Yeah. Pretty much.
I'd consider the fact that I can take an LM3886 amplifier and reproduce the data sheet performance with the goop but fall 20 % short on the output power without the goop to be a hint.
I guess it's just one of those unknowns in life that people choose to live with. It's not like we're trying to build a space shuttle here.
I'll get right on it!
Tom
Last edited:
Actually, if one can still find the classic application notes on heat sinking from Motorola and Burr-Brown (two that I recall), they specified Belleville washers to apply the right amount of force without ramming the mounting screw as far as possible. Heat sink mounting kits from companies like Thermalloy (including insulating washer, 'silpad', screw, nut and washer) use split lock washers. These two examples should indicate the amount of mounting force necessary.
The spring washers or split lock washers are common in Europe, actually. On this side of the pond people seem to favour the serrated lock washers. Go figure!
On-Semi has the most comprehensive guide to semiconductor mounting that I'm aware of: http://www.onsemi.com/pub_link/Collateral/AN1040-D.PDF
According to figures 5 and 6, the thermal resistance doesn't seem to vary a whole lot once the mounting screw torque exceeds 3-4 in-lbs (0.34-0.45 Nm in human comprehensible units).
One could pick up a set of torque screwdrivers and just torque the mounting screws. I've often drooled over the Wera 7440/41 set.
Tom
On-Semi has the most comprehensive guide to semiconductor mounting that I'm aware of: http://www.onsemi.com/pub_link/Collateral/AN1040-D.PDF
According to figures 5 and 6, the thermal resistance doesn't seem to vary a whole lot once the mounting screw torque exceeds 3-4 in-lbs (0.34-0.45 Nm in human comprehensible units).
One could pick up a set of torque screwdrivers and just torque the mounting screws. I've often drooled over the Wera 7440/41 set.
Tom
Attachments
Last edited:
The LM3886 is not suitable for bridging except for 8 Ω loads and above. Hence, the Modulus-86 is also not suitable for bridging except for 8 Ω loads and above.
The LM3886 can "only" deliver 7 A of output current and in a bridged configuration, each amp half "sees" half the load impedance. If you calculate V = I * R, you'll find the limits on the output swing imposed by the output current.
Tom
And here's the math: http://www.diyaudio.com/forums/chip-amps/286969-rowland-lm3886.html#post4619486
Perfect!
Anand.
Tom,
Speaking generally when considering paralleling chip amps, do you have any concerns over the initial settling time (those first few milliseconds following power on) of each stage when using devices that may see differing time constants on some of their pins (such as bootstraps, input caps and feedback return caps etc) and how that might impact on reliability i.e. the amp going bang at power up ?
It occurred to me that this phase of operation could place heavy reliance on the chips internal protection circuitry.
Speaking generally when considering paralleling chip amps, do you have any concerns over the initial settling time (those first few milliseconds following power on) of each stage when using devices that may see differing time constants on some of their pins (such as bootstraps, input caps and feedback return caps etc) and how that might impact on reliability i.e. the amp going bang at power up ?
It occurred to me that this phase of operation could place heavy reliance on the chips internal protection circuitry.
50 year old technology
I get your point but I have time to mess around. Also describing 50 year old technology as advanced seems a contradiction.
I get your point but I have time to mess around. Also describing 50 year old technology as advanced seems a contradiction.
I wish I had as much time to answer your questions as you have messing around and creating questions. As we discussed a few (many?) pages back, procrastination can be a real killer of productivity and what better way to procrastinate than to spend time on DIY Audio, eh? Although, one could argue that I'm not obligated to respond to any questions here, I do feel a sense of obligation to provide an answer in a timely manner. It's who I am. I enjoy being helpful and delivering good service.
It is a double-edged sword, though. If I spend too much time answering questions, I won't have enough time to develop new products, improve existing products, and my business will fold. Tradeoffs, tradeoffs... Please be keenly aware of this when you ask questions.
Please read the responses to your questions, including the references. Many of the responses contain good advice and good information. In some cases, the correct answer is, "it doesn't matter". Please accept that.
