There has been quite a lot of discussion about the "XY" LM3886 amplifier kit that is available from multiple Ebay sellers. The XY kit is quite inexpensive but the layout is "sub-optimal" and may be supplied with fake parts.
I happened to be poking around Ebay today and I happened to see what looks like a much better implementation of a mono LM3886 amplifier. This uses two LM3886 in parallel and includes perks like a Zobel output network, the cap across the LM3886 inputs that suppresses ringing, trimpots for balancing out the DC offset, and so on. The circuit schematic can be found here:
http://i.ebayimg.com/images/g/a0EAAOSwGotWmMlg/s-l1600.jpg
The only thing that looks a bit questionable is the input RC filter, and the overall input impedance as 47k||10k||10k which is getting a bit low (around 4k). I think the 47k to ground at the input should be removed and the input cap sized against 10k||10k=5k for the desired corner frequency. My guess is that the designer started with the PA100 circuit but then switched to inverting configuration for the LM3886. I can't comment on the layout because it's mostly not shown.
Of course twice the LM3886 chips and a larger and more complex layout don't come for free. These boards cost about US$26 each plus shipping, but you are getting a step up from the XY kit and these are pre-assembled (hopefully without fake parts!). The boards could drive 2R loads comfortably if they had a bit more on board capacitance IMO.
Anyway, here is a link to a current auction if anyone is interested:
Assembled LM3886 mounted in parallel 100W mono DC audio amplifier board
Comments welcome.
I happened to be poking around Ebay today and I happened to see what looks like a much better implementation of a mono LM3886 amplifier. This uses two LM3886 in parallel and includes perks like a Zobel output network, the cap across the LM3886 inputs that suppresses ringing, trimpots for balancing out the DC offset, and so on. The circuit schematic can be found here:
http://i.ebayimg.com/images/g/a0EAAOSwGotWmMlg/s-l1600.jpg
The only thing that looks a bit questionable is the input RC filter, and the overall input impedance as 47k||10k||10k which is getting a bit low (around 4k). I think the 47k to ground at the input should be removed and the input cap sized against 10k||10k=5k for the desired corner frequency. My guess is that the designer started with the PA100 circuit but then switched to inverting configuration for the LM3886. I can't comment on the layout because it's mostly not shown.
Of course twice the LM3886 chips and a larger and more complex layout don't come for free. These boards cost about US$26 each plus shipping, but you are getting a step up from the XY kit and these are pre-assembled (hopefully without fake parts!). The boards could drive 2R loads comfortably if they had a bit more on board capacitance IMO.
Anyway, here is a link to a current auction if anyone is interested:
Assembled LM3886 mounted in parallel 100W mono DC audio amplifier board
Comments welcome.
Last edited:
Using the input bias current into a 100 kΩ pot to adjust for the input offset voltage differences is really, really bad circuit design. Good idea. Horrible execution.
They goofed on the values of the resistors in the Zobel and Thiele networks.
I'm not a fan of leaving input caps with one end floating as shown in the schematic.
Interesting with the inverting design though.
If you'd like to put two LM3886es in parallel, a better starting point is the BPA200 app note (AN-1192) from TI: http://www.ti.com/lit/an/snaa021b/snaa021b.pdf
Tom
They goofed on the values of the resistors in the Zobel and Thiele networks.
I'm not a fan of leaving input caps with one end floating as shown in the schematic.
Interesting with the inverting design though.
If you'd like to put two LM3886es in parallel, a better starting point is the BPA200 app note (AN-1192) from TI: http://www.ti.com/lit/an/snaa021b/snaa021b.pdf
Tom
You're right about the input impedance and that the 47k is not required where shown.
Moving that 47k to the other side of the input cap and changing it to ~1M0 would work.
I think each chipamp needs it's own Zobel located before the 0r5 output resistor.
I would like to hear Tomchr's ideas on a properly executed offset balancing circuit.
Moving that 47k to the other side of the input cap and changing it to ~1M0 would work.
I think each chipamp needs it's own Zobel located before the 0r5 output resistor.
I would like to hear Tomchr's ideas on a properly executed offset balancing circuit.
When Ti announced the removal of the 4780 dual chip, I thought about whether it was worth trying to design a 3886 parallel amp "done right".
I would like to see some of Tomchr's expertise applied to the PA03 style PCB but with two 3886.
http://www.diyaudio.com/forums/chip-amps/119752-pa03-pavel-dudek-gainclone.html#post4541364
http://www.diyaudio.com/forums/chip-amps/119752-pa03-pavel-dudek-gainclone.html#post2537817
I would like to see some of Tomchr's expertise applied to the PA03 style PCB but with two 3886.
http://www.diyaudio.com/forums/chip-amps/119752-pa03-pavel-dudek-gainclone.html#post4541364
http://www.diyaudio.com/forums/chip-amps/119752-pa03-pavel-dudek-gainclone.html#post2537817
Last edited:
Great idea.
Back a few years there was this 3x parallel (used in a bridged-parallel config using two boards) build by alexw88 that I always was interested in:
http://www.shine7.com/audio/pa150.htm
You will see some similarities between the design (schematic) of the PA150 and the Ebay 2x parallel board in the 1st post of this thread:
Tom - since you have extensive experience with the LM3886 would you consider extending your existing design to a 3x parallel board? This brings up the power and output current capability to a more serious level. If you can keep the same ballpark THD specs in a 3x LM3886 parallel implementation you would really have something there. Perhaps the design could allow 1x, 2x or 3x version to be built from the same board by varying the level of component population. Also, I would love to have a way to implement muting using an off-board relay or the like. It's food for thought at least.
The shine7 PA150 layout gerber files are found here:
http://www.diyaudio.com/forums/chip-amps/65131-bpa300-mono-block-finished-measured-3.html#post787803
It uses smd components.
