schmitt triggered LVDS receiver help!!!

[gilbertomaldito]

lvds receiver

Hi, I got this circuit from an IEEE paper. It says that the differential voltage signal at the input is detected by schmitt trigger M1-M6 which ensures a 25mV/50mV hysteresis. I know what schmitt triggering is, but my question is, how does this circuit able to perform schmitt triggering action? and how am I going to measure its hysteresis in HSPICE simulation?
thanks a lot

 

[erikl]

lvds receiver hysteresis

how does this circuit able to perform schmitt triggering action?

M5 & M6 provide positive feedback. The W/L ratios of M5:M3 & M6:M4 define the hysteresis.

and how am I going to measure its hysteresis in HSPICE simulation?

Put a fixed comparison voltage to one of the inputs and a saw tooth voltage to the other one. Then run a transient simulation.

 

[snafflekid]

reciver schmitt

You can run a DC sweep in the positive direction and the negative direction, then overlay the two curves.

 

[gilbertomaldito]

lvds schmitt trigger

@snafflekid, can you explain further? I understand what you mean but I dont understand why would I do that? How am I going to meause the hysteresis in that curve?

 

[Vozhyk]

schmitt-triggered

1. Add two variable: Vin, i
2. set dc voltage in the vdc source: i*Vin+3*0.5*(1-i)
3. set in dc analysis that Vin change from 0 to 3
4. run parametric analysis by i: 1, -1

in the result vdc will change from 0 to 3 when i = 1 and change from 3 to 0 when i = -1

 

[erikl]

measuring hysteresis receivers

... and this can all be done with one single transient simulation, including the hysteresis switching dynamics ;-)

 

[gilbertomaldito]

lvds receivers with hysteresis

@erikl, I have done what you suggested. I applied DC on Vin- and a saw tooth on Vin+ with its DC equal to Vin-. When I set the ratios(M5:M3 & M6:M4) equal to 1, the output is almost a square-ish fullswing signal with no. However, when I tried to make my ratio equal to 0.25, the output has slower transition time. It seems like the gain decreases. Its not really what I expected.

Added after 1 minutes:

here is the simulation results.

 

[gilbertomaldito]

how to measure hysteresis for lvds

I was expecting that after setting the ratio to 0.25, there will be transition delay of the output from the intersection of the two input signals.

 

[erikl]

@erikl, I have done what you suggested. I applied DC on Vin- and a saw tooth on Vin+ with its DC equal to Vin-. When I set the ratios(M5:M3 & M6:M4) equal to 1, the output is almost a square-ish fullswing signal with no. However, when I tried to make my ratio equal to 0.25, the output has slower transition time. It seems like the gain decreases. Its not really what I expected.

Hi andrew,
you shouldn't trust anyone (at least not me ;-) ), and always check the info yourself! I'm sorry, but the relevant ratios are not (M5:M3 & M6:M4), but (M6:M3 & M5:M4), because M6 provides feedback to the M2-M3-G(M7) node, and M5 to the M1-M4-G(M9) node.

For hysteresis, these ratios should be asymmetric. For high switching speed, they could be even > 1 . Just try by simulation!

Good luck! erikl

 

[gilbertomaldito]

@erikl. I got the correct behavior now by making the ratio greater than 1. thank you very much.
However, I still cant explain how it provides hysteresis. Is it ok if you can explain it to me please? Im really sorry if im too slow.

 

[erikl]

@erikl.
However, I still cant explain how it provides hysteresis. Is it ok if you can explain it to me please?

@andrew: Your Schmitt-Trigger owns 2 concurrent gain loops for each input.
Hereafter, their signal paths together with their corresponding phase conditions (+) & (-) are shown (always related to the outputs, i.e. drains of the transistors). The 1st line represents the main loop, the second line is the loop which adds additional gain (actually not a common feedback scheme, but it works sort like a positive feedback, because it adds positive gain via a different path) :

VIN+(+) (M1 M4)(-) (M9 M10)(+) (M8 Vout)(-)
VIN+(+) (M1 M4)(-) (M6 M2 M3)(+) (M7 Vout)(-)

VIN-(+) (M2 M3)(-) (M7 Vout)(+)
VIN-(+) (M2 M3)(-) (M5 M1 M4)(+) (M9 M10)(-) (M8 Vout)(+)

This should give us the following info:
1. The input signal names should be inverted, because it's usual to relate their names to the output phase condition.

2. The 2 IO-paths have different lengths and number of phase inversions (3|3 resp. 2|4). This creates a certain asymmetry.

3. The gain values - as always - depend on the respective W/L ratios. So if you change the gain ratios equally within the 2 concurrent (but gain-added) paths, you change the overall gain, and, by this, the switching speed.

4. Some hysteresis could only be achieved (if ever), if the two individual IO-path gain ratios are different, which creates additional asymmetry. But I can't explain exactly, why. All the 4 gain paths should result in an overall gain Vout/VIN(diff); they just have different IO-delays. May be these delay differences can create some hysteresis. Rely on your sim. results!

I must admit that your Schmitt works a bit differently than I thought after first sight! ;-)

Cheers, erikl
Thanks for your thanks points!

 

[gilbertomaldito]

@erikl. Thank you for the details. Yeah, It could be delay differences due to aymetric signal path gains, but I think Ill investigate further through simulations. Thanks for giving me the idea.

"thanks points" is the least I can do..