My advice will always be: Please follow the instructions. All circuits I offer will perform as intended when assembled according to the instructions. All my circuits are built with off-the-shelf parts available from the various distributors (Mouser, Digikey, et al.) No voodoo, resistor dowsing, clamping bars, or whatnot are required. Just build it according to the instructions and plug it in.
I do admire your curiosity and many of your questions are interesting in their own ways. For the questions that end in "it doesn't matter", I suggest that you start a separate thread rather than continuing the discussion in this thread. It helps keeping this thread on topic so other builders can find the answers they need without having to scroll through page after page on how to polish an LM3886, which does not need polishing in the first place. If you do decide to take part of the discussion here to a separate thread to get your curiosity satisfied, you're always welcome to drop a line here saying, "Those who'd like to join my discussion on The Optimization of LM3886 Mounting Methods can find my thread here (link)". That way those interested can get their curiosity satisfied and those of us who are a little more "build it according to the instructions and be done" can continue here. I'll be more than happy to participate on any of your threads to the extent my time allows.
Thanks,
Tom
Excellent question! Startup is tricky for any circuit. I spent a fair amount of my career designing precision biasing circuits and large amounts of time in design reviews convincing my colleagues that, yes, the circuit will start up. I appreciate the thought.
In case of the Modulus-86, the LM3886 is in the mute state until the supply voltage reaches about 2/3 of its nominal value. The internal circuitry of the LM3886 is active way before it comes out of mute, so the amount of negotiation between the two halves of a parallel amp is really just the two DC servos adjusting to the input offset voltage of the main input pair versus the mute pair of the two LM3886es. This takes a few seconds during which a few 10s of mA may flow in the ballasting resistors. Once settled, I measured 45 µV drop across 0.1 Ω (450 µA) of imbalance between the two halves of the MOD86 Parallel amp.
I have not seen any signs of distress during startup in a parallel amp. I would have noticed pretty quickly during the Parallel-86 design.
It is certainly possible that the protection circuits kick in during start-up, just as they could in an LM3886 amp by itself. Considering that an LM3886 (and Modulus-86) is able to start up without any thumps or pops in the speakers at all, I'm thinking that the designers at National have taken care of any startup and shutdown-related concerns.
Now, that's not to say that you can't force it to fail. I don't know how, but I'm sure with enough abuse something will fail. The only scenario I can think of is if the VCC supply was shorted to ground and VEE at the full voltage. I've actually driven the MOD86 a few times by accident with only the VEE supply turned on. It happens at times in the lab as I'm driving the two rails with separate supplies and sometimes a finger slip causes only one ON/OFF button to be pressed. I've not seen any failures or anything that would indicate failures even under these conditions. In an actual build, if one supply was to short to ground the mains fuse would blow immediately on startup.
I have noticed that if the VCC rail collapses before the VEE, you will get a burp on the output on turn-off. This makes perfect sense as the MUTE signal is derived from the VEE rail. I discovered the effects of this when I built the 4xMOD86 + MiniDSP amp for a friend and powered the MiniDSP via an LM317 from the positive rail. Occasionally, the amp would thump on power-off. The quick-n-dirty solution was to add a CCS on the VEE rail. The CCS was set up to draw about 10-15 % more current than the MiniDSP. It's a bit wasteful on the power, but it did fully prevent any thumping both on startup and shutdown as the VEE rail always collapses slightly before the VCC rail, thus, muting the LM3886 on shutdown.
Awesome. Thank you. I had exactly that post in mind. I should get that onto my Taming the LM3886 pages one of these days.
Tom
Thanks for your time Tom
I appreciate your advice and I admit I get carried away with the what ifs. I'll curtail that. The link to ON-SEMI was very interesting.
Thanks,
henry
I appreciate your advice and I admit I get carried away with the what ifs. I'll curtail that. The link to ON-SEMI was very interesting.
Thanks,
henry
Actually I've rather been enjoying Henry's discourses into things metallic. We all go around the loop on this thinking that improving the thermal coupling to the heatsink by 20% will give us more powa, but then you quickly realise the harsh reality that the overall gain is to small to hear. But IMO the hobby of seeing how far you can push SOA by employing interesting tricks and brute force is a harmless one, esp if you have the ability to press copper slugs into aluminium.
