Neurochrome LM3886DR Build

Thanks Tom for quick response.

Out of curiosity again:
1. In your design you are using a bit different values for Thiele Network, why?
2. In some designs they recommend putting NFB resistor directly on chip pin. In your PCB one lead of nfb resistor is close to the chip pin while the other has a thin trace leading to output (before Thiele Network). I bet there is reason for this.
Can you elaborate on those points. Thanks.
 
Most designs use a *way* under-damped Thiele network. That results in excessive ringing with a capacitive load. I aim for a clean response with a 8 Ω || 100 nF load.

Yeah. I'm actually one who recommends using short traces on the feedback resistor. If you are not using the Thiele network, the trace inductance in series with the feedback resistor can cause stability issues with capacitive load. When the Thiele network in place these issues go away as the capacitive load is effectively isolated from the feedback network.
The voltage at the point where the feedback trace taps off is the voltage that is controlled by the amplifier. In the ideal world, this point would be right on the voice coil of the speaker and there would be a similar arrangement on the ground side of the voice coil. Unfortunately, this puts the feedback on the outside of the Thiele network which may cause some issues. The best point is therefore the point where the output voltage enters the Thiele network.

Tom
 
I've been absolutely flat out with work, but slowly chipping away at getting this finished - thankfully a lot of the work is getting me familiar with CAD which I will use to the benefit of this project.

I ended up buying a new toroid as the other one is going towards a build with an integrated pre-amp. Nothing fancy, 2x25v/300VA.

20160903_195342.jpg

Got the two eBay regulator modules (set to 30v output) set up on an old heatsink with a mica washer, along with one amp module (was doing this at 8pm at night so didn't bother hooking up both).

20160903_195348.jpg

Ran a 50Hz sine wave (as a stress test) through one of the spare speakers for 10 minutes continuously @ ~30w output. Zero voltage sag even when increasing this to 50w output. Pretty happy for $30 worth of Chinese power supplies!

Good news is there is no audible background noise, and both the regulators stayed completely cool in this time, with the LM3886 getting fairly warm to the touch (but still cool enough to hold my finger onto). I know this is a fairly big chunk of heatsink and only one module was running, but on the final assembly the results should be similar.
 
Hi Tom,

Are there any guidelines for power rating of Thiele network resistor? I know 2W is suggested but is the 1W enough?
I got the following resistor under my hand that I could choose:
1R5 5W wire-wound "non-inductive" design - afraid of its inductance anyway
10R 2W metal film - afraid that value is to much
2R2 1W metal film - afraid that power rating is to low

Which one would be the best to use in my case?
 
if you never apply HF test signals then the correctly rated Zobel and/or output damping resistors will never operate hot.

It's the HF that passes easily through the capacitances that will heat the resistors.
Take a 4r7 Zobel resistor in series with a 100nF capacitor.
The capacitor reactance @ 20kHz is ~80ohms
If you combine that with the 4r7 resistor you get a total impedance of sqrt (4.7² + 79.6²) ~ 79.7ohms.
The current passing when 20Vac is applied is ~251mAac and the power dissipated in the Zobel resistor is <300mW

But apply a 50kHz squarewave and the Zobel resistor will smoke very quickly and the esr heating in the capacitor may well damage the capacitor as well.

BTW,
1r0 600mW 1% resistors can be stacked in either parallel, or series, or a series parallel, combination to make accurate high power "specials" that are cheaper and quicker (but not nearly as pretty !) than buying in two values of a special when you have to pay postage for low quantities and wait days, or weeks, for the delivery.
I am nearly 40% through my box of 1000 1r0 600mW 1% metal film I bought about 5years ago.
 
Last edited:
1R5 5W wire-wound "non-inductive" design - afraid of its inductance anyway
10R 2W metal film - afraid that value is to much
2R2 1W metal film - afraid that power rating is to low

Of those choices, 1.5 Ω, 5 W is the best option. I generally stick with 2 W types. During sine wave testing above the audio band those resistors do get hot. Same for the Zobel resistor as Andrew mentioned.

Tom
 
Last edited:
Thanks Andrew, Thanks Tom,

I soldered PCBs. However I used bit different components, because I wanted to adjust it to my specific needs. Basically I wanted to build gainclone dedicated to Soekris dam1021.

so, as a source I am useing dam1021 Sokeris:
- 640ohm static output impedance
- 0 DC offset on output
- I also use built in bit perfect digital volume control, so there is no need for analog volume control between Soekris and my gainclone.

Due to this in my gainclone build:
- I wanted to get rid of Input cap because: there is no DC offset
- No feedback cap because source impedance is static, so DC offset from lm3886 chip once set should be stable (maybe drift a bit due to temperature).
- as a speaker protection I furtherly want to use some upc1237 based circuit

Since I do not want to use feedback cap I wanted to minimize DC with other means. This is what I found:
- value of feedback resistor should be of same value as input/load resistor, so I used 22K on those position
- resistances to ground from V+ and V- should be equal. In my case V+ is (640ohm source impedance + ~1kohm resistance from emi filter paralleled with 22K load resistor) V- is (value of Rf resistor 780hm in my case). So the difference is less than 1kohm with source connected.
- furtherly I want to use lm3886 instead of lm3886tf to have better heat dissipation. For now I used desoldered lm3886TF from other boards.

