told to me there was no hesitation in the story. I was told by a Naim staff member.
If one looks at the JLH distortion graphs it could be gain setting cap. Douglas Self almost likes it to be down to DC. The bootstrap also could be tuned by ear. Try one much smaller to the bootstrap and one much larger. I used 2200 uF to the bootstrap from a circuit I assumed to be right. It was fine. 10 uF to the bootstrap might do 180 Hz. As the amplifer has good loop gain when LF it might be better than you might think. That could be a film type. If anyone likes to do a sim of that it would be interesting. I have a hunch 0.1% distortion is still possible 10 to 500 Hz. Assume a 1000 uF feedback cap.
In a power amp, personally I wouldn't mess with a bootstrap cap; imho it would be best set at least a decade below the low freq. of interest; otherwise it would introduce phase distortion at LF and also there would result amplitude distortions in upper-half waveform because the (in JLH design) the upper output device would starve.
This would in consequence alter the mean (integrated) value of the output vaweform, and the NFB being DC and AC coupled; this would in consequence upset the amp's voltage bias - the NFB time constant would then also dictate how amp's bias points would "dance around" to the music.
In a power amp, imho the bootstrap should as closely resemble an (idealized?) current source to avoid this (unless intentionally).
Otoh, I used a "tailored" bootstrap in a small 3-transistor signal instrument preamp, to introduce some "color" and a "saggy-bouncy" vibe via (ab)using the above described mechanism of uppsetting the voltage bias point and output drive.
This would in consequence alter the mean (integrated) value of the output vaweform, and the NFB being DC and AC coupled; this would in consequence upset the amp's voltage bias - the NFB time constant would then also dictate how amp's bias points would "dance around" to the music.
In a power amp, imho the bootstrap should as closely resemble an (idealized?) current source to avoid this (unless intentionally).
Otoh, I used a "tailored" bootstrap in a small 3-transistor signal instrument preamp, to introduce some "color" and a "saggy-bouncy" vibe via (ab)using the above described mechanism of uppsetting the voltage bias point and output drive.
Playing with bootstrap caps won't be a problem. I have done it often enough. Remember the positive feedback is too low to be a problem and high enough to be useful. I never had a problem. Smaller can work. Douglas Self did a very similar project on uniform feedback with frequency as he imagined it might be something that someone would claim better. This was his Blameless design. Lets for arguement say the amplifier could give between 110 dB and 80 dB of correction, the 110 dB being at 5 Hz. It made me think this would be universal. I have never measured the JLH but assume the open loop gain to be about 60 dB at 1 kHz. I would guess it could be 75 dB at 5 Hz. If we say 180 Hz - 3 dB it might be -8 dB at 90 Hz and -15 db at 40 Hz. This would look much like Self's idea with much greater reason to to think it could be useful. As Self says the deliberate reduction of loop gain is bad engineering. However it could be virtuous here.
It is said humans can not hear phase distortion. However it might cause listening fatigue as our brains can. The proof is a squarewave played over speakers with good phase ability have no difference in sound when phase shifted. David Mate of SSL said his own work said it was fatiguing to process the sound. My own way of doing this is saying what is safe. Next I listen with an open mind. One has to say that standard speakers most own have considerable phase distortion. My own are not bad on this type of distortion. If you notice mine cross at 200 Hz and 8.3 kHz. I even get reasonable dispersion which should be impossible. A pure fluke of a cheap tweeter that seems ideal. The Qts being 1.2 and 0.49 helps. As said, the output caps do mostly what an active crossover might. The centre unit has no filtering except the 20Hz. The Q being enough to get the last drop of bass power without overshoot.
I have built a similar to JLH design that needed to be 1uF bootstrap before the distrotion was seen to increase. My LF loop gain must have been greater than 75 dB at 40 Hz. If one thinks it through an amplifer mostly needs bootstrap help 1 kHz to 10 kHz.
Self also did an input bootstrap. I nearly died of fright when I tried it. It worked. To keep noise down it's a miracle. I would have too many doubts to use it, when class AB crossover distortion junk to the input. It's just feedback loop upper leg ( LTP base 2 ) to input biasing split resistors ( 1K + 1K ) . However, many preamps used it. Two transistors can look impressive when you do. Lenco Phono stage if I remember did.
