Load dump is not about connecting two batteries in series to achieve 24V. It's about (accidentally?) disconnecting the battery while charging from the alternator - which is an inductive source. When the load taking the current is disconnected that current is going to continue flowing through the vehicle electrical system which will cause a voltage spike until the current goes down.
On a related note, I heard this 2nd hand from a guy who was talking to CAT about a module they wanted to have 80V over voltage protection for a crazy amount of time (minutes). My friend asked what scenario required that and it was something like this: Guy at a construction site can't start his bulldozer. He doesn't have jumper cables, but he's got an arc welder... I don't know if that actually happened but it would not surprise me.
Yes, that was a weird mistake for someone in the author's position to make.
That said, I may not understand what a 'load dump' condition is, myself -- I always thought it referred to the common case where a high current load is switched off abruptly, causing the voltage to spike until the alternator's regulator circuit can respond by reducing the field coil current.
Disconnecting the battery would not be expected to cause such a condition, since the charging current is not all that high most of the time compared to other loads. I wouldn't expect much of anything to happen when disconnecting the battery from a running car. It should just keep running.
In old cars it was a quick test to see if the alternator was working so some people still think that this is a good idea, IT IS NOT! newer alternators are controlled via LIN bus by the BCM so to generate more current when needed not just using battery voltage as a reference, if you disconnect the battery the voltage can go up to 30v even a bad battery or one with very low charge can make the charging system either shut off or go nuts.
The battery acts as a big capacitor to smooth the voltage and depending at which speed the alternator is turning I'd expect the voltage to vary wildly. This may make weird things happen especially on newer cars
Just to expand on this a bit. The load dump is not directly caused by the inductance of any component. That would only produce a very short voltage spike that would be easy to suppress. A load dump can involve greatly increased voltage for up to half a second.
Alternators tend to produce a constant current in proportion to the field winding current (the variable the voltage regulator controls). When you disconnect the battery the alternator attempts to maintain that current by increasing the voltage. The voltage regulator has to reduce the current in the field to close to zero to stop this voltage surge. The field has some magnetic energy (inductance) that the voltage regulator has no good way to get rid of quickly. So the output voltage zooms up to maybe 10 times the normal voltage for an extended period and any electronics in the car might be very sad.
Mig welders are used as temporary power supplies all the time. They allow for variable DC voltage (some offer AC), current, wave forms, and the bigger one have close to 100% duty cycle. Here's a guy using a mig welder to power a 15 ton radial arm dill. https://www.youtube.com/watch?v=2zYrabDGIEg
I have great memories building small guitar amplifiers out of its little brother the LM386.
Adding just a transistor to handle the high impedance of the guitar and a few cents worth of resistors and capacitors, one could build a tiny amplifier that once plugged to a 12" guitar speaker could make some serious rock'n'roll!
This is a fantastic read. At the risk of being the guy that needs the joke explained, I can't make sense of following paragraph:
> While the -6 volt supply spec caused little trouble for
the LM383, it meant that all the other electronics in a car needed
to voltage regulator that could provide protection.
The other components need protection anyway. What is the connection to
the LM383?
> Some one made the joke that the lateral PNP's were so bad that even delco would
not be able to destroy them.
Why are bad lateral PNP's hard to destroy?
Is he talking about PNP's inside the LM383 or inside other power supply circuitry that's
supposed to provide protection?
>A skunk works layout of a lateral PNP regulator ultimately made its way to
delco. And sure enough, they could not destroy it. They loved it.
And National Semiconductor got into the business of making PNP output regulators.
I think the -6V connection was simply that the LM383 designers were able to trivially integrate something the automotive industry considered non-trivial due to historical reliance on discrete processes and "build it sturdy" engineering. Nowadays chips are so highly integrated, so cheap, and so ubiquitous that it's hard to imagine dealing with messy power any other way, but it wasn't always so.
I have vague intuition that p-type silicon has lower mobility and worse ohmic-junction performance than n-type silicon. A real semiconductor engineer should verify, but I believe the core of the dig is that the low performance acts like intrinsic protection: the PNPs simply couldn't conduct enough current to destroy themselves, at least not in the multitude of spectacular fashions as the NPNs.
Electrons are the majority charge carriers for NPN vs holes in PNP which does mean that NPN transistors are faster, they also are usually a _bit_ better than PNP in other specs (beta, etc).
However I don't think this was the main part of the joke, rather it's that lateral transistors are _way_ worse in pretty much every conceivable metric than vertical transistors save two - they are hard to destroy and cheap. So add 'lateral' to 'PNP' and you basically have 'worst possible transistor', so bad that you can't kill it.
Thanks for clarifying. I chalked up the difference between 'a bit' and 'can't kill it' to mitigations that have happened since the LM383 was designed but now that I'm reading up on lateral transistors it looks like you're almost certainly right.
