|
Forseti posted:I have a quick question that I think is easy to answer, and probably is answered earlier in the text book but I'm jumping around and don't know where exactly to check. I'm looking at the section on oscillators and reading about the Colpitts oscillator, which they are showing with this diagram The grounded tap between C1 and C2 can be thought of as setting up a ratio between the inductor voltage and the base voltage. The fact that the tap, not one of the ends, is grounded is important since the filter portion of an oscillator built around an inverting amplifier needs to invert its input at the resonant frequency in order to oscillate.
|
#
?
Jul 27, 2021 02:11
|
|
|
|
| # ? May 28, 2024 17:57 |
|
|
Think of the two caps as a voltage divider. It's AC, they are of equal impedance, so imagine 12VAC across L, you're now grounding the 6VAC point of that. Just like a transformer's centre tap
|
|
#
?
Jul 27, 2021 03:22
|
|
|
poeticoddity posted:I once accidentally damaged an 18650 cell in such a way that fire came out while I was holding it, so I'm Nth-ing the suggestion of, "Do your experiments outdoors," and adding "away from flammables" and "have a fire extinguisher handy". 4 words: Explosion containment pie dish. I don't screw around with lithium cells myself, but i often charge grey import/ebay/aliexpress stuff with batteries in a cookie tin.
|
|
#
?
Jul 27, 2021 03:40
|
|
|
LimaBiker posted:4 words: Explosion containment pie dish. I actually have a fireproof battery box specifically for charging lipos in, which I totally forgot about until now.
|
|
#
?
Jul 27, 2021 10:00
|
|
|
poeticoddity posted:I once accidentally damaged an 18650 cell in such a way that fire came out while I was holding it, so I'm Nth-ing the suggestion of, "Do your experiments outdoors," and adding "away from flammables" and "have a fire extinguisher handy". "A fire extinguisher" isn't necessarily a good plan. Depends on the specific type of Li chemistry. For a lithium metal battery, only a class D extinguisher will help. For Li-ion, class ABC and BC extinguishers work. I've also read the dousing Li-ion in water is an option, but take that with a grain of salt. Personally, I just do my experiments with the batteries inside a fire safe enclosure/cabinet. If something goes wrong, I shut the lid and let the process work itself out. Since we were discussing thermal cameras recently, I'll also point out that a thermal camera, even a lovely one, is very useful for this. Gives you advanced notice of whether a battery is heading towards thermal runaway or not.
|
|
#
?
Jul 27, 2021 14:38
|
|
|
Forseti posted:
Fun fact: back when BJTs were hot new technology, PNP transistors were far easier to manufacture, so most circuits used them as much as possible. As for the negative supply voltage convention, this is just conjecture on my part: For analog circuits, you want your signals to be referenced to GND in order to avoid power supply ripple/noise from contaminating everything. Therefore your amplifiers should be referenced to GND as well. If you're using PNPs in common emitter amplifiers, the emitters must face GND. So the collector must be biased below GND, hence the negative supply voltage.
|
|
#
?
Jul 27, 2021 14:47
|
|
|
The fire extinguisher is for the things surrounding the battery, because the battery itself is just going to dump all its energy no matter what kind of stuff you spray at it. Like on airplanes, the procedure for dealing with them is throw them in a fireproof box and close it and just let them do their thing in there where they can't catch the rest of the plane on fire. That's why you can't put lithium ion batteries in checked baggage, the flight attendants need to be able to grab it and chuck it in the box.
|
|
#
?
Jul 27, 2021 14:48
|
|
|
ANIME AKBAR posted:Fun fact: back when BJTs were hot new technology, PNP transistors were far easier to manufacture, so most circuits used them as much as possible. You also needed to bias them way the hell up near their breakdown voltage to get them to work worth a drat, which is why early transistor radios needed like 20-30V batteries.
|
|
#
?
Jul 27, 2021 14:52
|
|
|
Thanks guys, your input is very helpful. I'm still not quite sure I've grasped it yet, but I have that feeling I've had in the past with other things where I feel like I'm just on the cusp and my mind just needs some synthesis time for it to click into place. This stuff is tricky IMO, but it's already making a lot of the graphs on various data sheets make a whole lot more sense to me, so definitely worthwhile for that alone to me
|
|
#
?