It is said humans can not hear phase distortion. However it might cause listening fatigue as our brains can. The proof is a squarewave played over speakers with good phase ability have no difference in sound when phase shifted. David Mate of SSL said his own work said it was fatiguing to process the sound. My own way of doing this is saying what is safe. Next I listen with an open mind. One has to say that standard speakers most own have considerable phase distortion. My own are not bad on this type of distortion. If you notice mine cross at 200 Hz and 8.3 kHz. I even get reasonable dispersion which should be impossible. A pure fluke of a cheap tweeter that seems ideal. The Qts being 1.2 and 0.49 helps. As said, the output caps do mostly what an active crossover might. The centre unit has no filtering except the 20Hz. The Q being enough to get the last drop of bass power without overshoot.
I have built a similar to JLH design that needed to be 1uF bootstrap before the distrotion was seen to increase. My LF loop gain must have been greater than 75 dB at 40 Hz. If one thinks it through an amplifer mostly needs bootstrap help 1 kHz to 10 kHz.
Self also did an input bootstrap. I nearly died of fright when I tried it. It worked. To keep noise down it's a miracle. I would have too many doubts to use it, when class AB crossover distortion junk to the input. It's just feedback loop upper leg ( LTP base 2 ) to input biasing split resistors ( 1K + 1K ) . However, many preamps used it. Two transistors can look impressive when you do. Lenco Phono stage if I remember did.
Hi all,
Is there anybody that knows the name of these green connectors as shown in the eBay pic refered by Stevec67 ? I am interested too. Thank you.
Best regards
rephil
hello Rephil,
those are common on all chinese PCBs.
something like this
10 pcs KF7.62-2P 2pin 7.62mm pitch Screw Terminal Block Good | eBay
here is the drawing
KF7.62-2-Pin PCB Terminal Block
regards
Prasi
Each time an LF time constant is introduced, the response to a step input is to give a damped oscillation which crosses the zero axis n-1 times. So for an output cap, input cap and feedback decoupling cap, that makes the zero crossing 2x for a step response. The amplitude of this "bounce" depends on the relative ratios of the time constants. This would almost certainly be worst for equal time constants.
This is one reason that the time constants need to be chosen carefully. The response to a (say 20Hz) LF sinewave with 2000mF is somewhat attenuated, perhaps to only one quarter power, for the leading sinusoid ("first cycle"). This makes the bass sound weak - bass signals are OK, but bass transients are "thin".
Therefore, I agree too that something more like 22mF is needed. That gives a better bass response, but then makes the choice of the other time constants more critical.
This is one reason that the time constants need to be chosen carefully. The response to a (say 20Hz) LF sinewave with 2000mF is somewhat attenuated, perhaps to only one quarter power, for the leading sinusoid ("first cycle"). This makes the bass sound weak - bass signals are OK, but bass transients are "thin".
Therefore, I agree too that something more like 22mF is needed. That gives a better bass response, but then makes the choice of the other time constants more critical.
Translation: kicks don't play back like they are/were recorded. Which may be cool if you want to "vibe" your amp to your liking, but it's a no-go if you're editing audio.
Your and mr. Pearsons' observations are similar to mine.
Was boring;-)
Have build in a REAL EMITTER, 2SC1381 (or so) in the Emitter-casing. A friends amp. May be, the only emitter in the "Emitter" - terrible amp of "S & R" - on earth;-)
Sounds great. Wide, open, clean, relaxed. And so on;-)
Not finished yet.
Have build in a REAL EMITTER, 2SC1381 (or so) in the Emitter-casing. A friends amp. May be, the only emitter in the "Emitter" - terrible amp of "S & R" - on earth;-)
Sounds great. Wide, open, clean, relaxed. And so on;-)
Not finished yet.
Attachments
![IMG_20180221_212753[1].jpg](/community/data/attachments/602/602878-9ff0ffd891cf9ff52ca3d9abe4828725.jpg)
![IMG_20180221_212659[1].jpg](/community/data/attachments/602/602897-5d3412e9f94bb2bd59ff20aef83dd914.jpg)
I think this time constants discussion very positive. The big companies like Chef's keep it secret. Not really secret as none are asking. I think to put the corner frequency very low when possible is a good idea. Then use the input cap to set the subsonic level ( by ear ). If you use 4700uF 100V at the output it should be a good compromise, 100V types are slightly better and avoid silly money being paid for Audiophile similar types. My bass speaker is 5R7 so could be larger. Tim de Paravacini though 5 Hz -3dB is the highest frequency one should have. Where practical yes.