When I read things like this it amazes me that computers work at all. As person who spends 99% of the time working on software, I so much take all of the EE level for granted.
Well, a computer is basically a handful of gates, instantiated millions of times. So you only have to optimize those basic gates, and because they are (supposed to be) small, the design-space is small as well which simplifies optimization.
The hard part is actually the parasitics that come into play (capacitive/inductive coupling between signal lines and the bulk), and dealing with that when synthesizing the design.
If this is intriguing, Shahriar from The Signal Path does know how this stuff works and does some amazing teardowns and repairs of six (and likely seven) figure RF equipment.
It’s not really. People like everyone to think it is.
The difference is the building blocks of circuits are different due to function and most of the time you are interested in power transfer (which requires different impedance matching relationship).
So the new LM386 is the Nsiway NS8002. This chip is better in that the output capacitor it not required (it has two LM386s making an h-bridge). On lcsc.com, they are $0.0329 each for QTY=10, or $0.90 for boards (each) on eBay. They are nice if you don't want to deal with class-D switching noise.
Cheap clock radios use them. They also use an 8-pin AM/FM digital radio IC: pins for 32 KHz crystal, I2C, antenna (loopstick for AM), and audio out. No coils, inductors or filters needed.
I wonder if I can get my hands on the radio IC - I'd love to do an "overkill breadboard" radio - connected up to a bluetooth transmitter or USB microphone so I could have it dump into a Raspberry Pi for LAN streaming.
"My neighbor had just got a dog that was
always barking. So I got tweeter, built
up a simple 30kHz oscillator using a LM324.
Set the two LM383 to swing rail to rail.
Got a lantern battery that could put out
the power. Then connected the tweeter."
Paging Steve Gibson: somebody reinvented your dog silencer[1][2]!
I love the part about how Chrysler somehow managed to defeat the thermal protection and destroy the chip via a heatsink that had just the right amount of thermal resistance to cause the transistor to heat up above its safe operating area and cook itself whilst keeping the thermal protection circuit cool enough not to trip.
I love the design (or lack thereof) of this site. Courier font plus the old scans of photocopies make it feel like i'm reading an old xeroxed white paper (in a good way).
I like it as well and find it delightfully easy on the eyes.
Funny thing is that I remember that back in the day I hated those Courier-font-style publications. They always seemed hard to read compared to well type-set documents. And I never understood screenplay writers obsession with Courier 12[1].
I guess the web is so noisy and cluttered today, that even harsh black Courier on a white page is a relief.
[1] Screenplays are usually written in Courier 12. I only know that because I sometimes read screenplays, for movies that interest me. It's fun and doesn't take as much time as it does to watch the movie.
>I only know that because I sometimes read screenplays, for movies that interest me. It's fun and doesn't take as much time as it does to watch the movie.
I don't know why I find this as hilarious as I do, but thanks for sharing.
Personally, I haven't read many screenplays, though I did read the original one for Monty Python and the Holy Grail and it bears very little resemblance to the final film. There's a whole section about buying pet ants from Harrods.
Courier 12 is my go to for cutting and pasting terminal output because so much depends on fixed width fonts lining up properly and Outlook doesn't offer an exact match of xterm's 9x15 font.
The package redesign remark is only directly relevant to NatSemi TO220-5 parts. There are various other TO220 packaged ICs (and discrete devices) made by other manufacturers which have exactly the described failure behavior.
this week I’ve been looking into making a wireless speaker similar to a Sonos play:1. In terms of the streaming mechanism is was going to use a raspberry pi running as a snapclient, wired to a class-d amp board and a, say, 3” driver. I thought I would find a load of projects using a pi + amp + driver in a small box combination, but I must be looking in the wrong place.
Does anyone have any top tips of examples of setting something like this up? What amp / driver combination to use? I’m not looking for anything amazing, and I’m sure round 1 will be a learning experience, but any help appreciated.
He mostly just lays out the speaker and crossover design; but as he mentions a search for "TPA3116 2.1 amp" will give you lots of results on Amazon, AliExpress, and Parts Express.
PE's forums are good, lots of people have tackled similar projects:
The product page for the C-Notes (a fantastic value at $100/pair!) kit has a build video which gives you a great idea of the process and the tools you'll need. Process is similar for any flat-pack kit.
There's about $80 worth of drivers in there, which is bonkers for a $100 kit that also includes crossover parts and MDF cabinets.
For comparison, the $130 Overnight Sensations kit contains about $55 worth of drivers. That's a good deal, and consistent with other DIY kits... and it illustrates what an outlier the C-Notes are.