Jul 27, 2021 15:27
|
|
|
Shame Boy posted:You also needed to bias them way the hell up near their breakdown voltage to get them to work worth a drat, which is why early transistor radios needed like 20-30V batteries. I love learning about the old school radio designs and seeing the circuits they came up with based on the constraints at the time. My favorite so far has been the reflex receiver (http://techlib.com/electronics/reflex.htm) which uses the same transistor to amplify both the RF and the audio signals. It makes sense looking at what they did, but that would have never occurred to me to try in these days of transistors costing a few pennies even at hobbyist quantities
|
|
#
?
Jul 27, 2021 17:10
|
|
|
Struggling again with power saving and op amps. I received and setup up boards that use an analog switch to cut off power to the opamps. It appears from experimentation that it takes about 2 seconds (!) after power is enabled for readings to stabilize. Eg, if I close the switch (Conntecting the amp to its power source), take readings, and open the switch, the readings are bogus. If I wait 2-3 seconds, they're fine. If I wait less than that, the readings are bogus. How would you troubleshoot this? Of note, I found some low-quiescent current opamps (eg Analog Devices LTC6259 or LTC2064. The latter has 2uA supply current! That's 500x less than the current ones I'm using. None of these come in SOIC8, which is the current footprint, so I can't just drop them in) that use uA power when active, and have a shutdown pin to reduce consumption further, but I'm hoping to get this particular design working. Circuit:  I don't see anything relevant in the opamp's datasheet. Why I'm not just respinning the boards with the low-power amps: #1: I need to ship this; need to arrest these perpetual 2 week delays (And would prefer not to eat the cost on this batch). #2: I'm not sure the low power amps won't also suffer from this, or something else; need to test. edit: I think this is my answer. Short-term drop-in: AD8617 Long-term replacement: LTC2064 And from reviewing AD's product catalog, it seems like SOIC-8 is a dated package, and DFN is the new hotness for amps. Dominoes fucked around with this message at 01:26 on Jul 28, 2021 |
|
#
?
Jul 27, 2021 22:21
|
|
|
Speaking of batteries, I think I found a life hack for getting 18650 for a decent price: cheap flashlights on Amazon. For twenty bucks you can order a lovely flashlight and get 4 to 6 ok enough LI-ION batteries with it. They're Chinese no-names, but I if you scroll through the comments there's usually several reviews of people who have drain and charge tested the batteries to see if they are decent or not.
|
|
#
?
Jul 28, 2021 03:14
|
|
|
Is there a place where I can get 1x1 female "dupont" connectors? The kind like the black headers on the wires in this picture? There's a whole bunch of kits on Amazon and stuff, but a lot of them look real shady and have reviews where people say the parts don't fit together. I'm looking for a kit to get a whole bunch of the 1x1 enclosures and metal pins that I have to crimp on the wires myself. Also, I was wondering if an Engineer PA-09 tool can crimp these kinds of connectors? I looks like the answer is yes, but I wanted to make sure.
|
|
#
?
Jul 28, 2021 06:38
|
|
|
Just get a female/female DuPont and cut one end off. I’ve got a bunch of them from Amazon with no issue.
|
|
#
?
Jul 28, 2021 06:51
|
|
|
Ah, that should work for now... Thanks!
|
|
#
?
Jul 28, 2021 07:12
|
|
|
Be careful, though. I've done that before and found that the creator used wire with shockingly-little metal in it and much thicker than average insulation. So you may need to buy a few different lots from different people just to maximize the chance you don't get bootleg-rear end wire.
|
|
#
?
Jul 28, 2021 07:34
|
|
|
BattleMaster posted:Be careful, though. I've done that before and found that the creator used wire with shockingly-little metal in it and much thicker than average insulation. So you may need to buy a few different lots from different people just to maximize the chance you don't get bootleg-rear end wire. Concur. I was using some as hookups for LED strings. I had a set that turned into fuses at about 800mA, which I didn't think was all that much current at all. No smoke, though, because the insulation expanded to about 4x diameter then was hot-knifed away. Got a good look at the LEW as the LEDs dimmed from lack of current before the wire parted.
|
|
#
?
