The first step is to use all components that are shown in the datasheet, including those marked as optional. Add Ci, you don't need 1000 µF there.
Then you need a decent power supply, i.e. the cascaded 1000 µF, 10 µF and 100 nF capacitors on board as specified in the datasheet plus bigger caps (4700 µF or bigger) off board. Forget about subberised power supplies. Use 100 nF capacitors directly on the rectifier diodes, then add as much capacitance as you can afford. Don't waste your money on a bleeder resistor.
Depending on your speakers it might be a good idea to complete the Tiele network, i.e. after the RC circuit add a 10 Ohm resistor with parallel 0,7 µH coil in series to the speaker.
As soon as you install all those components you will probably have to use a different PCB. That of itself may also help to get rid of oscillation, if you avoid the mistakes of the original. E.g. input and output traces shoud not run in parallel. If you can route the input at a right angle to the output and both at a right angle to the power supply traces, you are on the right track.
The 2,7 Ohm resistor is a good inidcator. You should be able to use a standard 0,25 W resistor there. As long as that blows, you can be sure that the amp oscillates which has a bad effect on the sound. Placing a 2 W resistor in that place is curing the symptoms instead of the illness.
No. Daniel referes to ESR, but that is unimportant in this case. A 100 nF capacitor will always have an ESR that is insignificant in relation to the 2,7 Ohm resistor that it has in series. The corner frequency of that RC circuit is nearly 600 kHz which puts it far above the audio range.
I concur with most of PacificBlue's information in post #81.
Yes, by all means add 1000uF (or 1500uF) to those amplifier boards. They already have the 100nF.
Yes, by all means "complete the Tiele network" at output.
Yes, try a 1/4 watt resistor as a test approach for speaker output zobel, because if the resistor becomes hot, turns black, goes open circuit, or blows out, that will tell you your amplifier is oscillating. However, please replace with a sturdy part before the amplifier is put into long term service. I still maintain that 2.7 ohms is an inappropriate load for LM3886.
And, yes, if you want to fit a lot more elegance onto your amplifier, then you'll probably want to replace the circuit boards.
My prior suggestions had to do with using Audiosector design on the chipamp.com boards because its answers the frequency response question and because it will fit the tiny little boards that you already own. Then you can use the information in the Audiosector support forum: Audio Sector - diyAudio <--link
Audiosector and chipamp.com kits had concurrent development and are one and the same design. However, the chipamp.com kit's frequency response and bandwidth are wack because of some component values and probably also oscillation. Neither have any technically correct design and that's fully expected of low component count gainclones.
Engineers hate skimpy design, as you can tell.
However, audiosector design does have a functional answer that fits onto the boards that you already own, like they were made for it, because indeed they were.
I don't like that answer any more than AndrewT or PacificBlue.
But I've absolutely no intention of making a brand new design for LM3886 or replacing your circuit boards.
Yes, by all means add 1000uF (or 1500uF) to those amplifier boards. They already have the 100nF.
Yes, by all means "complete the Tiele network" at output.
Yes, try a 1/4 watt resistor as a test approach for speaker output zobel, because if the resistor becomes hot, turns black, goes open circuit, or blows out, that will tell you your amplifier is oscillating. However, please replace with a sturdy part before the amplifier is put into long term service. I still maintain that 2.7 ohms is an inappropriate load for LM3886.
And, yes, if you want to fit a lot more elegance onto your amplifier, then you'll probably want to replace the circuit boards.
My prior suggestions had to do with using Audiosector design on the chipamp.com boards because its answers the frequency response question and because it will fit the tiny little boards that you already own. Then you can use the information in the Audiosector support forum: Audio Sector - diyAudio <--link
Audiosector and chipamp.com kits had concurrent development and are one and the same design. However, the chipamp.com kit's frequency response and bandwidth are wack because of some component values and probably also oscillation. Neither have any technically correct design and that's fully expected of low component count gainclones.
Engineers hate skimpy design, as you can tell.
However, audiosector design does have a functional answer that fits onto the boards that you already own, like they were made for it, because indeed they were.
I don't like that answer any more than AndrewT or PacificBlue.
But I've absolutely no intention of making a brand new design for LM3886 or replacing your circuit boards.
Last edited:
I'm looking at the datasheet:
http://www.national.com/ds/LM/LM3886.pdf
Compared with the chipamp.com manual on page 3:
http://www.chipamp.com/docs/lm3886-manual.pdf
There are 2 diagrams that I can see in the datasheet. The first is labeled "Typical Application." This is very close to the diagram of the chipamp kit, except that the parallel resistor R2 is not on the datasheet, nor is the Zobel/RC on the output. The datasheet includes RL, an 8R resistor to ground after the output that the kit does not include.
