That's the idea, yes.
Even in the RCA version of the amp, you're still getting good rejection of anything injected on the input lines. See the attached graph. With differential input, you get a CMRR following the red line. With RCA input on the Modulus/Parallel-86, you get the purple line. With any other single-ended amplifier design, you'll get the blue line (straight at 0 dB, no rejection at all). Ignore the green line.
With the MOD/PAR86, take the if it ain't broken, don't fix it approach. I see no need for ground breakers and I can't recommend them as they're against the electrical code in most places.
My prototype circuits typically involve a PCB mounted on a heat sink with the power supply and wiring in spread eagle formation on the lab bench. I've never had audible hum in the speakers. The highest mains hum component is at -120 dBV (MOD86). That's 1 µV. Most tube guys would be envious of <1 mV of mains hum. The MOD86 is 1000x better...
~Tom
Attachments
Last edited:
Funny how people approach the same situation differently. After about four times of explaining LM3886 stability and layout, I got tired of it and set up my Taming the LM3886 website. I saw a market need and went after it.
~Tom
IMO, the LM3886 is too far inside the board. I cannot get the heatsink to be flush with the output device as the board is in the way. Even after filing the board and pulling hard on the chip.... it will barely reach. Careful not to file too much, since the ground plane is very close to the edge of the board as well.
It is possible that a metal shim or a foil will be needed so that it will reach the heatsink. I do not wish to put keratherm between it and the heatsink as the part is already electrically isolated.
Moving it towards the edge by at least 2mm would work very well.
This is the rev 1.01 of the Modulo86 board.
It is possible that a metal shim or a foil will be needed so that it will reach the heatsink. I do not wish to put keratherm between it and the heatsink as the part is already electrically isolated.
Moving it towards the edge by at least 2mm would work very well.
This is the rev 1.01 of the Modulo86 board.
It works a bit better if you mount the part to the heat sink first, then solder it. I've had no trouble with the TF package (isolated). The T package (non-isolated) is the same dimension from the pins to the rear of the package. It should fit just as well.
Mounting the IC is covered quite well by Richidoo's post, Post #73.
I did move the LM3886 a tad closer to the edge of the board for Rev. 2.0. That'll give you at least 0.5 mm between the board and the heat sink. I did not feel comfortable putting the IC closer to the board edge. I do some amount of copper available for the supply connections...
~Tom
Last edited:
Good idea to screw the chip to the heatsink before soldering it to the PCB. Even with more forgiving PCB holes, the chip will still slouch due to gravity. This will stress the pins when bolted to heatsink. I have read that is to be avoided, but maybe it doesn't matter.
You can use quikchip to desolder the IC to try again. Chip Quik -
You can use quikchip to desolder the IC to try again. Chip Quik -
Thanks Tom
Yeah, I've mounted to the heatsink first in the past, and should have done it this time too. I did not think the clearances were so demanding.
I just followed the build guide. Perhaps a footnote on installing U2 to the heatsink first would help the less experienced like me. I just thought that with so many pins, the placement would be perfect.
Thankfully, it was not too much out, filing down the edge was enough.
Yeah, I've mounted to the heatsink first in the past, and should have done it this time too. I did not think the clearances were so demanding.
I just followed the build guide. Perhaps a footnote on installing U2 to the heatsink first would help the less experienced like me. I just thought that with so many pins, the placement would be perfect.
Thankfully, it was not too much out, filing down the edge was enough.
From the build guide:
This is mostly an issue for the Rev. 1.0 and 1.01 boards as the edge of the board is flush with the back of the IC. For Rev. 2.0 I did move the LM3886 a bit closer (forget if it was 25 or 50 mils) to the edge of the board.
Sorry you ran into trouble on that one. A piece of copper bar stock between the LM3886 and the heat sink should take care of things, though. It'll look pretty too. Just add thermal goop on the thermal interfaces.
~Tom
This is mostly an issue for the Rev. 1.0 and 1.01 boards as the edge of the board is flush with the back of the IC. For Rev. 2.0 I did move the LM3886 a bit closer (forget if it was 25 or 50 mils) to the edge of the board.
Sorry you ran into trouble on that one. A piece of copper bar stock between the LM3886 and the heat sink should take care of things, though. It'll look pretty too. Just add thermal goop on the thermal interfaces.
~Tom
I got the idea to check out the heatsink fitment and alignment from reading this step in the build instructions. But it was a fluke, normally I don't read the instructions either.
Thanks Tom. I have a 39K to try if it does not start up.
I'm going to put my other concern from the other thread here too
I misread the RCA drawing and did not connect shield to signal- at the RCA.
It looks like it's for the Pin1 + Pin 3 connection that has to be established.
Can this be done with a jumper at the PCB as well/instead? That would leave the shield disconnected at the RCA end.
I will be finishing it tomorrow, and will post a shot of the implementation.
I'm going to put my other concern from the other thread here too
I misread the RCA drawing and did not connect shield to signal- at the RCA.
It looks like it's for the Pin1 + Pin 3 connection that has to be established.
Can this be done with a jumper at the PCB as well/instead? That would leave the shield disconnected at the RCA end.
I will be finishing it tomorrow, and will post a shot of the implementation.
You can make the Pin 1 to Pin 3 short at the board. You will get slightly better CMRR performance if you make the Pin 1 to Pin 3 short at the RCA connector rather than at the board. Even better performance, yet if you make the connection at the source using a pseudo-differential interconnect cable. And best performance if you use differential signalling.
Tom
Tom
The offset is about 1 mV while the LM3886 is muted. As soon as it comes out of mute, the offset will spike to around 50-75 mV and settle to about 200 uV in a few minutes (MOD86 R1.0). MOD86 R2.0 and PAR86 R1.0 improve this to a 35 mV spike and settling to 200 uV in about 10 seconds.
The initial spike and the settling are caused by the DC servo. The capacitors need to charge and that takes time. When designing the DC servo for MOD86 R2.0, I spent quite a while minimizing the turn-on spike. 35 mV was the best I could do. The DC servo in MOD86 R2.0 (and PAR86 R1.0) was actually one of the main design efforts. The servo needs to settle fast, provide as little turn-on pop as possible, and not compromise the THD at 20 Hz, all at the same time. That's a pretty tall order, actually. It was a fun design process...
A 50-75 mV pop is well below audible on my speakers. Even with the MOD86 R1.0 in my living room, I only occasionally notice a small pop when turning it on, and I have my head right up against the speaker when I turn it on. The 35 mV pop of the MOD86 R2.0 is below what I can hear.
Tom
The initial spike and the settling are caused by the DC servo. The capacitors need to charge and that takes time. When designing the DC servo for MOD86 R2.0, I spent quite a while minimizing the turn-on spike. 35 mV was the best I could do. The DC servo in MOD86 R2.0 (and PAR86 R1.0) was actually one of the main design efforts. The servo needs to settle fast, provide as little turn-on pop as possible, and not compromise the THD at 20 Hz, all at the same time. That's a pretty tall order, actually. It was a fun design process...
A 50-75 mV pop is well below audible on my speakers. Even with the MOD86 R1.0 in my living room, I only occasionally notice a small pop when turning it on, and I have my head right up against the speaker when I turn it on. The 35 mV pop of the MOD86 R2.0 is below what I can hear.
Tom
Last edited:
It sounds terrific, and it is only doing 100 -> 900 Hz. Much better than the previous amp that was in that spot -- Marantz 250M.
I'll have to build some more of these!
This on has no on/off switch and no indication of on/off. I leave my amps on all the time anyway.
I'll have to build some more of these!
This on has no on/off switch and no indication of on/off. I leave my amps on all the time anyway.