Jul 28, 2021 09:49
|
|
|
Hah. That's because most of those lovely wires are steel. Like, stick to a magnet and everything. You wouldn't think that copper wire is too expensive for them to use but here we are
|
|
#
?
Jul 28, 2021 14:18
|
|
|
-
|
|
#
?
Jul 28, 2021 14:18
|
|
|
Dominoes posted:Struggling again with power saving and op amps. I received and setup up boards that use an analog switch to cut off power to the opamps. It appears from experimentation that it takes about 2 seconds (!) after power is enabled for readings to stabilize. Eg, if I close the switch (Conntecting the amp to its power source), take readings, and open the switch, the readings are bogus. If I wait 2-3 seconds, they're fine. If I wait less than that, the readings are bogus. How would you troubleshoot this? I am far from an expert, but I thought I remembered that settling time is one of the disadvantages of single supply op amps. Quick google search brought up this app note which you may have already seen https://www.analog.com/en/analog-dialogue/articles/avoiding-op-amp-instability-problems.html The very last part may be something to look into: quote:In applications where the circuit's turn-on time may become excessively long, a Zener or active biasing method may be a better choice. Have you looked at it on a scope? Does it bounce around before settling or is it slowly rising to the correct reading?
|
|
#
?
Jul 28, 2021 14:23
|
|
|
By the way, I think I came up with a pretty elegant solution to the problem of "weirdly-shaped holes" I mentioned a few weeks ago: I just got an acrylic front panel laser cut and etched by Ponoko, that goes in front of the metal case front panel. This tackles a bunch of things at once:
I also got some acrylic paint pens (Posca ones, I read somewhere they work real well and was not let down) and painted over all the etched areas, then buffed off the overpaint (it doesn't stick to the smooth areas so you just scrub it with a microfiber rag and it comes right off) to make all the text and lines really pop, and I think it came out fuckin' fantastic. You can't even tell how much I butchered the actual metal case  The only thing I need to change is that big 10mm "Status" LED, it's supposed to actually poke out a bit so it can be more visible from off-angles, but I soldered it too close to the board. Easy enough fix, just gotta get around to it.
|
|
#
?
Jul 28, 2021 15:06
|
|
|
Looks great!
|
|
#
?
Jul 28, 2021 15:10
|
|
|
The key is to keep people from just randomly turning on the thing with the big zappy capacitors inside it. The toggle switch explicitly arms the interlock circuit, which is a high voltage relay that's normally closed and by default connects a 1Meg resistor across the output caps. Keeping that kind of load on the power supply all the time would just kinda bake the transistors in the power supply and lead to higher ripple, so when in use the relay is energized to disconnect the discharge resistors. However, if you slap that red cover down (or remove the key, or open the lid), the power to the high voltage circuitry is cut and the interlock relay is released, slapping that resistor across the output caps and taking them down to safe levels very quickly. The important industrial button is designed in such a way to make it very hard to accidentally press, which is good cuz it turns on and off the output. They are definitely real important parts and not at all things I wanted to add to make it feel like I was starting up a nuclear reactor whenever I turn it on. 
|
|
#
?
Jul 28, 2021 15:15
|
|
|
Dominoes posted:Circuit: What madman would make a dual opamp symbol like that? I don't suppose your issue is the delay in charging C8, this is a fairly obvious source of issues. The network should charge in around ~300 ms from what I can tell but you may have some other networks like that as well? Do you actually legitimately need such a long time constant there? If you want to make it start up a fair bit faster, add a second capacitor mirroring C8 but to VCC, though at that point removing the capacitor altogether may accomplish the same thing. Incidentally if you're only switching 3.3 V supplies you could possibly use a logic gate instead to save money/space? Not sure about quiescent current on those (never had to care about it), but e.g. LCX/LVC gates can supply decent current at least. One key difference is that CMOS gate actively drive the output high or low, whereas your analog switch only drives the output high and will float low. This may cause unexpected issues due to sneak paths in the powered off circuitry.
|
|
#
?