(The reason why the C-Notes are so affordable is because Parts Express and Dayton share corporate ownership, so presumably they're about to source those Dayton drivers for next to nothing)
FYI there's an MTM (two woofers, one tweeter) version of the C-Notes coming in ~45 days for $100/speaker ($200/pair)
I have built my own wireless bookshelf speakers using a combination of GE5654 Tubes, TPA3116D2 [0][1] and customized woofers and tweeters (Rosewood finish) [2]. I have to say, it really came out much better than I expected. [3]
Here's how I did it:
Last year, I visited Taiwan and China, I basically went to scout for speaker suppliers. I shortlisted a couple, but I realized, the popular they are, the pricier they are - regardless of the quality. Then, I realized something shocking. Actually, most speaker manufacturing is done in Nanjing (and one more location starting in N, forgot). These factories just stick the brand names on top and license it to these "big brands". And you pay a premium for that. So, I visited one direct supplier in Nanjing. It was a very small shop. I just walked in, placed an order for a pair. The whole set of 4 speakers (2 woofers and 2 mid-ranges) costed me in total of $200 US roughly. I purchased two 6.5" aluminium cone woofers and 2 golden capped full ranges. I like woofers with large Xmax, so I bought these for bass units.
You can purchase from AliExpress too, although it's slightly pricier. But most speakers from China, contrary to popular belief, are very very high quality. In fact, the consumer audio space such as Sony, Samsung, etc. use much lower grades than what's normally found on AliExpress.
As for amplifier, buy a half-decent board. Strip out all the cheap components, spend 3x the price of the board on really good quality components. For the TPA3116D2 board, this place is a good start [4]
There are many projects of bluetooth amplified boxes using class-D modules including BT receivers. Searching for "diy bluetooth speaker box" on Google Images returns some interesting results. You just need then to find the suitable amplifier without BT and stick the RPi inside the box.
A word of advice though: the Raspberries have a sub par audio output which isn't good for music, so you may want to consider either a different SBC or an external USB sound card, as even the cheapest will sound a lot better than the RPi on board one.
I wish I had half a million bucks to dump into Snapclient - it's so close to being good, but needs some automated way to get the multi-speaker delay down to a good level.
In regards to your project - it depends how many sub-projects you want to spawn. I've got https://www.ebay.com.au/sch/i.html?_from=R40&_trksid=m570.l1... hooked up to my PC and some old speakers - 3.5mm jack, USB DAC and Bluetooth support. It's variable voltage up to 12V (I think). Funnily, if you supply it with 5~7 volts, you'll get a "low battery" audio cue - despite there being no battery connector anywhere.
The newer version of this has the volume knob in a silly place.
I worked on almost the exact same idea for my senior design and the audio output from the dac built into the pi is terrible. The HiFiBerry board was a life saver.
I've had a lot of luck (good and bad) with the ESP8266 wifi module for little applications similar. I'm interested in trying something like ESP8266Audio[0]
As a kid, the LM383 had particular appeal: it was the highest-power amp chip available that ran off a 'safe' voltage. I didn't need to do any mains wiring.
> "But it is humorous when the TDA2002 is often referred to as a replacement for the LM383. As a matter of fact, the LM383 is really a reverse engineering of the TDA2002. Credit should be given to the design engineer Bruno Murari who was working for SGS at the time. He was the one who came up with a way to do an automotive power amplifier using only five pins."
This first paragraph is almost gibberish to me. In a good way. I just find it humorous that there are so many in-the-know words here.
TDA200 and LM383 are obviously product or device names. Bruno Murari is obviously a person's name (a design engineer, as the sentence points out!), SGS is obviously a place where one works, likely a company.
You should have no issues with the term "reverse engineering" if you are on Hacker News. So that only leaves "automotive power amplifier using only five pins". You probably know what "automotive" and "five pins" means, though, so even less than that.
And then it's absolutely obvious that you won't know what a "power amplifier" is if you don't have some connection to electrical engineering, but it's useless to point that out. Imagine if every electrical engineer did that on articles on biology, chemistry or economic sciences.
These are names of things. The sentence structure makes that clear, and makes it clear how they relate. They are not "in-the-know" terms that are lacking definitions. It is not gibberish, just like the sentence "Jane gave Sue a letter from Bob" is not gibberish, even though nobody defined "Jane", "Sue", or "Bob".
I don't do any analog work, so I wasn't familiar with the LM383 but it made perfect sense to me.
From the article, I knew the LM383 was a 5 lead audio amp. From context in the sentence it was clear that the TDA2002 is a predecessor of the LM383. I don't know who Bruno Murari or SGS are, but again, context makes it clear that Murari worked at SGS when he designed the TDA2002.
On a related note, I heard this 2nd hand from a guy who was talking to CAT about a module they wanted to have 80V over voltage protection for a crazy amount of time (minutes). My friend asked what scenario required that and it was something like this: Guy at a construction site can't start his bulldozer. He doesn't have jumper cables, but he's got an arc welder... I don't know if that actually happened but it would not surprise me.