Buffer/Gain stage suggestions for LM3886

[zakman35]

By the way what happened to Mr.zakman35? Did he completed his project? I would like to see some great pics of his live working project. LM3886 has also to come my way, sooner or later, i am sure.

Hi noddy55, i am here doing my math as you did in order to calculate proper RC values according to Andrew's suggestions.

Anyway i start with a "simple" question about a buffer stage and i end up re-building my 8 channel amp from scratch . But i guess this is what DIY is all about and i enjoy "almost" every moment of it.

I decide to rebuild only 4 channels at the moment but i need to make some rearrangements inside the amps case in order to reduce cabling.

I decide to go for a point to point construction instead of pcb, but i still need to decide about filters - gain - input sensitivity, but mostly about the values - size of the caps.

Of course i will post my schematic for your comments before i build it.

[zakman35]

Ok, here is what i come up with based on your suggestions.
Sorry for my bad drawing.

Can you suggest if this needs further changes - adjustments?

[danielwritesbac]

RM may have insufficient current, suggest smaller value--see the kits for a possible reference point.
RI versus CI mismatch slightly because of assuming perfect cap, suggest bigger values for RI and RF--such as 2K2, 62k, which won't need a preamp.
22k input load may lack slightly, suggest 10k~12k input load to reduce dc pollution of the CI area.
There are performance aspects more important than having a non-inverting amplifier's feedback resistor same value as its input load.
Perhaps you need to test some options with one amp before soldering all four.

Aribitray bypass cap value applied to "1000u Panasonic" and instead suggest to Temporarily use a sample 1000u as either an NFB cap or an output cap while you select an attractive bypass cap for it--ideal bypass cap value is not always 100n. So, "eavesdrop" on that and find a good match but try to select the value within a plausible range. It doesn't have to be perfect--main point is to avoid causing peakish.

2.7R with 100n output RC infers that the 100n is a polyester dip cap with an ESR of approximately 4 ohms (the typical part selected for mass production amplifiers); however, if instead you use an efficient cap, then increase the resistor value to 5.6R in order to maintain approximately correct resistance value (2.7R is off the mark if efficient cap is with it).

Stability assistance part is missing--that is Part CC from the datasheet (range 220p to 330p). Any treble lack from installing that can be recovered by selecting attractive bypass caps ("Plan A" simple method) to add at CIN and also at CI, and since you want to do that anyway, suggest installing part CC prior to performing fine tuning. Do please consider the "Part CC" capacitor (lm3886 datasheet) to assure stability prior to doing any sort of fine tuning work.

Ideal topologies for LM3886 support include nesting, composite and current pump, like the MyRef-FE and those should be considered before investing a lot of time with LM3886.

If you're not using one of those designs, but you are doing a non-inverting LM3886, I suggest 2 more parts.
*From Allied Electronics, 1 of 250v 4.7uF Mallory SEK capacitor to use from V+ to V- at the chip, for cancelling audible harmonic noise. That's 1 of those caps per each chip. This filter can make your audio fine tuning easier and perhaps the amp may run a bit cooler.
*A buffer to help source devices input a more level frequency response--although a stiffer input load at the chip (my previous suggestion of 10k to 12k range) will help slightly, a buffer can help greatly, depending on the source used.

[zakman35]

RI versus CI mismatch slightly because of assuming perfect cap, suggest bigger values for RI and RF--such as 2K2, 62k, which won't need a preamp.

RF = 62K? I thought the maximum value for RF was close to 33K.

22k input load may lack slightly, suggest 10k~12k input load to reduce dc pollution of the CI area.

Yes, but this will decrease the input sensitivity to 1.4V, miniDSP outputs only 0.9 V.

Stability assistance part is missing--that is Part CC from the datasheet (range 220p to 330p)

I have include CC 220pf, but instead of the usual position between +In and –In I have place it between +In and signal ground. This decision made after I read this thread.

to add at CIN and also at CI, and since you want to do that anyway, suggest installing part CC prior to performing fine tuning

Are you suggesting to omit Cin & Ci?

Ideal topologies for LM3886 support include nesting, composite and current pump, like the MyRef-FE and those should be considered before investing a lot of time with LM3886.

MyRef-FE, hmmmm, I have four chips left!!!

A buffer to help source devices input a more level frequency response--although a stiffer input load at the chip (my previous suggestion of 10k to 12k range) will help slightly, a buffer can help greatly, depending on the source used.

I think i will not avoid the buffer after all.

[danielwritesbac]

Originally Posted by zakman35

RF = 62K? I thought the maximum value for RF was close to 33K.

Divide the feedback resistor value and the feedback-shunt resistor value and then add 1, to check your gain settings. The feedback resistor versus the feedback-shunt resistor form a voltage divider, really just an ordinary volume control. If want lower resistor values then. . .
For feedback of 33k the feedback shunt resistor is 1.2k (gain is 28.5x) and the NFB cap is in the range of 330u (warm) to 470u (neutral). Yes, you'll need a bigger cap. Those are more difficult to select for seemly results, but very much doable.
Both lower resistor values and higher resistor values represent trade-offs, but they're different.