I am I will note lousy at metalwork, so the case for mine is going to elicit lots of giggles when I start trying to fit everything in.
I am I will note lousy at metalwork, so the case for mine is going to elicit lots of giggles when I start trying to fit everything in.
The supply voltage presents another harsh reality. Even if you obtained the bare die from TI and soldered it directly onto the heat sink and cooled the heat sink with liquid nitrogen, you still would not be able to get more than 28 Vp (56 Vpp) swing from a ±28 V supply.
We can all geek out about something. That's part of DIY. I suggest that we try to move the field forward and help new builders by providing information that is directly applicable rather than launch on tangents of what-ifs. What would happen if gravity caused us to be repelled by the earth? Could we get more power of of the LM3886 under these circumstances?
I do agree that the few bits of build process that Henry has shared and the stories about his machine shop capabilities are interesting. I would love to see more build pictures and stories about how well the bits came together.
Tom
I would also note that we have had people fall foul of mounting using the tab when they solder the LM3886 down before they do a mount check. The advantage of clamping across the body is that you can't get that twist (well you can but it's less likely). For the DIY'er the knowledge that its down flat and staying that way does help the feeling of security.
There is also the fact that less precision in the drilling and tapping of holes is required for a bit hoofing bar. Us drill gorillas like that...
There is also the fact that less precision in the drilling and tapping of holes is required for a bit hoofing bar
. Us drill gorillas like that...
You can't fit a clamping bar Rowland-style across the LM3886 in the MOD86, though. The supply decoupling caps are in the way. I suppose you could mount the caps on the bottom of the board - same with the mute cap - but you're really trying to solve a problem that isn't there to begin with.
If you're going to use any mounting method other than a screw through the hole in the tab of the LM3886, I suggest using a spring clip. Mouser P/N 532-MAXCLIP08G should work.
Tom
If you're going to use any mounting method other than a screw through the hole in the tab of the LM3886, I suggest using a spring clip. Mouser P/N 532-MAXCLIP08G should work.
Tom
Henry, exactly my point. There has been little new in the world of 'standard' heat-sinking in a good 50 years. The "ON Semi" app-note that Tom dug up is, in fact, a very old Motorola app-note that was 'rebranded' when ON was split off from Motorola.
Yes, it is useful to minimize the thermal resistance at the device-to-heat-sink interface, but that doesn't improve the other thermal resistances in the thermal circuit -- (1) thermal resistance between transistor junctions and the surface of the IC package and (2) the thermal impedance between heat sink and the ambient air. The most useful thing to do, and which is in the DIYer's control is to reduce the impedance of the heat sink, either by larger/better size/shape or better ventilation.
Tom has a very good section on his web site (and in the construction manuals) about choosing heat sinks. Ultimately, your amp's output capabilities are limited by maximum supply voltages, output current limits, safe-operating-area (SOA) of the output transistors, etc., even with 'ideal' heat sinking. Taking 'extreme' measures beyond standard 'best practice' to improve the device-to-heat-sink interface provides minimal additional benefit.
Yes, it is useful to minimize the thermal resistance at the device-to-heat-sink interface, but that doesn't improve the other thermal resistances in the thermal circuit -- (1) thermal resistance between transistor junctions and the surface of the IC package and (2) the thermal impedance between heat sink and the ambient air. The most useful thing to do, and which is in the DIYer's control is to reduce the impedance of the heat sink, either by larger/better size/shape or better ventilation.
Tom has a very good section on his web site (and in the construction manuals) about choosing heat sinks. Ultimately, your amp's output capabilities are limited by maximum supply voltages, output current limits, safe-operating-area (SOA) of the output transistors, etc., even with 'ideal' heat sinking. Taking 'extreme' measures beyond standard 'best practice' to improve the device-to-heat-sink interface provides minimal additional benefit.
Yeah... If the laws of physics have changed recently, I certainly misted the change notification.
Thank you.
Exactly. Surfaces free of nicks and gouges, thermal goop, and a machine screw will do. Anything beyond that won't matter even one imperial gnat fart.
Tom