Thanks for the clear explanation. I would like to mention the TXS0104E device that I have been using. The TXS0104E contains 4 level translators similar to what you showed. Additionally it has internal 10k pull-up resistors, and edge detection on the input signal from either side. When a transition from low to high is detected, the internal circuitry actively pulls up the input to speed up the transition. After the transition the active pull up is again released. Very handy for long distance one-wire and two-wire applications.
Anyone who likes to play with microcontrollers is probably all to familiar with this schematic.... but what we were lacking was a really good description of what's actually going on.... until now that is.
I'm confused - but that's good. Makes me do some head scratching. One thing though. In High School and the Air Force all of my instructors used electron flow! When back in college, some of the professors would use electron flow and the majority would use current flow. It was the same way with the text books. I graduated with PTSD- or something. There was hole flow but I don't want to get into that. To many memories. Well that's maybe why I'm confused. Thanks for the time spent on a great presentation!
"That means you have 4 level converters on one board - That's what I have over here" - getting in on the high side and coming out at the low side." With five of them for a buck and a half I was sure to see the same signal converted from high to low and back 10 times in a row. But thank you for the perfect explanation. 😄
The NXB0101 chip (from Nexperia) is yet another dual supply transceiver, better suited for bus translation up to ~100MHz (3.3V↔5V, and it can also go high-Z). NXS0101 is the open drain cousin, though it's much slower.
A few points: there is a 2N7002 which is like the 2N7000 but with tightened spec (threshold voltage = 2.5V max) in an SOT-23 package. However the BSS138 is cheaper, so good catch. The body diode in these small signal MOSFETs isn't going to be dropping as much as 0.7V in conduction, more like 0.6V when loaded with 10K from a 3.3V rail, so Vgs is at least 2.7V. At 9:00 when the MOSFET starts conducting as the diode is turning on, the channel resistance becomes so low that the source is effectively dragged down to 0V. You now have 3.3V Vgs, not 2.6V. At 8:20 we really don't want the MOSFET to turn on "hard", just "hard enough", otherwise the increased gate charge reduces switching speed. Better not to in level-shifting applications.
This is timely and awesome! I've been using the classic BSS138 circuit for several products for the last 5 years. I just adapted the same circuit as a cutoff switch for an i2c bus that was conflicting with our AVR programmer. Both sides are at 3v3, but the gate is tied to the programmer's separate 3v3 bus.
The 74CBT16211 bus switch is my favourite device for 5V3.3V logic level conversion. It has 24 lines, so you can level convert an entire chip at once, and the fact that it doesn't use pull-up resistors makes that it works at much higher speeds than a MOSFET solution. You should make a video about it!, Using the 16211 for logic level conversion ain't too obvious from the data sheet, so a video would be helpful for people.
Are you sure it really _can_ do logic level translation? Sure, there's a bullet point mentioning logic levels but there's no way to tell it the range for a given port. The description says switch, and another bullet point discusses analog applications among others. I struggle to see how it could function as a level translator while also being suitable for analog applications. Have you hooked a scope to an input and output and verified that it's actually reducing/increasing the voltage?
@@markp5726 Yes, I told this ain't obvious from the data sheet at all. Search for the "Texas Instruments CBT (5ĆV) and CBTLV (3.3ĆV) Bus Switches data book". At the end of the book, there is a chapter about using these CBT chips for voltage translation. Basically the FET switch reduces incoming voltages to 4 Volt. To make them limit voltages to 3.3V you insert a diode into the Vcc line, and you can buy these switches with the diode already built in, they have a D into their part number: 74CBTD16211.
@@markp5726 I said it was badly documented :) The 74CBT74211 will by default limit voltages to 4V. You need to insert a diode in the Vcc line and you will get 3.3V. But there is more: There exists an 74CBTD16211 with this diode already integrated into the chip, making it ideal for voltage capping an entire bus at once.
@@danielmantione oh that is weird. The datasheet for 74CBT16211 (which I found first) is fairly different from the CBTD version /me dives for the rabbithole...
Keeping in mind these are only good for low speed like serial ports and slow spi etc. There is way too much capacitance in the junction to use for, say 10Mhz, and you have to resort to the proper chips with more stuff inside them. It can be easily seen on the oscilloscope at higher frequencies and can be confusing on the logic analyser.