The second is labeled "Single Supply Application Circuit," which does not seem to apply in my case, correct? That one certainly is more complicated...
I also notice that the "Typical Application" does not include an input DC blocking cap or RF filter as suggested by Andrew.
Would you use the R2 resistor from the Chipamp.com diagram? Would you put a DC blocking cap between the pot and Rb in the datasheet diagram? Would you put an RF blocking cap somewhere beside across the RCA jacks? What does RL do? Could this missing be part of my problem?
I think I might try to wire this up point to point and see how that sounds.
Thanks for your good humor in answering my questions so far!
R2? Chipamp.com shows 12k input load (25k//22k). National Semiconductor's Typical Application shows 10k input load. Either works.
But, I would most certainly use National Semiconductor's example (as a 10k resistor for input load) because it will help prevent DC offset design mistakes.
Gainclone design is plagued by omission, and therefore Page 8 of National Semiconductor's datasheet should be a most interesting read if you'd like a more complete amplifier.
"Typical Application" shows a scrap of an amplifier along with a few assumptions and Page 8 helps to fill in some of the blanks.
RL? If that's missing, you'd consume the lowest amount of electricity possible, but you'd also have 0% electrical efficiency. RL is Resistive Load, and they assume that you will use a speaker.
Point-To-Point? Well, that's dandy, but since your tiny circuit boards aren't much different than point to point, I'd just use the boards. If you want some point to point wiring, then perhaps you'd enjoy creating a small signal ground above the board. I'd try sticking a 3.3 ohm resistor into signal ground terminal and using the 3.3 ohm resistor as the ground for the 680R, the input load and the signal cable (aka, a signal star ground). There's probably better examples, but it'll work. Look up star grounding for more (and probably better) information.
Would I put an RF blocking cap somewhere besides the RCA jack? Yes, most certainly! See post #80. The number 1 problem of a gainclone is oscillation, but the little cap also blocks the amplification of RF.
See the National Semiconductor datasheet on page 8.
Its part labeled Cc, a 220pF cap between pins 9 and 10.
That value is right. Generic name is "stabilizer cap" because of its function. The bare minimum value is 100pF. Omission is severely inappropriate for non-inverting LM3886 amplifiers, and omission of this part Cc may cause unlevel frequency response, excess heat, and reduced longevity, eventually including the burnout of both amplifier and speaker. Using Cc reduces that risk.
But, I would most certainly use National Semiconductor's example (as a 10k resistor for input load) because it will help prevent DC offset design mistakes.
Gainclone design is plagued by omission, and therefore Page 8 of National Semiconductor's datasheet should be a most interesting read if you'd like a more complete amplifier.
"Typical Application" shows a scrap of an amplifier along with a few assumptions and Page 8 helps to fill in some of the blanks.
RL? If that's missing, you'd consume the lowest amount of electricity possible, but you'd also have 0% electrical efficiency. RL is Resistive Load, and they assume that you will use a speaker.
Point-To-Point? Well, that's dandy, but since your tiny circuit boards aren't much different than point to point, I'd just use the boards. If you want some point to point wiring, then perhaps you'd enjoy creating a small signal ground above the board. I'd try sticking a 3.3 ohm resistor into signal ground terminal and using the 3.3 ohm resistor as the ground for the 680R, the input load and the signal cable (aka, a signal star ground). There's probably better examples, but it'll work. Look up star grounding for more (and probably better) information.
Would I put an RF blocking cap somewhere besides the RCA jack? Yes, most certainly! See post #80. The number 1 problem of a gainclone is oscillation, but the little cap also blocks the amplification of RF.
See the National Semiconductor datasheet on page 8.
Its part labeled Cc, a 220pF cap between pins 9 and 10.
That value is right. Generic name is "stabilizer cap" because of its function. The bare minimum value is 100pF. Omission is severely inappropriate for non-inverting LM3886 amplifiers, and omission of this part Cc may cause unlevel frequency response, excess heat, and reduced longevity, eventually including the burnout of both amplifier and speaker. Using Cc reduces that risk.
If you're interested in point to point, the build methods at Decibel Dungeon will give you plenty of room to work so that you're not limited to skimpy design.
Gainclone chip amp index page <--link
The insulator that they have set on/near the heatsink is great to attach parts and wires so they won't stray. They also cover star grounding, which is appropriate.
You might like some Gel Flux (radio shack has it) to assist in desoldering on those thick chipamp.com boards. They also have some sorts of sleeve type insulator tubes that may be helpful to keep wires out of trouble. If using them, I'd double-up because those tubes are thin. And, I'm pretty sure you can find a couple of 100pF or 220pF for stabilizer cap (pin 9 to pin 10).
