That's a good explanation. Covers all the basics, including mentioning that FETs fail in the ON direction. This is why switching power supplies badly need protection circuitry, and why you don't want wall warts and battery chargers that didn't pass UL testing.
He mentions electrostatic discharge sensitivity. Yes. It's very easy to blow out the gate of an FET just by touching the gate lead. Once it's attached to a driver transistor, it's protected, but loose FETs need to be stored in conductive foam and handled while wearing a wrist strap.
FETs are really amazing. ON resistances in the milli-ohm range, and OFF resistances in the megohm range. It's surprising that's physically possible.
Proper galvanic isolation is most of what protects you from mains. The YouTube channel Diode Gone Wild does a lot of charger teardowns explaining what makes them safe or dangerous.
Unfortunately, we often rely on trusting secretive vendors with insulation specs, trusting assemblers with following instructions and not damaging hard to inspect wiring and insulation, and trusting brand and certification markings on the product since you can't open it without destroying it.
> Speaking of the Pinecil, it uses an unorthodox driving circuit – it has an NPN transistor, but its base driven through a capacitor, so that only the AC component of the driving signal gets through. As a result, if the main MCU hangs and the control GPIO is stuck high, the FET won’t stay enabled!
The Pinecil implementation is essentially a watchdog, it just operates proactively instead of reactively. The end result is the same but relies on simple hardware. See also: dead man's switch [1] used in trains and other safety critical infrastructure.
If the end result is the same you're just wasting components that already are builtin into most microcontroller.
And you'd need to remember to not use autonomous part of microcontroller (like a timer output) to implement it.
Considering the connected net is named "tip_control" I'd assume its PWM that is fed into MOSFET controlling the heater... and in vast majority of cases that PWM will happily run when core is hanged.
Looking at datasheet of the microcontroller PA6 does indeed go to tip_control and one of alternate functions is TIMER2_CH0, and at least glancing at code
/* enable TIMER1 - PWM control timing*/
timer_enable(TIMER1);
timer_enable(TIMER2);
They need both. Watchdog can't help if the GPIO hardware is stuck high or if the MCU is browned out and doing undefined behavior. The npn can't help if the code is stuck.
Jfets are used a lot in guitar pedals (j201, 2n5457, 2sk30a, ...). Unlike MOSFET they only need one gate resistor (10M gate to ground) in a input buffer configuration. The drain and source resistors are same as MOSFETs.
I feel like guitar pedals are obsolete because of modeling amps now. Why bother messing around with a bunch of separate pedals with 9V batteries and interconnect cables when you can just program the effects you want all in software (using your phone, no less), and instantly switch between completely different configurations (sound profiles) by pressing a button?
Yes, with some qualifications. Below, with some edits, is what I originally wrote for the comment you replied to, before scratching it for being too wordy.
Some personally compelling reasons for myself as a guitarist:
1) With physical gear, I own the means of sound production. Less so for pedals that integrate closed computing elements into their circuits though, and even less if it has IOT crap (why??). Analog devices are generally best at this aspect.
2) Along those lines, longevity and repairability. Again this applies most strongly to analog pedals. I can keep using a well-designed pedal for decades, and keep it going with occasional repairs and maintenance. Digital modelers (including digital pedals to an extent), not so much, for a number of reasons.
3) In general, physical signal chain elements, particularly analog ones, strike me as more malleable as an end-user. I can do tweak the gain of individual stages, rework the inter-stage EQ and attenuation, etc. with a physical tube amp, and do similar things with analog pedals. Perhaps to a lesser extent with digital pedals too. I'm not aware of a modeling system with comparable flexibility, though I'd be interested in being proven wrong.
4) Some analog stuff, to my limited knowledge, is still difficult to model.
>2) Along those lines, longevity and repairability.
Why is this an issue? When was the last time you had digital electronics fail, and it wasn't a victim of the early 2000s capacitor plague (in which case, replacing the affected caps fixed it)? "Repairability" is vastly overrated: most electronics don't fail in the first place, and when they do, it's almost always electrolytic caps, and this affects both digital and analog circuits. There's nothing "unrepairable" about a digital circuit: it's not like a logic gate on your CPU is likely to fail.
>and even less if it has IOT crap (why??).
If you're referring to amps (like Fender Mustang series) that use your phone for the UI, that should be pretty obvious: the phone is an ubiquitous device that you can assume the customer owns, and provides a large touchscreen for controlling the amp remotely. It's not IOT; it's a Bluetooth connection. It's not necessary for using the amp, it's just an option, and a really helpful one when playing with effects settings, and much easier to use than the tiny LCD screen on the amp plus its jog wheel. Once the settings are done, you can assign settings to footpedal buttons and never mess with it again. With the 7-button footpedal, for instance, you can easily set it up with 5 different sound profiles you can easily switch between, with the other buttons you can switch between "banks" of 5 effects if you need more for one song/gig.