Jul 28, 2021 15:18
|
|
|
He did request fixes that don't require a board spin but gently caress it: it would be more efficient to switch the ground in and out using an N-channel MOSFET, instead of a mux switching power.
|
|
#
?
Jul 28, 2021 15:28
|
|
|
ante posted:He did request fixes that don't require a board spin Could an open drain output off the MCU be used here? Could just be a bodge wire I think longview posted:What madman would make a dual opamp symbol like that?  I thought the same thing, that schematic symbol is nuts! Usually you get two separate op amp symbols for the schematic from a dual package op amp.
|
|
#
?
Jul 28, 2021 15:32
|
|
|
So as noted that 1 uF capacitor needs to charge up through 110k ohms every time you power the op amp on. It's possible that the fact that the 1/2 3.3V reference takes so long to stabilize is why the output takes time to stabilize. I figure one of the two options may help: 1. Connect that divider to the normal 3.3V rail. It will draw 200k ohms worth of current constantly but the capacitor won't have to charge up more than once. 2. Lower both the resistor values and the capacitor values. It will draw more current in operation but charge up faster. The on resistance of that mux is 2 ohms and its current rating is 30 mA so I doubt that the mux is a significant factor.
|
|
#
?
Jul 28, 2021 16:01
|
|
|
Forseti posted:I am far from an expert, but I thought I remembered that settling time is one of the disadvantages of single supply op amps. Quick google search brought up this app note which you may have already seen longview posted:What madman would make a dual opamp symbol like that? I didn't think of the floating low part! Not sure what consequences that is having. Forseti posted:Could an open drain output off the MCU be used here? Could just be a bodge wire I think I thought the same thing, that schematic symbol is nuts! Usually you get two separate op amp symbols for the schematic from a dual package op amp.[/quote] Hah. Yea it's not intuitive or clear as a single, and is messy. When I make split part symbols in KiCad, the section between turns into a deadzone where any operation inside causes KiCad to ask me if I'm referring to the thing I clicked, or the pair of amps surrounding it. I made a quad amp symbol at one point in this approach, and changing things was a mess.
|
|
#
?
Jul 28, 2021 16:04
|
|
|
I was going to do an edit, but since you posted: 3. WILD AND CRAZY OPTION: Use a zener diode in an appropriate package in place of R12. If there is no such thing as a 1.65 volt zener, use the closest option and adjust the firmware to use a different zero point in the calculations??? Also reduce the value of its companion resistor and capacitor. edit: I misread IN A as EN A because of the crossing of the net wire with the text, so disregard that comment if you saw it before my edit BattleMaster fucked around with this message at 16:11 on Jul 28, 2021 |
|
#
?
Jul 28, 2021 16:08
|
|
|
Dominoes posted:This begs the question of once I replace with a low-power amp that has a shutdown pin in the next iteration, do I use the shutdown pin, or just leave it on given the low supply current? No reason not to try it, and either solution would be acceptable. I assumed that the low quiescent current meant you were going to leave it on all the time. It would depend on the specific part, but I think in general low power amplifiers are even slower to settle? I always think of it in terms of CPU overclocking. To really push the limits, the voltage needs to be turned up which makes the rise times faster and improves stability at higher clock rates. In that case though, it's obviously otherwise the exact same part. It is certainly possible I think that the low power op amp could be lower power and still achieve the same rise and fall times as your current op amp.
|
|
#
?
