Differential phase splitter with level shift capability

[Reactance]

Hi all

Ive recently been playing around with class-d building blocks and trying out various arrangements, im at a point where i feel "xor phase splitting" with separate transistor level translation is unnecessary. (i know there are cases where the xor stage do both level-shifting and phase inversion) however the discrete version should have a faster response time.

I'm at the moment trying to do level "shifting and phase splitting" all in one go, using a "differential stage" that's able to make a perfect copy of the source signal, level shifting it and allowing phase splitting at the same time.

Ive seen Eva done something similar, where her level shifter/phase splitter comprised of a differential pair only.

Here is a partial example of a differential stage acting as a phase splitter and level shifter at the same time.

[Reactance]

bump?

[ChocoHolic]

You did not ask anything. What should we answer?

Some measurements on a similar thing here:
1kW Gen2
posting #1 for schematic
posting #5 for measurement

...but finally I stepped to an integrated comparator + fast discrete level shifter.
==> improved gain and flipping precision.

[Reactance]

Sorry, i only realized i didn't post a question that was dumb !

Anyway my confusion/question is why does so many circuits employe XOR gate for phase splitting and separate level shifting?
Where a simple long tail pair does phase inversion + level shift all together performance is better propagation delay faster, the only reason why is to actually benefit from longer propagation delays using a xor with separate level shifting.

Thanks
Yes post #5 measurements are very interesting im looking at it now.

[ChocoHolic]

Originally Posted by Reactance

..the only reason why is to actually benefit from longer propagation delays using a xor with separate level shifting.

It depends on the overall circuitry if you wish to have a prolonged propagation delay. For UcD you might be right.
For most clocked designs the a propagation is not helpful, longer propagation delay there typically worsens carrier aliasing...

[darkfenriz]

XOR gates easily provide trimmable dead time, which is very constant with parameters spread, temperature and so on.
I cannot see how it is possible to easily tweak the dead time in your arrangement, I'd prefer some gates for prototyping-friendly design.

[ChocoHolic]

XORs provides deadtime?
Or do you mean XORs with a resistor-diode-capacitor-timing-network at the input?
If we consider such a network, it could basically also be used at the inputs of the IR21xxx, without XOR. Or did you already experience issues with tolerances and/or temp drift in that arrangement?

[darkfenriz]

Of course I mean with RCD networks. XOR gates are as good as any other gates or IR21xx input, but they are the best for phase splitting if one is working as buffer and the other as an inverter.
They do so with minimum skew and very well matched delay regardsless from the fact that their threshold is 30% or 70% Vcc or wheather their input is sharp edged or slowly slewing.

[elevator]

Here's what I use FWIW.
No diode needed, works fine for me.
Cheers

[Reactance]

I'm really get bad signal response time, im using a single BC557 PNP, F=250Khz (wired up in a common base configuration), the output from the LM311 is a clean square wave, however after level shifting rise time on the pulse increased up to 1 us !! is this normal for level shifting.?

I can confirm XOR splitting is very clean, but its a bit bulky vs a standard long tail pair with increased performance. (i think)