# Fast voltage translation from 3.3v to 5.7v

I need to translate a 3.3v signal to 5.7v with transition of around 60ns or less. Voltage translators such as the TXS0101 usually specify the logic high level for the input as VDD x ~0.65 (= 3.7v in my case), and their upper level is usually 5.5v. I have tried with a P and N MOSFET (see diagram below) but the P MOSFET is not saturated because 3.3v - 5.7v = -2.4v which is insufficient. Adding a resistor between 5.7v and the P MOSFET source fixes the saturation but the switching time becomes too long.

Is there in fact an IC which can do what I want? I've looked around but can't find one.

One other consideration is that a voltage drop at the output (to around 4.7v) is acceptable, but doing that without burning quite a lot of watts seems to be tricky. I tried using a 4.7v zener to drop the P input to 4.7v but it wastes a lot of current.

Update with 2nd attempt here. It's not quite as quick as I'd like but think it can work. Thanks for the input, the baker clamp was new to me, works like a charm.

2nd attempt: Fall time: Rise time:

simulate this circuit – Schematic created using CircuitLab

 relaxed timing requirement to 40ns

 relaxed timing requirement to 60ns

• can you use a transformer? – Jasen Nov 3 '18 at 20:17
• I don't know, footprint needs to be less than a few square mm, can they switch that fast? – Charlie Skilbeck Nov 3 '18 at 20:23
• How large is the load on the output? There are small MOSFETs with lower thresholds. – CL. Nov 4 '18 at 8:41

simulate this circuit – Schematic created using CircuitLab

Rise time should be in the 25ns range typically and fall time better (10%-90%). If you only need 50ns typical you can increase R1.

Do not use a larger than necessary MOSFET for M1 or the drain source capacitance and Miller capacitance (assuming some source resistance) will kill your speed.

• Wouldn't a BJT be more suitable? Considering they usually have lower input & output capacitance compared to MOSFETs for the same price => larger margin => faster slew-rates? - In other words, just replace M1 with a BJT + 1 kΩ base resistor. – Harry Svensson Nov 3 '18 at 23:03
• @HarrySvensson Maybe with a Baker clamp, otherwise saturation will kill the speed. – Spehro Pefhany Nov 3 '18 at 23:04
• My power budget isn't quite that generous, I'll look into the BJT option – Charlie Skilbeck Nov 3 '18 at 23:10
• A Baker clamp is a small (the smallest you can realistically use) Schottky diode between collector and base so it steals base current when 'on' and prevents the transistor from saturating. – Spehro Pefhany Nov 3 '18 at 23:12
• @CharlieSkilbeck You could check out the 2SC3355. I've never used it, but it's relatively cheap and overkill and will most certainly be fast enough. Don't forget to add a baker clamp as Spehro said. – Harry Svensson Nov 3 '18 at 23:13

This IC is quite a bit overkill, but it would do the job.

I am sure there are single-channel level shifters/translators. The issue is finding them.

You could probably also use a TTL LVDS receiver in a pinch, by biasing one of the inputs and attenuating the input signal.

• Thanks Edgar, yes, that's a bit much IC for this task – Charlie Skilbeck Nov 4 '18 at 7:46

There is one way to do this with a single transistor (NMOS), I used the BSS123 (you can get it in small sot-23 packages). You could try something like this:

This bidirectional—you could send a signal from the high voltage side to the low one, too. But say, you want to push a 0 from the DL (low side) line to the DH (high side) line, the NMOS will turn on bring the 0V level to the drain. If you leave DL floating or apply 3.3V, the NMOS will be off and DH will be high due the 10k pullup $$\R_2\$$.

As far as the speed, well, I did a quick test in LTSpice and found that a 0V signal propagates through (from DL to DH) at a reasonably high speed (~5ns). Of course, in practice this will depend on the input impedance of whatever you attach to the DH side.

The green plot shows a 0V signal going from the DL to the DH side (blue).

You can find readily available versions of this circuit on DigiKey for example check this out

Hope this helps.

• Thanks for this, the fall time is good but the rise time is a few hundred nanoseconds. Lowering the value of the resistors improves it, but to get it fast enough, the current draw is too high – Charlie Skilbeck Nov 4 '18 at 7:45