How fast can you get with two NPN BJT's connected together in reverse orientation (emitter to collector)? With the two bases pulled up to the respective voltages you want translated as well as pullups on the emitter/collectors. example: 1K |------/\/\/\------| 5V | 1K | |------/\/\/\------| 5V | ___|___ 1K | * |/ \ |------/\/\/\-----| 3.3V | __/ \__ | A o__|__| |__|__o B | __ __| \ /| *2N3904 _\___/_ | 1K |------/\/\/\-----| 3.3V
@@pol.kraine7890 You would have to draw it up in spice. It would be faster as FETs have much more capacitance then ye olde BJTs. Then again, the dedicated translation devices have Schmidt triggers, giving them even less limitations at high bandwidth.
I use 74LVC245 for level shifting. The 74LVC245 is great for higher speed signals vs discrete FET based level shifters & have 8 lines. Additionally TXB0108 is a great level shifter too though I have quite a few SN74LVC245AN’s around.
4-ch PCBs from CN are Sparkfun module from 2013 clones btw. A lot of those exist thanks to Adafruit, Sparkfun and the like OSHW contributions cloned and mass-produced often using CN analog chips (which are statistically quite good) if such exists, with genuine brand parts too, but generally CN modules almost always work.
Nice very useful, for better results TXS(0104) series from Texas Instruments for open drain applications like I2C or MDIO buses and TXD(0104) series for push-pull CMOS logic like data line of SRAM are also available.
interesting, You also got TXS0101 (single), TXB0108 (8 fold), but require 2 100nF capacitors on both supply sides. But the Sizklai also works pefectly, but requires 1 more transistor than your circuit, but not bi directional.
There are indeed much better MOSFETs available nowadays than the BS170, and they are also much cheaper. The only advantage it has is that it cab be brought in through-hole package, making it easy to use on breadboards and experimental PCB's and so.
Basically, it is a common-gate topology. The source & drain of the FET are generally interchangeable. Moreover, there is a unique EM mosfet which has a body diode.
I've seen bipolar transistors used in the same topology. The only change needed is a base resistor. As with a mosfet the transistor will work backwards.
nice introduction of this nice MOSFET and the use in level shifters. However, the real resistance is of minor importance for a level shifter. Even for I2C lines it only has to be significantly below the pull-up resistors' values, but 10s or even 100s of Ω should not matter much, unless you have long lines with high parasitic capacitance and high frequency signals.
Hello Mr engineer I want to help me how to find circuit schematic all of the electronics appliance for repairing like this power switching & main board of the other circuits Thank you Mr Engineer
I'd like to convert TTL to 24 volts 60mA for FSK Radio teletype of Old.Bidirectional so i can get my 1950s Modem to think the computer is a Teletypewriter?
Hi Sir please reply... this question is very very confusing... I have 2 identical 12V DC 7.5Ah batteries say X and Y. One side of a normal bulb is connected to postive terminal of X battery, and another side of the bulb is connected to the negative terminal of the Y battery. The negative terminal of X battery and the positive terminal of Y battery are not connected to anything. So, in this case, I want to know whether there will be a current flow to the bulb, and whether the bulb will glow??? Give reasons if bulb will glow, and reasons why bulb will not glow. Thanks in advance.
Thanks. The datasheet shows maximum Vgs(th) is 1.5V. It means if the gate is 3.3V and 0V at source in the circuit, then this MOSFET will get damaged. So it cannot be used for 3.3V to 5V level shifting. Please correct me if i am wrong....
The parameter you are listing is the threshold voltage, not a maximum voltage. There must be a voltage difference between the gate and the source for the MOSFET to turn on. That value in the datasheet is saying that for some devices that difference may be as low as 0.8 volts, for most it will be 1.3 volts, and for some the difference must be at least 1.5 volts before the device will turn on. The value you probably were thinking of is Vgss, which my datasheet lists as + or - 20V for the BSS138.
You create a wide variety of topical videos. The number of views sometimes vary wildly. Do you bother to look for correlations to number of views? Your database for analysis increases daily. :-)
Great explanation, love the detailed analysis. But 0.7 volts is (kinda) valid for 3.3 volt devices. Also given the light loading, the diode will (probably) be less than 0.7 volts. "Engineering is the art of what you can get away with".
@@jspencerg Except yeah they do. There's a reason for the saying "Any idiot can build a bridge that stands, but it takes an engineer to build a bridge that barely stands."