Decision paralysis is a thing - while having digital multieffect might open you to near-unending combination of effects, exploring within limits of few single function (not neccesarily analog) pedals is less complex.
Also some say some quirks of analog designs are hard to replicate in simulation althought I haven't seen double blind tests of that so I'm mildly skeptical of that
True, and they work remarkably well as preamps. Only problem is that they're becoming scarce, at least in THT case, and a lot of people has been burned by online sellers who sent them relabeled BJTs. To my knowledge, J201 and many 2SK* low noise ones haven't been manufactured for years, most reliable sellers don't have them in stock anymore, and those sold online by generic sellers are almost invariably either relabeled random ones (if lucky) or the 1st thing they had at hand, which often is a BJT, not even a jFET.
However, if you don't need lots of gain a decent preamp can be built around the venerable 2N7000. Yes, that's a MOSFET, definitely not aimed at audio, but it's dirt cheap, available everywhere ...and noisy. But if you keep the gain to reasonable levels it works quite well.
22 comments
[ 3.0 ms ] story [ 57.3 ms ] threadHe mentions electrostatic discharge sensitivity. Yes. It's very easy to blow out the gate of an FET just by touching the gate lead. Once it's attached to a driver transistor, it's protected, but loose FETs need to be stored in conductive foam and handled while wearing a wrist strap.
FETs are really amazing. ON resistances in the milli-ohm range, and OFF resistances in the megohm range. It's surprising that's physically possible.
Unfortunately, we often rely on trusting secretive vendors with insulation specs, trusting assemblers with following instructions and not damaging hard to inspect wiring and insulation, and trusting brand and certification markings on the product since you can't open it without destroying it.
Clever!
[1] https://en.wikipedia.org/wiki/Dead_man%27s_switch
And you'd need to remember to not use autonomous part of microcontroller (like a timer output) to implement it.
Considering the connected net is named "tip_control" I'd assume its PWM that is fed into MOSFET controlling the heater... and in vast majority of cases that PWM will happily run when core is hanged.
Looking at datasheet of the microcontroller PA6 does indeed go to tip_control and one of alternate functions is TIMER2_CH0, and at least glancing at code
...https://github.com/Ralim/IronOS/blob/1fbcdcdf987a28541524e4f...
... timer_init(TIMER2, &timer_initpara);
...they are using it for PWM. The watchdog itself also appears to be enabled tho, so there is a chance it will kill itself the normal way
Some personally compelling reasons for myself as a guitarist:
1) With physical gear, I own the means of sound production. Less so for pedals that integrate closed computing elements into their circuits though, and even less if it has IOT crap (why??). Analog devices are generally best at this aspect.
2) Along those lines, longevity and repairability. Again this applies most strongly to analog pedals. I can keep using a well-designed pedal for decades, and keep it going with occasional repairs and maintenance. Digital modelers (including digital pedals to an extent), not so much, for a number of reasons.
3) In general, physical signal chain elements, particularly analog ones, strike me as more malleable as an end-user. I can do tweak the gain of individual stages, rework the inter-stage EQ and attenuation, etc. with a physical tube amp, and do similar things with analog pedals. Perhaps to a lesser extent with digital pedals too. I'm not aware of a modeling system with comparable flexibility, though I'd be interested in being proven wrong.
4) Some analog stuff, to my limited knowledge, is still difficult to model.
Why is this an issue? When was the last time you had digital electronics fail, and it wasn't a victim of the early 2000s capacitor plague (in which case, replacing the affected caps fixed it)? "Repairability" is vastly overrated: most electronics don't fail in the first place, and when they do, it's almost always electrolytic caps, and this affects both digital and analog circuits. There's nothing "unrepairable" about a digital circuit: it's not like a logic gate on your CPU is likely to fail.
>and even less if it has IOT crap (why??).
If you're referring to amps (like Fender Mustang series) that use your phone for the UI, that should be pretty obvious: the phone is an ubiquitous device that you can assume the customer owns, and provides a large touchscreen for controlling the amp remotely. It's not IOT; it's a Bluetooth connection. It's not necessary for using the amp, it's just an option, and a really helpful one when playing with effects settings, and much easier to use than the tiny LCD screen on the amp plus its jog wheel. Once the settings are done, you can assign settings to footpedal buttons and never mess with it again. With the 7-button footpedal, for instance, you can easily set it up with 5 different sound profiles you can easily switch between, with the other buttons you can switch between "banks" of 5 effects if you need more for one song/gig.
Also some say some quirks of analog designs are hard to replicate in simulation althought I haven't seen double blind tests of that so I'm mildly skeptical of that
However, if you don't need lots of gain a decent preamp can be built around the venerable 2N7000. Yes, that's a MOSFET, definitely not aimed at audio, but it's dirt cheap, available everywhere ...and noisy. But if you keep the gain to reasonable levels it works quite well.
Here's some info.
http://www.muzique.com/schem/mosfet.htm
(the entire site is worth bookmarking)
https://hackaday.com/2023/05/02/fet-fun-endeavors-together/#...