Jul 28, 2021 16:31
|
|
|
BattleMaster posted:1. Connect that divider to the normal 3.3V rail. It will draw 200k ohms worth of current constantly but the capacitor won't have to charge up more than once. 1) No. While the exact opamp isn't listed in the schematic, one thing basically all opamps have in common is: input voltage can not exceed the supply voltage beyond +-0.3 or 0.6 V depending on the opamp process. Since the supply voltage is either 3.3 V or 0 V, the inputs can't be powered by something unswitched. It is possible to use a second analog switch to isolate the divider and opamp but that's silly IMO. 2) the time constant is the crucial part, the exact same effect can be had by reducing the capacitance as you get from reducing the resistor values. The capacitor can possibly be reduced all the way to 0 for that matter, this would give close to 0 startup time. I would put somewhere around 1 nF in there instead, mainly to "guard" against nearby trace capacitance. In case anyone's not familiar with circuit analysis that circuit is equivalent to a simple RC circuit with a 60k series resistor, 1 µF capacitor, connected to half the supply voltage. The time constant is trivially calculated from that. Note also that the exponential charge up means that you're potentially waiting 10 or more RC time constants for "proper" settling, depending on what the application is exactly. I would also note that R13 seems to have no purpose, though it isn't causing any issues being there. Dominoes posted:I didn't think of the floating low part! Not sure what consequences that is having. Driving the switched supply to ground may highlight other design issues you have; leakage between unswitched and switched circuitry will often result in the switched 3.3 V sitting at some intermediate voltage. E.g. if you were to power the divider off normal 3.3 V, some current would flow through the opamp input protection diodes into the 3.3 V switched supply, charging it up to some voltage. Depending on circumstances this will often end up stabilizing at some moderate leakage level, depending on the loading on the switched supply. I.e. it will often end up stabilizing to a voltage where the minimum leakage current flows. If you drive the switched supply to 0, this stabilization won't happen since the 3.3 V switched voltage will be forced to 0 V, potentially increasing idle consumption. Now, the fact that this happens is typically caused by design error and you should try to avoid designs with this type of leakage, but some leakage is typically always there no matter how correct the design is. In some circles this is called a sneak path, and they can cause significant issues in some cases.
|
|
#
?
Jul 28, 2021 16:34
|
|
|
longview posted:1) No. I should have known that, oops. I guess I've never switched power to an op amp before so it was never on my mind. Well, if the op amp has protection diodes on the inputs then they should be able to sink the small current through the high value resistors just fine, but I wouldn't want to rely on that. And I guess you'd risk powering the circuit through the divider. edit: I discovered that last bit when I forgot to hook up the VCC to a part but it still worked anyway, and thankfully the protection diodes on the I/O it was being powered through could handle enough current that it wasn't an issue before I noticed it! longview posted:Driving the switched supply to ground may highlight other design issues you have; leakage between unswitched and switched circuitry will often result in the switched 3.3 V sitting at some intermediate voltage. E.g. if you were to power the divider off normal 3.3 V, some current would flow through the opamp input protection diodes into the 3.3 V switched supply, charging it up to some voltage. So what about putting a resistor between the op amp VCC and ground of such a value that it can give the leakage current a path to ground with a small enough voltage drop that things effectively turn off when the mux is off? I guess it would have to also discharge the bypass caps to turn everything off properly though. BattleMaster fucked around with this message at 16:49 on Jul 28, 2021 |
|
#
?
Jul 28, 2021 16:39
|
|
|
longview posted:In case anyone's not familiar with circuit analysis that circuit is equivalent to a simple RC circuit with a 60k series resistor, 1 µF capacitor, connected to half the supply voltage. The time constant is trivially calculated from that. Note also that the exponential charge up means that you're potentially waiting 10 or more RC time constants for "proper" settling, depending on what the application is exactly. I did the math earlier and came up with a time constant far shorter than what he was observing the settling time to be and figured that wasn't it. I totally didn't think about that since this is a measuring device, it probably needs to wait for well after the cutoff frequency to be sufficiently settled 
|
|
#
?
