LM3886 PCB vs Point-to-Point (with data)

Jump to Latest
P

proid

Member
Joined 2011
tomchr said:
I don't understand your question. My LM3886DR has a low-inductance path to ground for the feedback network. You can see the performance measurements on its product page: LM3886 Done Right: Best DIY implementation of the LM3886 80W gainclone – Neurochrome

Is that what you're asking about?

Tom
Click to expand...

In post 221 you state that R1 inductance doesn't have impact to THD but in the post i quote before, your statement is other way, may be i misunderstand?

tomchr said:
As noted in my LM3886 Bypassing Revisited Thread, the supply impedance is dominated by the resistance of the connectors and wires at frequencies below 1 kHz. In the mid audio band (1~10 kHz), the impedance is dominated by the ESR, ESL of the on-board electrolytic can. My simulations show that 470~1000 uF is about optimum. I saw minuscule improvement when going from 1000 uF to 10000 uF. Above 10 kHz, the supply impedance is determined by the local bypassing network (C1, C4, C13, C16 in my circuit).

Basically, this says that a regulated supply will provide a benefit at frequencies below 1 kHz. Above 1 kHz, the performance is determined by the on-board components. The improvement I measure at 20 kHz is real and must be caused by the on-board components or layout changes. I used the same generic 1000 uF, 10 uF, and 47 nF caps that I used in my previous PCB and P2P experiments, hence, the improvement must be from the layout.

The new layout improves five things:
  1. The supply inductance is much, much lower due to the use of planes for the supply lines.
  2. The decoupling has moved as close to the LM3886 as possible.
  3. The ground connections of the decoupling caps have been placed as close to each other as possible. This to get as close to a star ground as possible, thereby, controlling the HF current flow in the ground plane.
  4. The parasitic inductance of the Zobel network has been greatly reduced and its ground connection has been moved to as close to the star ground as possible.
  5. R1 has moved off the star ground and is grounded at the load ground. This should minimize the impact of any error signal introduced between the load ground and the star ground.

My guess is that the better grounding and elaborate use of planes and pours is responsible for the improvement at 20 kHz. The overall improvement across the entire audio band is caused by better grounding. See page four-ish of this thread for my grounding experiments.

Parasitic inductance in R1? Compared to my previous layout, there's probably an additional 30 nH of inductance in series with R1. Does this impact the performance at 20 kHz? Let's run the numbers: XL = 2*pi*f*L --> XL = 2*pi*20E3*30E-9 = 3.77 mΩ. This is negligible compared to the 1 kΩ resistance, hence, the parasitic inductance in series with R1 is not responsible for the improvement in THD at 20 kHz.

Regulated supply? There's no doubt that keeping the charging pulses off the board will have an impact on the performance of the amp. The question is how much charging current one would see if using reservoir caps of, say, 22000 uF off the board and routing a heavy wire to the board. My guess is that around 1/22 of the current would run through the caps on the board. How much crap this introduces is a good question. I do plan to measure this...
That said, regulated supplies are rather inexpensive, so why bother with the heavy transformers?

~Tom
Click to expand...
 
tomchr

tomchr

Member
Joined 2009
neurochrome.com
Paid Member
Uh... You are asking me to provide an exact account of what I thought in 2014, seven years ago. I can barely remember what I had for dinner yesterday... :)

If the "R1" is referring to the R1 in the schematic in Post #1 of this thread then I'm not sure what your argument is. In my LM3886DR layout the inductance there is minimized by using a ground plane. In Post #221 I say that the inductance in series with R1 is low anyway. So I'm taking something that's low and making it lower and somehow you find this problematic? I'm confused... I tend to use the available copper. I pay for it whether I use it or not...

Tom
 
You must log in or register to reply here.
Top Bottom