Gainclone chip amp index page <--link
The insulator that they have set on/near the heatsink is great to attach parts and wires so they won't stray. They also cover star grounding, which is appropriate.
You might like some Gel Flux (radio shack has it) to assist in desoldering on those thick chipamp.com boards. They also have some sorts of sleeve type insulator tubes that may be helpful to keep wires out of trouble. If using them, I'd double-up because those tubes are thin. And, I'm pretty sure you can find a couple of 100pF or 220pF for stabilizer cap (pin 9 to pin 10).
Last edited:
this is the one that has not omitted the compulsory "optional" components.
Read through the table for what each component does and add it into the "typical". National should have a big label on there saying "advanced builders only
For each chip that gets damaged, there is potential to sell a replacement. Could this colour the advice given by the manufacturer?
For each stereo PB100 or PA100 adopted that's an extra two chipamps sold and one BPA300 sells 5 extra chipamps.
Last edited:
I'm intrigued by the idea of changing the power supply. Am I to understand that cascading capacitors means caps in parallel? Are resistors needed anywhere in the power supply after the diodes?
When you say use 4700uF caps or bigger off board, do you mean Cs in the amplifier circuit? When you say add as much capacitance as you can afford, what do you mean?
I think I see how to implement all the features of the datasheet including optional components, except for CA and RA, which see to have to do with DC bias. I haven't seen anyone include these on gainclone designs. Are these important for frequency response?
I am also not sure if the 10uF DC blocking cap is also for the dual PS design, and if so, whether it would go before or after Rb. The datasheet diagrams don't use the 22k resistor to ground (that the chipamp.com design uses) after Rb. Maybe the best thing is to do what Andrew suggested to me in the first place, which is to leave the 1k Rb in place, as well as the 22K to ground, and just put the coupling cap between the pot and Rb.
Thanks for any additional thoughts. I realize that I'm probably frustrating you by now, but I am appreciative of your help!
Last edited:
The entire datasheet is directed to advanced builders, that is why there is no such advice.
Yes, with the smallest caps next to the chip and the biggest next to where the wires form the power supply connect to the PCB.
No.
Yes.
Just that. This is one of the few cases where bigger is really better.
Those components are only necessary for single supply. You don't need them.
Yes.
Now I'm getting somewhere! Thank you.
I'm not sure if I understand what you mean by chip (I assume the 3886), and to which PCB you are referring.
So here's what I'm thinking: I can use the chipamp PCB from the kit. I can remove the 2.2k bleeder resistor, then I can put a 1000uF cap instead of the 10KuF nearest the diodes, a 10uF instead of the first 0.1uF, and leave the third 0.1uF in place, putting a wire in place of the 1R snubber resistor. Then I can maybe use the 10KuF caps for Cs at the amp boards.
The diagram from chipamp.com is on page 4:
http://www.chipamp.com/docs/lm3886-manual.pdf
So do I have it right, or backwards? I read that the problem with using large caps in the PSU is that they "take away some of the magic" in the mids and highs of the chipamp. In light of that, I don't understand why you are recommending large smoothing caps, which seem closely related to the caps on the PSU.
Thanks.
I'm not sure if I understand what you mean by chip (I assume the 3886), and to which PCB you are referring.
So here's what I'm thinking: I can use the chipamp PCB from the kit. I can remove the 2.2k bleeder resistor, then I can put a 1000uF cap instead of the 10KuF nearest the diodes, a 10uF instead of the first 0.1uF, and leave the third 0.1uF in place, putting a wire in place of the 1R snubber resistor. Then I can maybe use the 10KuF caps for Cs at the amp boards.
The diagram from chipamp.com is on page 4:
http://www.chipamp.com/docs/lm3886-manual.pdf
So do I have it right, or backwards? I read that the problem with using large caps in the PSU is that they "take away some of the magic" in the mids and highs of the chipamp. In light of that, I don't understand why you are recommending large smoothing caps, which seem closely related to the caps on the PSU.
Thanks.
That's the problem in a nutshell.
There are posts out there on the net that say ANYTHING YOU COULD POSSIBLY IMAGINE, and many things that you probably can't. Imagine.
This is more so in audio than in any other science-based subject.
You need to take note particularly when words like 'magic' are used.
Anybody giving advice in these circumstances faces a difficult problem, especially when dealing with a comparatively unsophisticated questioner. I don't mean you any insult when I say this. While there may be elements of truth in some of the wilder suggestions, it's better to steer the inexperienced away from them, so that they can gain some solid experience in sound practices.