If you use 4700uF you could put a Panasonic FC 220 uF in paralell and 2u2 400V polyester. I am not totally convinced by this so do listen. If your tweeter needs a similar cap have it's own that avoids the 4700 uF. That's the smart way to do it.
If you use 4700uF you could put a Panasonic FC 220 uF in paralell and 2u2 400V polyester. I am not totally convinced by this so do listen. If your tweeter needs a similar cap have it's own that avoids the 4700 uF. That's the smart way to do it.
Thanks Nigel. I think I can follow your reasoning. It's a bit like speaking Italian to a Romanian at work, my Italian is p-poor but his is great, better than his English, so we just about get by but my Italian, like my O level Electronics and A level Physics from the 80's here, is hanging by a thread. For now I will build it as per schema and get it working. This may in itself be a challenge. I will then think about having a play with component values.
A level French is "un passeport a nowhere" to (mis)quote Miles Kington. Because you study literature for A level your language skills advance but minimally. You then have a LONG way to go to get to a good conversational standard if you go on to study Mod Lang. A friend has a daughter, aged 18 and about to do an A level in French prior to a Mod Lang degree, I tried her French out against mine (I lived and worked in France for 3 years) and it's very weak. I only have O level French but like you, I'm the talker. It doesn't matter where I go.
Back to amplifiers. I haven't studied electronics for over 30 years. I did once understand what a time constant was, bear with me while I catch up. Thank God for the internet.
Edit - light dawns! It's the k part of V=Vo*e^kt, isn't it? I'm a (bio)chemist, we use N=No*e^kt to describe the kinetics of first order chemical reaction kinetics because, guess what, the same set of physical laws apply. In chemistry "k" is the rate constant, iirc. Light dawns indeed. I will go away and have a lie down before I think about the implication of this to an ac situation.
Back to amplifiers. I haven't studied electronics for over 30 years. I did once understand what a time constant was, bear with me while I catch up. Thank God for the internet.
Edit - light dawns! It's the k part of V=Vo*e^kt, isn't it? I'm a (bio)chemist, we use N=No*e^kt to describe the kinetics of first order chemical reaction kinetics because, guess what, the same set of physical laws apply. In chemistry "k" is the rate constant, iirc. Light dawns indeed. I will go away and have a lie down before I think about the implication of this to an ac situation.
Last edited:
5Hz -3dB will introduce low-end phase smear upto 50Hz-something.
But being overly optimistic at setting huge time constants (so low corner f) can and will upset the DC operating points in the amp. I intuitively understood some simplistic rookie aspects of this way back in 80s when was in e.e. school (and built simple amps to blast punkrock since my tween years, ha) but can't imagine one being able to find the optimum simply "by ear".
Setting the amps' low corner f with the input cap could possibly be done this way, but setting other important constants shoud be imho monitored first by a sim and then with a scope (monitoring the V and I across the constant-setting cap).
The NFB cap will average/integrate the "music" mean value (the DC across this cap will "dance" with music), and this caps' time constant must be imho be set with that in mind. So if you want to "virtually eliminate" this behavior, set it super-low. But then, the whole amp's settling time will be super slow ... and so on and on. Neverending story.
But being overly optimistic at setting huge time constants (so low corner f) can and will upset the DC operating points in the amp. I intuitively understood some simplistic rookie aspects of this way back in 80s when was in e.e. school (and built simple amps to blast punkrock since my tween years, ha) but can't imagine one being able to find the optimum simply "by ear".
Setting the amps' low corner f with the input cap could possibly be done this way, but setting other important constants shoud be imho monitored first by a sim and then with a scope (monitoring the V and I across the constant-setting cap).
The NFB cap will average/integrate the "music" mean value (the DC across this cap will "dance" with music), and this caps' time constant must be imho be set with that in mind. So if you want to "virtually eliminate" this behavior, set it super-low. But then, the whole amp's settling time will be super slow ... and so on and on. Neverending story.