Two were used in the BPA300 project. Some measurements are found here:
http://www.shine7.com/audio/bpa300_3.htm
Last edited:
Getting three LM3886es to play nicely together is a bit of a challenge. Getting them to play nicely in a composite amplifier is really quite difficult. If you really need 20+ ampere of output current and good performance, I think you'll be better off with a discrete output stage driven by an opamp.
Tom
Tom
Well I decided to give the PA150 from alexw88 a try so I bought a couple of PCBs. I plan to make a couple of mods when I build it:
Eliminate the 47k to ground on input and the input cap (see below)
Reduce input resistors to 1k or less (see below) to keep noise low
reduce feedback resistors and bypass cap accordingly to give desired gain
break out the mute connection so that I can mute the output stages remotely
I have not decided what to do about the trim pots. Since the input impedance is much lower, the trimpot value can also be reduced significantly and that should keep noise down.
If I reduce the input impedance for each amp to 1k and there are three in parallel that's only 333R input impedance. That's too low to be driven directly, however, I can implement a "preamp" circuit to drive the load. I developed and built a headphone amp that uses 16 NE5532s (as followers) in parallel per channel as output buffers, fed by an LME49720, with onboard PS. I measured very low distortion into 32R, so I could even reduce the input impedance 10 times LOWER and use that to drive the board. Not sure I really need to do that, but these would work nicely together. Alternately I could simply drive the amp using a DAC that can drive a couple hundred Ohms comfortably. If the DAC has no DC output I can eliminate the input cap.
I have an upcoming active speaker project for which I need two channels of 80-100W power into 4R and I might use these boards with rail voltages around +/-35V to +/-40V using some Mouser-sourced LM3886s that I bought previously. I may only populate the board with two LM3886s, not sure.
Eliminate the 47k to ground on input and the input cap (see below)
Reduce input resistors to 1k or less (see below) to keep noise low
reduce feedback resistors and bypass cap accordingly to give desired gain
break out the mute connection so that I can mute the output stages remotely
I have not decided what to do about the trim pots. Since the input impedance is much lower, the trimpot value can also be reduced significantly and that should keep noise down.
If I reduce the input impedance for each amp to 1k and there are three in parallel that's only 333R input impedance. That's too low to be driven directly, however, I can implement a "preamp" circuit to drive the load. I developed and built a headphone amp that uses 16 NE5532s (as followers) in parallel per channel as output buffers, fed by an LME49720, with onboard PS. I measured very low distortion into 32R, so I could even reduce the input impedance 10 times LOWER and use that to drive the board. Not sure I really need to do that, but these would work nicely together. Alternately I could simply drive the amp using a DAC that can drive a couple hundred Ohms comfortably. If the DAC has no DC output I can eliminate the input cap.
I have an upcoming active speaker project for which I need two channels of 80-100W power into 4R and I might use these boards with rail voltages around +/-35V to +/-40V using some Mouser-sourced LM3886s that I bought previously. I may only populate the board with two LM3886s, not sure.
You may want to consider the LME49600 buffer if you need to drive a low input impedance. Even at 50 Ω load, I can't measure the THD+N of the LME49600 (when used in a properly designed composite amp with an LME49710 or OPA1611). At 32 Ω load, my Audio Precision APx525 can just barely measure the THD+N above 10 kHz. Below 10 kHz, the THD+N falls below the noise floor.
You could also make the amp non-inverting and not have the issues with low input impedance. Do make sure that you read the bits in the data sheet about Cc, Rf2, and Cf. These three components are needed for stability near the rail. Consider adding the Zobel and Thiele networks as well.
If you haven't read it already, I suggest reading my Taming the LM3886 series: Taming the LM3886 Chip Amplifier.
Good luck!
Tom
You could also make the amp non-inverting and not have the issues with low input impedance. Do make sure that you read the bits in the data sheet about Cc, Rf2, and Cf. These three components are needed for stability near the rail. Consider adding the Zobel and Thiele networks as well.
If you haven't read it already, I suggest reading my Taming the LM3886 series: Taming the LM3886 Chip Amplifier.
Good luck!
Tom
Last edited:
You want to inject a small DC voltage into the amp in a low-noise way. This is commonly done with a reasonably low value pot (say 1 kΩ) powered from a reference voltage of some sort. Typically, this reference voltage is just a well bypassed voltage divider off the supply rail. Franco goes through a couple of different schemes in Ch. 5 of his op-amp book. You've previously grumbled a bit that textbooks are expensive and I absolutely agree with you. Thankfully, the 2nd edition is every bit as good as the 3rd edition and available for less on the used market.
The challenge with the DC adjust circuits is that they tend to be noisy. The 100 kΩ pot in OP's schematic is a prime example. In some cases they also compromise the PSRR of the amp (such as the voltage divider off the rail I just mentioned).
Rather than relying on adjustment to drive the DC offset down, I suggest using a DC servo. Using a modern low-offset, low input current op-amp, such as the OPA277, it is possible to obtain offset voltages in the 2-digit uV range. In fact I have one channel of Modulus-86 running on my lab bench right now with an offset in the single digit uV range.
The challenge with DC servos is that you need to push the cutoff frequency down to avoid the servo degrading the THD within the audio band. With the LM3886 by itself, you won't need to worry too much, but I do suggest keeping an eye on it. You need a filter topology (a DC servo is a lowpass filter) with good stop band attenuation. Multi-feedback topologies are handy here. Jim Karki's (TI) app note on the topic is useful: http://www.ti.com/lit/an/sloa049b/sloa049b.pdf
For whatever it's worth, National used DC servos to minimize the offsets in the BPA200 design.
Tom
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.