AndrewT and pacificblue are both experienced constructors. If you let them steer you, you will produce a working amplifier, probably with a constitution not very far removed from the manufacturers recommendations, which will do justice to any material you choose.
Remember that, despite much of what you may read here, there is very little difference between one good amplifier and another, other than the absolute power output, and when differences are heard this is because the amplifier has fallen short of the very high standard which is achievable with little more effort than constructing a jigsaw puzzle.
w
Page 3 of AN1192 lets us know that we need to cool with clean power plus good stable design (and filters!!!) to stop extraneous non-audio workload, because no amount of heatsink can stop the generation of 40 watts--that's like shut barn door after horse got out. As usual, design choices that naturally run the coolest, may sound much better too.
Filter review
At the RCA jack, the RF filter (47pF suggested?) blocks RF (non-audio signals) so that it doesn't go through the amp.
(and now referring to the LM3886 datasheet)
Between pins 9 and 10, the stabilizer cap (Cc on the datasheet) not only helps prevent oscillation, but also prevents the amplification of non-audio signals.
Rsn, Csn block non-audio signals from speaker cables
L, and R drop the load of non-audio signals
The amplifier needs its workload confined to the audio band.
Filter review
At the RCA jack, the RF filter (47pF suggested?) blocks RF (non-audio signals) so that it doesn't go through the amp.
(and now referring to the LM3886 datasheet)
Between pins 9 and 10, the stabilizer cap (Cc on the datasheet) not only helps prevent oscillation, but also prevents the amplification of non-audio signals.
Rsn, Csn block non-audio signals from speaker cables
L, and R drop the load of non-audio signals
The amplifier needs its workload confined to the audio band.
When I said swap the pairs of 100uF power caps on the amplifier board for pairs of 470uF and that was ridiculed, I found this odd because I was quoting National Semiconductor's documented recommended power supply for LM3886, which is on page 14 of AN1192.
For your convenience, here is the link to the LM3886 chip manufacturer's document: http://www.national.com/an/AN/AN-1192.pdf
The big 10,000uF caps aren't stealin your treble.
For your convenience, here is the link to the LM3886 chip manufacturer's document: http://www.national.com/an/AN/AN-1192.pdf
The big 10,000uF caps aren't stealin your treble.
Last edited:
That's backwards. By cascading it is meant the largest capacitors are farthest from the LM3886, the smallest are closest. If you have 10mF on the power supply that's fine. If you have 1mF then 10uF then 0.1uF on the amp boards, that's pretty good. Don't change that setup.
Last edited:
Thank you for your answer. Right now I have 10mF then 0.1uF on the PS board, then Cs on the amp boards is 100uF // 0.1uF. What did Pacific mean about putting the most capacitance I can afford off board?
Are you suggesting that I get rid of the 100uF cap (Cs) on the amp board, and replace it with 1000uF // 10uF? Would it be the same thing putting 10,000uF then 1000uF on the power supply PCB, and then changing the 100uF Cs cap over to a smaller 10uF? I thought the goal was to INCREASE the size of Cs.
Are you suggesting that I get rid of the 100uF cap (Cs) on the amp board, and replace it with 1000uF // 10uF? Would it be the same thing putting 10,000uF then 1000uF on the power supply PCB, and then changing the 100uF Cs cap over to a smaller 10uF? I thought the goal was to INCREASE the size of Cs.
The most capacitance you can afford is in the power supply, your 10mF caps. Bigger is better, although there are diminishing returns. This has most effect on the bass, and less effect on the highs.
I think your power supply capacitance is fine and very likely not the source of your lack of highs.
Have you verified that all resistors are in the correct positions and that all polarized capacitors are the correct way around?
I think your power supply capacitance is fine and very likely not the source of your lack of highs.
Have you verified that all resistors are in the correct positions and that all polarized capacitors are the correct way around?
Ah, so having more capacitance doesn't lessen the highs? I read somewhere that the Gaincard had no capacitance at the power supply. I guess we all know you can't believe everything you read, though.
Hi,
there is much disagreement on that "fact".
I believe that cutting low on the capacitance impacts the overall sound quality more than what is gained in specific areas of the frequency and quality spectrum.
But equally some argue the exact opposite, that the overall sound quality benefits from using reduced capacitance more than can be made up for the gains in other areas when high capacitance is used.
You have to listen to each for your self and decide which gives overall satisfaction (if any).
I would suggest you re-check that the same conclusions are valid when a new amplifier is being developed. I'm am afraid to admit that I have missed this last stage due to my laziness or maybe lack of discerning hearing ability.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.