@@AttilaAsztalos "...with the correct safety factor thrown in" Some engineering fields permit or even encourage wanton cost saving, planned obsolescence or life till the end of warranty. Bridge building in the USA is not one of those. Neither are the probe building wizards at NASA.
The 0.7V is only there before the fet turns on - once it's on, you get the 5 ohm (or less) path through the fet, and the diode is essentially out of the circuit. The diode just bootstraps the fet into the on condition, and then the fet takes over, and only in the direction of a low on the right driving the left side low. It's actually extremely clever how it works!
Thanks, I did not know about this topology for level shifting. @5:11 remember the source and drain of the normal MOSFET transistors are interchangeable. So when the output is low the source and drain are exchanged. Here the body diode does not get engaged until much higher currents( that is when I x Ron > 0.7 V) The advantage of this topology is its speed and it is bi-directional, the disadvantage is use of constant static current. Normally level shifters are made of cross coupled pmos transistors on top of nmos transistors. The disadvantage is that it slower than this topology. Also, if you have MOSFET transistor that you could connect to its body, it is better to connect the body to the lowest voltage( ie. in this case 0 V)
The static current situation isn't as bad if you're already using a protocol, like I2C, that is designed to use open collector outputs, pullup resistors, and logic high idle state.
Thanks for the clear explanation. I would like to mention the TXS0104E device that I have been using.
The TXS0104E contains 4 level translators similar to what you showed. Additionally it has internal 10k pull-up resistors, and edge detection on the input signal from either side. When a transition from low to high is detected, the internal circuitry actively pulls up the input to speed up the transition. After the transition the active pull up is again released. Very handy for long distance one-wire and two-wire applications.
I was just going to post the same thing. I've had nothing but trouble with a board from Sparkfun with the described design.
You have to remember that a lot of logic fets, were designed when logic levels were 5v and not today's lower levels
Anyone who likes to play with microcontrollers is probably all to familiar with this schematic.... but what we were lacking was a really good description of what's actually going on.... until now that is.
I'm confused - but that's good. Makes me do some head scratching. One thing though. In High School and the Air Force all of my instructors used electron flow! When back in college, some of the professors would use electron flow and the majority would use current flow. It was the same way with the text books. I graduated with PTSD- or something. There was hole flow but I don't want to get into that. To many memories. Well that's maybe why I'm confused. Thanks for the time spent on a great presentation!
"That means you have 4 level converters on one board - That's what I have over here" - getting in on the high side and coming out at the low side."
With five of them for a buck and a half I was sure to see the same signal converted from high to low and back 10 times in a row.
But thank you for the perfect explanation. 😄
The NXB0101 chip (from Nexperia) is yet another dual supply transceiver, better suited for bus translation up to ~100MHz (3.3V↔5V, and it can also go high-Z). NXS0101 is the open drain cousin, though it's much slower.
A few points: there is a 2N7002 which is like the 2N7000 but with tightened spec (threshold voltage = 2.5V max) in an SOT-23 package. However the BSS138 is cheaper, so good catch.
The body diode in these small signal MOSFETs isn't going to be dropping as much as 0.7V in conduction, more like 0.6V when loaded with 10K from a 3.3V rail, so Vgs is at least 2.7V.
At 9:00 when the MOSFET starts conducting as the diode is turning on, the channel resistance becomes so low that the source is effectively dragged down to 0V. You now have 3.3V Vgs, not 2.6V.
At 8:20 we really don't want the MOSFET to turn on "hard", just "hard enough", otherwise the increased gate charge reduces switching speed. Better not to in level-shifting applications.
This is timely and awesome! I've been using the classic BSS138 circuit for several products for the last 5 years. I just adapted the same circuit as a cutoff switch for an i2c bus that was conflicting with our AVR programmer. Both sides are at 3v3, but the gate is tied to the programmer's separate 3v3 bus.
I've used a 2N7000 to level shift 5V to 3.3V and it worked fine. Also used an A2SHB smt mosfet. The max threshold for the A2SHB is a Mere 0.85V
The 74CBT16211 bus switch is my favourite device for 5V3.3V logic level conversion. It has 24 lines, so you can level convert an entire chip at once, and the fact that it doesn't use pull-up resistors makes that it works at much higher speeds than a MOSFET solution. You should make a video about it!, Using the 16211 for logic level conversion ain't too obvious from the data sheet, so a video would be helpful for people.
