| Electronics > Projects, Designs, and Technical Stuff |
| Composite amplifier: LM3886 + LME49720 |
| (1/4) > >> |
| JeanLeMotan:
In the past weeks I've been working on an audio amplifier using the LM3886 and LME49720 in a composite, bridged topology. I started off with TINA-TI simulations of the basic circuit until I got it stable with a decent phase and gain margins, step response and all and then I moved on to the actual schematic and PCB. The whole thing is finished (so I think) and I want a second opinion before I move on to actually prototype this. There's the info. 1. Goal: - 100W into 4 and 8 ohms. Not a hard requirement, I'm happy with anything in the 60-100+W range but since I didn't choose the speakers yet and I don't know their sensitivity I want some headroom. - THD as low as possible (for a hobby builder with zero XP). - Fun build that will teach me something. This, plus the previous point is why I choose the composite topology. Building just a LM3886 feels too much like lego: someone already did the work and I just connect the dots. The composite topology is a different beast on the other hand. - I want to design it to spec, not end up with smth good (or bad) by accident The attachments are in this order: a) Inverting part schematic used for simulations b) Non-inverting part schematic used for simulations c) Schematic used for phase-gain margin simulation/calculation (done according to this: https://training.ti.com/ti-precision-labs-op-amps-stability-3) d) Phase/Gain plot obtained from the schematic above e) Gain rise calculated according to this (https://training.ti.com/ti-precision-labs-op-amps-stability-4). This confirms the values from (d) f) Step response calculated according to this (https://training.ti.com/ti-precision-labs-op-amps-stability-4) g) Clipping transient response I attached the schematic used for simulation (non-inverting and inverting parts separately), the step response, the gain rise over the frequency range, the phase/gain margin plot. The full schematic with BOM and PCB is here: https://easyeda.com/jeanleflambeur/amp-LM3886-bpa200 The initial discussion is here (https://www.diyaudio.com/forums/chip-amps/334273-composite-amplifier-lm3886-lme49710-3.html) Extra info and questions: In the PCB, I tried to keep the track lengths for the feedback loops (both the inner one and the outer one) as short as possible. This forced me to place every second LM3886 on the bottom layer to keep the design symmetrical. The PCB in that area is also crammed. Does it make sense? Am I going too far here (or not far enough)? I have 2 grounds in the schematic: - The signal ground (QGND) used for the input reference and the signal opamps - The high-power ground for the LM3886 On the PCB I feed these in 3 points: High-power GND using a spade connector in the bottom-center of the PCB, input-gnd together with input using a 0.1" connector and low-power-gnd together with opamp-vcc and opamp-vee using another 0.1" connector. The idea is to merge these grounds on the PSU board. Does this grounding make sense? There are 4 power rails in the schematic: V+ (28V), V- (-28V), OV+ (15V) and OV- (-15V). V+ and V- are fed separately for each half of the amplifier: on the PCB I have 2 V+ and 2 V- connections. Is this ok or should I try to connect the power of the 2 halves on the PCB and feed it in only one point? I have a signal-ground plane over the central part of the 2 halves - both top and bottom in an attempt to reduce the QGND impedance (and to use the copper I'm paying for). Is this ok? The resistors I chose are metal-film, 0.1% and they are somewhat expensive and hard to find. The packages are all over the place (some 0603, some 0805, lots of MELFs) which ruins the aesthetic. Does it make sense to go for metal-film in this application? I'm not an audiophile, I like to measure and I'd like to choose components based on data and realistic expectations rather than feelings. Any other feedback you might have, it will be much appreciated. Thanks a lot! |
| nick_d:
That looks awesome. I am not really an expert on this, I just wanted to jump in and congratulate you on the detailed simulations and design work done so far. I will read the linked reference about how to simulate later on. From what you describe I think your grounding scheme sounds sensible. I will try to find time to examine the schematic later. In the meantime I'm keen to hear of progress. I also think the composite amp should be excellent when properly designed. There was another thread last month in which many engineers panned the idea, but despite several requests I never received a clear answer as to why, except the feeling seemed to be that it would be unstable and whatnot, my answer to that was the feedback loop might be harder to design but once done it should be OK. cheers, Nick |
| David Hess:
--- Quote from: JeanLeMotan on February 28, 2019, 09:17:36 am ---The resistors I chose are metal-film, 0.1% and they are somewhat expensive and hard to find. The packages are all over the place (some 0603, some 0805, lots of MELFs) which ruins the aesthetic. Does it make sense to go for metal-film in this application? --- End quote --- Metal film resistors do not suffer from excess noise so they are a good choice for audio applications but such tight tolerances are not required except in circuits like instrumentation amplifiers where common mode rejection is important. |
| JeanLeMotan:
--- Quote from: David Hess on February 28, 2019, 02:36:42 pm --- --- Quote from: JeanLeMotan on February 28, 2019, 09:17:36 am ---The resistors I chose are metal-film, 0.1% and they are somewhat expensive and hard to find. The packages are all over the place (some 0603, some 0805, lots of MELFs) which ruins the aesthetic. Does it make sense to go for metal-film in this application? --- End quote --- Metal film resistors do not suffer from excess noise so they are a good choice for audio applications but such tight tolerances are not required except in circuits like instrumentation amplifiers where common mode rejection is important. --- End quote --- Regarding the tolerance, The amp consists of 2 bridged halves of 2 paralleled LM3886 each. The paralleled amps have to have the same gain so that they don't fight each other and the 2 bridged halves should have the same gain to reduce distortion. Using 0.1% resistors was my attempt at this balancing. There are DC servos for each amp to make sure the voltage offsets don't cause problems. |
| JeanLeMotan:
--- Quote from: nick_d on February 28, 2019, 09:53:51 am ---That looks awesome. I am not really an expert on this, I just wanted to jump in and congratulate you on the detailed simulations and design work done so far. I will read the linked reference about how to simulate later on. From what you describe I think your grounding scheme sounds sensible. I will try to find time to examine the schematic later. In the meantime I'm keen to hear of progress. I also think the composite amp should be excellent when properly designed. There was another thread last month in which many engineers panned the idea, but despite several requests I never received a clear answer as to why, except the feeling seemed to be that it would be unstable and whatnot, my answer to that was the feedback loop might be harder to design but once done it should be OK. cheers, Nick --- End quote --- Thanks a lot! It was quite difficult to get it stable, I spend more than one week tweaking it. The inner loop was fine as it's stable at >10x gain but the outer loop tended to oscillate at ~5 MHz due to the LME49710 being so much faster than the 3886. Anyway, the 2 compensation caps fixed this and I think I have now a very stable amp... simulation. The next step is to build it. The PCB layout is as tight as I could make it - the only unknown is the grounding scheme. I don't know what PCB parasitic capacitance and inductance will to do stability. Also not sure how reliable are the TINA-TI models for the chips. The composite topology is definitely proven and this exact choice of components and configuration is already present in a line of products, see here: https://www.neurochrome.com/modulus-86/ So I know it's possible to stabilize. |
| Navigation |
| Message Index |
| Next page |