Jul 28, 2021 16:50
|
|
|
BattleMaster posted:I should have known that, oops. I guess I've never switched power to an op amp before so it was never on my mind. No worries, designing for multiple power rails like this is kind of its own mind-set that I've seen very experienced EEs mess up. I've found it very helpful to analyze circuits (coarsely, not with mathematical rigor) for the case of: Supply off, instantly turns on Supply on, instantly turns off (i.e. shorted) And variations on this where multiple supplies are involved. This often helps catch fuckups before they break stuff, a lot of designs ignore what happens if the input supply voltage is suddenly shorted. This can often break power supply regulators, for example. Hence why many old linear regulators specified a reverse diode across the input/output, to catch this reverse current. Re relying on input protection diodes: At least one of the major manufacturers precision opamps is specified for input protection diode current. In the last two years I have seen around 5 different cases of broken opamps of one specific make, going up from ~0 the last 10 years. One thing all of the circuits had in common was that they relied on the protection diodes for at least one operational mode, and the current was always below the specified maximum. In multi-package versions of this amp, the bad one was always the one using the protection diodes. Now this was a precision bipolar opamp, which used input differential voltage protection and not a standard CMOS type, but I don't trust opamps to be reliable when those diodes are used regularly. In one case the opamp decided it wanted to short out a 60 A+ rated power supply, that left a decent mark on the board!
|
|
#
?
Jul 28, 2021 16:55
|
|
|
I have a breadboard with a DIP-28-shaped melted spot in the middle of it. I didn't build an ICD header into a USB device design I made (imo dumb mistake I'll never repeat, because lol at writing from-scratch USB device firmware with only the host-side debugger) so to program it I popped the chip out and put it into a breadboard that was connected to the ICD via a breadboard cable. It worked well for several cycles until one time the ICD couldn't find the micro. I had no clue what was wrong until I smelled burning plastic and then burned my finger touching the top of the IC. It turns out I socketed it in one set of pins over, so ground was connected to VCC and one of the I/O pins was connected to ground, which caused a short through a protection diode. Of course, it was limited to the couple hundred milliamps that the Microchip ICD2 could source so it wasn't too catastrophic. Because Microchip makes their micros to last, the micro worked fine after I let it cool down. The I/O pin that was shorted didn't work but every other function of the micro seemed to. It wasn't a pin I needed for the current application so I never even discovered that until much later when I needed it (it was one of the SPI pins.)
|
|
#
?
Jul 28, 2021 17:03
|
|
|
I successfully ran a PIC at 12V recently, accidentally. It ran code, blinked my LED and everything. For three or four seconds, then never worked again. Still incredibly impressive.
|
|
#
?
Jul 28, 2021 17:12
|
|
|
I had one case where 1/4 of a quad opamp was completely open, like the output bond wire broke. This caused a lot of confusion and improper operation since it was an inverting amplifier circuit driving a high impedance load. Same opamp type as I mentioned above. With the opamp broken, the "amp" was effectively non-inverting. Since this was a moderately high power laser power control circuit it was a fairly annoying case of either getting 500% power or no power. Fortunately I decided to power it up initially into a power meter so I could see immediately that it was broken. Ended up gluing an AD8605 opamp on top of the old amp and dead-bug wiring in a new one in parallel. Last time I touched a PIC it was a 16F series that seemingly decided to stop living, it was a weird one that I've seen a few other cases of in similar products. Part was completely dead except for responding to the hardware interrupt line (i.e. wakeup from lowest power sleep), which made it power on the oscillator for 1-2 seconds. The weird part was I could read the flash data off it, and reprogramming it into a new part gave the same behaviour. Then I hosed around with the config fuse settings, setting them to other values then back, and then it came back to life after the ~5th reprogramming. I've also seen other PIC16s just stop working years after manufacture, though that's the only one where someone paid me to bring it back to life.
|
|
#
?
Jul 28, 2021 17:36
|
|
|
For your amusement, crossposting this (it's a 1965 stereo receiver): petit choux posted:.... and I did go back for it, in part just because it looked so nice on the outside. But when you open it up it's not so pretty ...
|
|
#
?
Jul 28, 2021 21:27
|
|
|
Yeah those look like transistors in a TO-3 package. Judging from the lack of big tubes and audio transformers those are probably the final stage of the audio amplifier.
|
|
#
?
Jul 28, 2021 21:58
|
|
|
|
| # ? May 28, 2024 17:57 |
|
|
Shame Boy posted:I just got an acrylic front panel laser cut and etched by Ponoko, that goes in front of the metal case front panel. This tackles a bunch of things at once: That does look good. If you don't mind sharing, what did that cost?
|
|
#
?
Jul 28, 2021 22:50
|
|