Are you sure it really _can_ do logic level translation? Sure, there's a bullet point mentioning logic levels but there's no way to tell it the range for a given port. The description says switch, and another bullet point discusses analog applications among others. I struggle to see how it could function as a level translator while also being suitable for analog applications.
Have you hooked a scope to an input and output and verified that it's actually reducing/increasing the voltage?
@@markp5726 Yes, I told this ain't obvious from the data sheet at all. Search for the "Texas Instruments CBT (5ĆV) and CBTLV (3.3ĆV) Bus Switches data book". At the end of the book, there is a chapter about using these CBT chips for voltage translation. Basically the FET switch reduces incoming voltages to 4 Volt. To make them limit voltages to 3.3V you insert a diode into the Vcc line, and you can buy these switches with the diode already built in, they have a D into their part number: 74CBTD16211.
@@markp5726 I said it was badly documented :) The 74CBT74211 will by default limit voltages to 4V. You need to insert a diode in the Vcc line and you will get 3.3V. But there is more: There exists an 74CBTD16211 with this diode already integrated into the chip, making it ideal for voltage capping an entire bus at once.
@@danielmantione oh that is weird. The datasheet for 74CBT16211 (which I found first) is fairly different from the CBTD version
/me dives for the rabbithole...
Keeping in mind these are only good for low speed like serial ports and slow spi etc. There is way too much capacitance in the junction to use for, say 10Mhz, and you have to resort to the proper chips with more stuff inside them. It can be easily seen on the oscilloscope at higher frequencies and can be confusing on the logic analyser.
How fast can you get with two NPN BJT's connected together in reverse orientation (emitter to collector)? With the two bases pulled up to the respective voltages you want translated as well as pullups on the emitter/collectors.
example:
1K
|------/\/\/\------| 5V
| 1K
| |------/\/\/\------| 5V
| ___|___ 1K
| * |/ \ |------/\/\/\-----| 3.3V
| __/ \__ |
A o__|__| |__|__o B
| __ __|
\ /| *2N3904
_\___/_
| 1K
|------/\/\/\-----| 3.3V
@@pol.kraine7890 You would have to draw it up in spice. It would be faster as FETs have much more capacitance then ye olde BJTs. Then again, the dedicated translation devices have Schmidt triggers, giving them even less limitations at high bandwidth.
I use 74LVC245 for level shifting. The 74LVC245 is great for higher speed signals vs discrete FET based level shifters & have 8 lines. Additionally TXB0108 is a great level shifter too though I have quite a few SN74LVC245AN’s around.
There's several ICs available that work the same way. One is the SGM4553.
This is always a strange transistor arrangement. Nicely explained, as usual !
4-ch PCBs from CN are Sparkfun module from 2013 clones btw.
A lot of those exist thanks to Adafruit, Sparkfun and the like OSHW contributions cloned and mass-produced often using CN analog chips (which are statistically quite good) if such exists, with genuine brand parts too, but generally CN modules almost always work.
Nice very useful, for better results TXS(0104) series from Texas Instruments for open drain applications like I2C or MDIO buses and TXD(0104) series for push-pull CMOS logic like data line of SRAM are also available.
It's wizardry I tell you. Wizardry. 🧙♂️👍
interesting, You also got TXS0101 (single), TXB0108 (8 fold), but require 2 100nF capacitors on both supply sides.
But the Sizklai also works pefectly, but requires 1 more transistor than your circuit, but not bi directional.
There are indeed much better MOSFETs available nowadays than the BS170, and they are also much cheaper. The only advantage it has is that it cab be brought in through-hole package, making it easy to use on breadboards and experimental PCB's and so.
Basically, it is a common-gate topology. The source & drain of the FET are generally interchangeable.
Moreover, there is a unique EM mosfet which has a body diode.
I've seen bipolar transistors used in the same topology. The only change needed is a base resistor. As with a mosfet the transistor will work backwards.
Excellent information. 🥳 Thank you.
Superb! Thanks.
nice introduction of this nice MOSFET and the use in level shifters. However, the real resistance is of minor importance for a level shifter. Even for I2C lines it only has to be significantly below the pull-up resistors' values, but 10s or even 100s of Ω should not matter much, unless you have long lines with high parasitic capacitance and high frequency signals.
A great low gate voltage mosfet is the Si2302, ultra cheap!