Originally Posted by zakman35

Yes, but this will decrease the input sensitivity to 1.4V, miniDSP outputs only 0.9 V.

Then decrease the 680R proportionately. For example, 10k with 330R at input. Loss or lack of a load in this area dramatically reduces the performance of the amplifier's LTP type input. That chip is not an integrated amp--it is a power amp. The actual sensitivity is the amplifier gain setting, and if that is looking too high (too much influence and noise) then consider a preamp. I would consider the MooseFet project.

Originally Posted by zakman35

I have include CC 220pf, but instead of the usual position between +In and –In I have place it between +In and signal ground. This decision made after I read this thread.

The need of these options will be different per each different layout.

Originally Posted by zakman35

Are you suggesting to omit Cin & Ci?

I wouldn't omit those caps. Even though there's a selection workout, there's grand performance benefits to be had, especially in dynamics. And, always zero offset is helpful for speaker safety.

Originally Posted by zakman35

MyRef-FE, hmmmm, I have four chips left!!!

Have fun!

Originally Posted by zakman35

I think i will not avoid the buffer after all.

Depending on the source, the frequency response can be helped with a buffer. My computer is a good example of that. Even my 6n3p experiment (no match for a diamond buffer) still managed to improve the computer's output tremendously, and that was a surprise. How much? Well, throw the EQ away--that much. Yes, surprising indeed.

[zakman35]

I would like to thank you all for helping me understand and build this amp.

I build 4 channels point to point construction.
The measure: DC offset 1,4 - 1,7 mV, Ac noise = 0.1mV.

Listening impressions: the amps are more detailed and noise free.
I can hardly listen to a tiny hiss with my ear stuck on the tweeter and preamp at full volume.

Here is the final schematic i use.

[danielwritesbac]

I wonder if you might like up to triple the values of both RI and RF. When done proportionately, it doesn't change the gain. But, it does change the feedback current and that will give you a different sound that you may (or may not) like better.

H1/H2 bass tone:
For one thing, adjusting RI up to the neighborhood of 2K~2.7K will match up with that 220uF at CI for neutral tone setting or 1k~1.5K with 220uF for warm bass setting.
Likewise adjusting RI up to approximately 1K~1.5K will match up with a CI value of 470uF for neutral tone setting or 680R with 470uF for warm bass setting.
However, the schematic's 680R with 220uF is "extra extra warm" bass setting that some people might think is either boomy or "old fashioned" sort of bass rendering, and yours won't be boomy thanks to Cin set at 1u.

2 ways to make CI "larger" (pass more/lower bass):
Larger values for CI (a bigger cap or a paralleled cap), or. . .
Larger values for RI (increase RF resistor by same proportion to keep gain)

H1/H2 balance is also reliant on Cin, input cap, size, and I believe that 0.68u to 3.3u values are valid choices--the schematic's 1u looks valid to me. At the input cap, smaller is cleaner; however, if Cin is made proportionately larger, likewise CI may also be made larger resulting in more low notes and with the same tone balance as before. But if Cin is made larger without making CI larger then the result is warmer/boomier bass. Conversely, if Cin is not changed but CI made larger, then the result is colder/deeper bass.

Currently, the schematic has CI set for warm warm and Cin set for clean. Given that LM3886's Spike system can cause "hot mids" sound, your setting for extra warm, yet clean, bass may be a practically perfect tonal balance compensation. Is that a "bari boost" setting? I do believe that your schematic shows a valid option for LM3886 and other overture chips. I've listed other bass options in case you want to try them.

[Robert Kesh]

I think Cin says 10u, though it's hard to tell with that hand writing.

[zakman35]

Cin= 10uf,

Thanks for the us-full infos, i do have some modules left so maybe i will try a version with increased Ci and one increased RI and RF.

The bass settin and the effects of it varies according to the used speakers and the room of caurse, for example with my Audio Spectrum Athena speakers (8'' Seas woofer, 1'' seas tweeter, F3= 33Hz) with active cross and bi-amplification the bass is definitely on the warm side but very pleasant and not boomie.

Today i connect (with passive crossovers no bi-amplification) my Focal's (Electra 906 with 6.5'' midwoofer, 1'' tweeter, F3 = 50Hz) which i rarely use because their lack of lower bass and i notice a significant increase in the lower extension (5 - 10 Hz i guess) with very natural bass tones, no warmness here.
The Focal's now sound full bodied and very natural and they certainly don't lack bass any more.

So my question is, can the adjustment of Ci & Ri effect beside the bass tone & the bass extension as well?

[bobodioulasso]

According to Andrew and Daniel long posts/explanations Cin should be 1uF.
You could just serial connect an 1 uF at your input ( keeping the 10 uF in place) and tell us if you perceive a difference in sound.
I am curious, not you?
Or at least to thank them.