So I build two boards like that and as test I connected 4R7 12W resistor on the output:
- one board works. With source connected I got stable ~20mV DC offset. PSU readings: -/+27.5VDC. Music plays well.
- the other board does not work. The connected test resistor got EXTREMELY hot and burned some of my desk. DC offset was around 19V! and PSU readings was: ~22V on (+) rail and 26V on (-) rail and then when connected to the speakers ... no, I obviously did not do that :)

My question is: Do you guys encountered such situation? Do you think it might be due to some faulty chip or capacitor or maybe kind of short on PCB in some specific place? Any ideas?

Thanks,
 
Last edited:
I soldered PCBs. However I used bit different components, because I wanted to adjust it to my specific needs.

Okay... Just beware that you can get yourself into trouble that way. My LM3886DR will work well when you use the specified components. Components similar in value and, for capacitors, similar dielectric may be substituted.

That said, the LM3886 is a popular circuit with people who enjoy tweaking the circuit to suit their needs. Before colouring outside the lines, I strongly suggest that you read the LM3886 data sheet, specifically the applications section in the back.

Due to this in my gainclone build:
- I wanted to get rid of Input cap because: there is no DC offset

Fair enough. Just beware that the DC gain is 20x (26 dB) in that case.

- No feedback cap because source impedance is static, so DC offset from lm3886 chip once set should be stable (maybe drift a bit due to temperature).

That's not such a good idea. The LM3886 can have as much as 10 mV of input-referred offset. 10 mV * 20 = 200 mV offset on the output of the amp. You'll need to deal with this elsewhere in the circuit.

Since I do not want to use feedback cap I wanted to minimize DC with other means. This is what I found:
- value of feedback resistor should be of same value as input/load resistor, so I used 22K on those position
- resistances to ground from V+ and V- should be equal. In my case V+ is (640ohm source impedance + ~1kohm resistance from emi filter paralleled with 22K load resistor) V- is (value of Rf resistor 780hm in my case). So the difference is less than 1kohm with source connected.

I can't quite follow what you're doing there. Do beware that if you set the LM3886 up for a gain less than 10 (20 dB), you will have an amp that is not stable.

- furtherly I want to use lm3886 instead of lm3886tf to have better heat dissipation.

There's no need for that, but also no harm in doing so.

- one board works. With source connected I got stable ~20mV DC offset. PSU readings: -/+27.5VDC. Music plays well.
- the other board does not work. The connected test resistor got EXTREMELY hot and burned some of my desk. DC offset was around 19V! and PSU readings was: ~22V on (+) rail and 26V on (-) rail and then when connected to the speakers ... no, I obviously did not do that :)

If you have one board that works and one that doesn't, I suggest looking for differences between the two boards.

My question is: Do you guys encountered such situation? Do you think it might be due to some faulty chip or capacitor or maybe kind of short on PCB in some specific place? Any ideas?

The LM3886DR PCBs are made in a commercial PCB fab. The traces and spaces are wide and do not push the fabrication process at all. It is extremely unlikely that you have a board with an electrical short or broken trace.

The LM3886 ICs are tested by TI before they leave the fab. If you used new ICs from a reputable source, it is extremely unlikely that you'll have a defective one. You mentioned you desoldered the ICs, so you could have damaged the IC in the process. They tend to be pretty rugged, though. I bet you have a wrong connection somewhere or accidentally swapped two components. I suggest comparing the two boards you have to see if you can find the difference.

Tom
 
Hi Tom, Hi Mark

Thanks Tom for explanations and Mark for the tips. It was my fault. Now two boards are up and running. I get around ~90mV of DC with source disconnected and ~25mV DC with source connected. After running of several hours with different volumes DC offset stays at same level. I can live with those ~25mV.

I got two more questions:
1. Since Soekrs R2R has LPF on its output with cut of frequency at 270kHz, do I need "heavy" EMI filtering in lm3886 input?
2. Is it worth using naked z-foil resistor on feedback loop and input? They are of lowest noise, however their inductance is <0,08uH, whereas I found somewhere that metals films inductance is around 2nH. Tom, do you have any thoughts on this one?

Thanks,
 
1. Since Soekrs R2R has LPF on its output with cut of frequency at 270kHz, do I need "heavy" EMI filtering in lm3886 input?

The output filter on the Soekrs will do nothing against any EMI induced on the interconnect cable. That's what the RF input filter on the LM3886DR is for.

The RF filter in the LM3886DR causes a 0.02 dB loss at 20 kHz, so it has no impact on audio frequencies. It starts rolling off around 500 kHz and hits -40 dB at 8 MHz.

2. Is it worth using naked z-foil resistor on feedback loop and input?

No. Regular metal film resistors are sonically transparent as far as I can measure. I see no reason for 'bling' resistors in this application.

Tom
 
Tom, thanks again for explanation and great product:
- great PCBs (very robust - I did lot of soldering/desoldering on them and they are still alive unlike other PCBs I had where sometimes traces fall of after first desoldering)
- great detailed manual sent directly to my email (lots of detailed explanations and instructions)
- great support on this forum (quick response and nice explanation)
 
Last edited:
Well, this takes things back a step. While I was working on machining the cast slabs of aluminium up, somebody was playing around in the electronics lab with a 2kw electric skateboard. Somehow, they managed to crash it into the storage shelves I was using which knocked all my stuff to the ground.

Snapped the IC's off both of the regulated board and bent the 3886 up really badly :( I've managed to carefully rebend the pins on the 3886 back into place, however the regs on the power boards aren't repairable (and it's more expensive to replace them than the entire board).

I've ordered new boards which should be here in the next week or so, however they're slightly different so I won't get too far with the machining just yet.