Hello Mr engineer
I want to help me how to find circuit schematic all of the electronics appliance for repairing like this power switching & main board of the other circuits
Thank you Mr Engineer
ua-cam.com/video/Ii--aTKocTA/v-deo.htmlsi=jtoMnPQBzQoGZu8K
How high in frequency can those little FET level translators boards go? Thanks.
Thanks. Do you have a link to that digikey doc, or the "basics" series they have?
www.digikey.com/en/blog/logic-level-shifting-basics
Awesome....cheers.
I'd like to convert TTL to 24 volts 60mA for FSK Radio teletype of Old.Bidirectional so i can get my 1950s Modem to think the computer is a Teletypewriter?
forum.vcfed.org/index.php?threads/mek6800d1-output-interface-for-60ma-teletype-current-loop.47518/
Hi Sir please reply...
this question is very very confusing...
I have 2 identical 12V DC 7.5Ah batteries say X and Y.
One side of a normal bulb is connected to postive terminal of X battery, and another side of the bulb is connected to the negative terminal of the Y battery.
The negative terminal of X battery and the positive terminal of Y battery are not connected to anything.
So, in this case, I want to know whether there will be a current flow to the bulb, and whether the bulb will glow??? Give reasons if bulb will glow, and reasons why bulb will not glow.
Thanks in advance.
sounds like a homework problem
Thanks. The datasheet shows maximum Vgs(th) is 1.5V. It means if the gate is 3.3V and 0V at source in the circuit, then this MOSFET will get damaged. So it cannot be used for 3.3V to 5V level shifting. Please correct me if i am wrong....
The parameter you are listing is the threshold voltage, not a maximum voltage. There must be a voltage difference between the gate and the source for the MOSFET to turn on. That value in the datasheet is saying that for some devices that difference may be as low as 0.8 volts, for most it will be 1.3 volts, and for some the difference must be at least 1.5 volts before the device will turn on. The value you probably were thinking of is Vgss, which my datasheet lists as + or - 20V for the BSS138.
@@tfrerich Heartly thanks for your clarification... Yes i got confused with Vgss & Vg(th)... Thank you...
You create a wide variety of topical videos. The number of views sometimes vary wildly. Do you bother to look for correlations to number of views? Your database for analysis increases daily. :-)
yes, I look at the statistics. If I only wanted views/subscribers, it would make my channel stupid. I just do what I want.
@@IMSAIGuy……….and thanks !
why do so many FET symbols show the body diode as a zener?
I believe it’s a Schottky diode. If so the forward drop would be about 0.3V not 0.7V.
Great explanation, love the detailed analysis. But 0.7 volts is (kinda) valid for 3.3 volt devices. Also given the light loading, the diode will (probably) be less than 0.7 volts. "Engineering is the art of what you can get away with".
I doubt bridge building engineers embrace that saying.
@@jspencerg One would hope, but I'm afraid some do.
@@jspencerg Except yeah they do. There's a reason for the saying "Any idiot can build a bridge that stands, but it takes an engineer to build a bridge that barely stands."
@@AttilaAsztalos "...with the correct safety factor thrown in"
Some engineering fields permit or even encourage wanton cost saving, planned obsolescence or life till the end of warranty. Bridge building in the USA is not one of those. Neither are the probe building wizards at NASA.
The 0.7V is only there before the fet turns on - once it's on, you get the 5 ohm (or less) path through the fet, and the diode is essentially out of the circuit. The diode just bootstraps the fet into the on condition, and then the fet takes over, and only in the direction of a low on the right driving the left side low. It's actually extremely clever how it works!
BSS214N is a little better in some aspects: 1.2V max threshold voltage and 0.25 ohm at 2.5V Vgs.
Thanks, I did not know about this topology for level shifting.
@5:11 remember the source and drain of the normal MOSFET transistors are interchangeable. So when the output is low the source and drain are exchanged.
Here the body diode does not get engaged until much higher currents( that is when I x Ron > 0.7 V)
The advantage of this topology is its speed and it is bi-directional, the disadvantage is use of constant static current.
Normally level shifters are made of cross coupled pmos transistors on top of nmos transistors.
The disadvantage is that it slower than this topology.
Also, if you have MOSFET transistor that you could connect to its body, it is better to connect the body to the lowest voltage( ie. in this case 0 V)
The static current situation isn't as bad if you're already using a protocol, like I2C, that is designed to use open collector outputs, pullup resistors, and logic high idle state.