More than 10 years old : 🙂
https://www.diyaudio.com/community/threads/point-to-point-lm3886.187527/post-2580556
Patrick
https://www.diyaudio.com/community/threads/point-to-point-lm3886.187527/post-2580556
Patrick
This I find more interesting. 😉
Patrick
Patrick
I have been lazy and used the cheap PCB's available from China stuffed with the best parts I can find at Digikey for my last few 3886 builds using these nice switching power supply boards that can often be had for $15 each. I may have to try a dead bug build next time as it does look great. Here are a few pics of the 3886 amp my son built last month using a 3D printed case. These do work well as transconductance amps.
Why bother to use CHN PCBs when the original Brian Bell Gerbers have been made public ?
https://enjon.uk/2019/05/28/chipamp-lm3886-amplifier-pcb/
Patrick
Perhaps because they are not very good? Just look at the mounting holes. Clearances are too small.
Why? These are super easy to buy for $6 on Ebay, I've built six of them and they work well. If I send someone else's Gerbers to some PCB house I have to wonder if the files are the right format for that vendor, etc. It's a hassle. Is the Brian Bell design measurably better? Maybe. Will I hear the difference. Nope. It does look like fun doing a build without a PCB as show by the OP.
I looked at that website with the Brian Bell Gerbers a bit more closely, only to find that they recommend having the boards built by JLCPCB, which is located in China.
I measured that exact board here: https://www.diyaudio.com/community/threads/xy-lm3886-kit-review-measurements.294032/
Tom
I'm not sure symmetry is the most important criterion. The traces on the XY board are very short, over lay the ground plain, and have small loop areas. Because the power supply connections are very close together this also provides for small loop area for the board to power supply wire connections. The Brian Bell board has longer traces, and the board to board connection points couldn't be farther apart, forcing a large loop area in the connecting wires. The importance of symmetry in this instance is lost on me.
I have purchased, but yet to solder these very same XY PCB. By symmetry, I meant the +ve &-ve traces are almost of equal lengths instead of -ve trace being closer than +ve trace, as in XY PCB and many other different versions of LM3886 PCBs. Plus, space for large caps & small 0.1uF near power leads, which is absent in XY PCB. BTW I'm not an expert like tomchr in designing and evaluating PCB
Thanks for your review, now I'm confident enough that something you pitted against your very own Modulous 86(clear winner) performed marginally well. Probably not so much upto your high design standards but workable. BTW my LM3886 are genuine, got them more than 14 yrs ago from National sample program
Just one question, how low can the supply voltage go down before LM3886 runs into problems like rise in measurable distortion or other critical parameters getting worse?
Datasheets says it can work upto as low as +/- 10V, but graphs shows test results only around +/- 20V or above region. Usually, manufacturers avoid publishing test results around absolute minimum or maximum ratings for obvious reasons
I'm planning to run it at +/- 14V for overall lower power dissipation resulting in lower carbon footprint.
Last edited:
Fig 6 seems clear?
It is dang near dead to +/-4V (8V total), way under-fed to +/-8V (16V), and well-fed past there, which fits page 4 "Electrical Characteristics Power Supply Voltage 18 V typ 20 V (AOQL limit)".
This may be related to the claim that "Upon system power-up the under-voltage protection circuitry allows the power supplies and their corresponding caps to come up close to their full values before turning on the LM3886...."
Last edited:
Actually, I compared it against the LM3886DR, not the Modulus-86. The Modulus-86 uses error correction to drive the THD down into the abyss. That results in improvements on many other parameters as well. It would not be fair to compare a basic LM3886 circuit against the Modulus-86. That's why I compared it against the LM3886DR, which is a basic implementation of the LM3886 (i.e., no error correction, "just" a good layout, good decoupling, etc.) You can see that comparison here: https://www.diyaudio.com/community/threads/xy-lm3886-kit-review-measurements.294032/post-4769520
Perfect.
If the data sheet says ±10 V for the minimum supply voltage, then I would expect the part to meet the min/max specs listed in the data sheet spec table at that minimum voltage. The spec table is a legally binding contract between you and TI. If you ever find a part that does not perform according to spec you should contact TI. They do care about these things. It's much better for them for you to find a flaw than someone like Apple to do so. At least I would expect a different level of panic in your case. 🙂
You are correct that the graphs tend to highlight the best performance, but I have yet to see a part that fell apart when operated within the limits given in the spec table. Also note that the numbers listed in the 'typical' column of the spec table are not test limits. They're derived from a measurement of a number of samples. If the parameter can be measured on the tester, the typical number may be based on the measurement of hundreds or thousands of units. The THD number for the LM3886 is likely in this category as there's a graph showing the statistical distribution in Fig. 28). But if the test requires a lot of manual work, it may just be based on the measurement of, say, 20 units. Or, these days, based on simulation. Note that the simulation models tend to be very close to reality. Remarkably close, in fact. So "guaranteed by simulation" is actually somewhat reasonable these days, though if I was the design engineer on such a product I would definitely verify in the lab as well. I like to sleep well at night. 🙂
The THD for the LM3886 doesn't seem to depend strongly on supply voltage, so I wouldn't worry too much about that. Do worry about the output voltage swing, though. The drop-out voltage is a few volt usually. On ±14 V, I wouldn't count on more than about ±11-12 V of output swing, in particular with a 4 Ω load. Also, take mains variation into account to make sure you don't drop below ±10 V even if the mains should run 10% low one day.
Tom