LMK00308: Question about LVDS AC coupling

Lee HongGyo

Part Number: LMK00308
Other Parts Discussed in Thread: LMK00304, LMK1D1204P,

In the past, there has been a discussion about where the termination resistor should be placed in relation to LVDS AC coupling.

The link below is a discussion of the location of the termination resistor.

https://e2e.ti.com/support/clock-timing-group/clock-and-timing/f/clock-timing-forum/722839/lmk00304-input-lvds-ac-coupling

Assuming that there is a weak bias inside the LVDS receiver, there is no difference if there is a clock at the input of both A/B structures.

If the clock is not input to the input pin, the structure of A has a possibility of chattering and

The structure of B explains that it has the advantage of maintaining a certain state without chattering.

However, TI often responds by recommending the B structure.

https://e2e.ti.com/support/clock-timing-group/clock-and-timing/f/clock-timing-forum/1058887/lmk00304-is-there-internal-dc-biasing-and-clarifying-terminations?tisearch=e2e-sitesearch&keymatch=LMK00304#

https://e2e.ti.com/support/clock-timing-group/clock-and-timing/f/clock-timing-forum/789767/lmk00308-lmk00304-08-bias-voltage?tisearch=e2e-sitesearch&keymatch=LMK00304#

However, in reality, application examples are provided in the datasheet by mixing various A/B structures.

This makes it very confusing in practical applications.

The links and datasheets below are some examples to confuse you.

e2e.ti.com/.../lmk00308-lvds-outputs-with-ac-coupling-to-receiver-with-internal-100-ohm-termination

LMK00304

LMK1D1204P

So, I ask again about LVDS AC coupling as below, and hope it gets sorted out.

If there is a constant clock on the receiver input, my personal opinion is that the A structure is good.

The reason is the duty-cycle change.

TI says that there is no duty-cycle change according to the A/B structure, but this answer applies when the clock's rising/falling time is very very small or the frequency is low( f<100MHz ).

https://e2e.ti.com/support/clock-timing-group/clock-and-timing/f/clock-timing-forum/789767/lmk00308-lmk00304-08-bias-voltage?tisearch=e2e-sitesearch&keymatch=lmk00308#

Actually, considering the rising/falling time of the input signal, it is thought that the duty-cycle change occurs as the frequency increases.

That is, below 100 MHz (approx. duty cycle = 50%), the duty cycle change is hard to see, but at 200 MHz, the duty cycle change is visible.

The reason for this is the different offset settings of P/N to prevent chattering.

Therefore, if the A structure is used, the other offset of P/N applied to prevent chattering is ignored and the P/N offset of the signal is equal to about 1.2V, so that the duty-cycle change does not occur even for a 200MHz signal.

In the case of LVDS AC coupling, if the clock continues to exist at the receiver input, is there any structure in the above A/B structure that is particularly problematic?

Is structure B still good?

Also I hope to provide a PSpice model for the clock buffer.

best regards

Lee HG

over 2 years ago

0 Lee HongGyo over 2 years ago

Prodigy 220 points

The link "e2e.ti.com/.../lmk00308-lvds-outputs-with-ac-coupling-to-receiver-with-internal-100-ohm-termination" is not properly connected.

--> https://e2e.ti.com/support/clock-timing-group/clock-and-timing/f/clock-timing-forum/589608/lmk00308-lvds-outputs-with-ac-coupling-to-receiver-with-internal-100-ohm-termination?tisearch=e2e-sitesearch&keymatch=LMK00308#\

0 Asim Khan over 2 years ago in reply to Lee HongGyo

TI__Genius 9250 points

Hi Lee,

Here are some comments related to Type A and Type B termination schemes as you mentioned.

During no clock input at input, the TYPE A does have the possibility of chattering because the 100 ohm resistor will act as short between the inputs and overriding any built in hysteresis. While B will have leverage in that case because 100 ohm is not there to override the input differential voltage set internally.

TYPE B is preferred for smooth startup as in case of LMK00304.

For continuous clock, structure A is good to use as well as mentioned in of the articles.

https://e2e.ti.com/support/clock-timing-group/clock-and-timing/f/clock-timing-forum/1058887/lmk00304-is-there-internal-dc-biasing-and-clarifying-terminations?tisearch=e2e-sitesearch&keymatch=LMK00304

Your question related to duty cycle distortion in TYPE A vs TYPE B is something I am still investigating with couple of different devices to make sure the behavior is similar to what you listed in your comments. On that I will get back to you by Oct 19, 2022.

0 Lee HongGyo over 2 years ago in reply to Asim Khan

Prodigy 220 points

Hi Asim Khan

Thank you for your kind explanation and quick response.
Generally
During no clock, TYPE B has no chattering potential and TYPE A has chattering potential.
It is said that Type A is good as long as there is a clock.

But I think the LMK003xx series is dedicated to TYPE B.
The reason I thought so is as follows.

You described it like this:
TYPE B is preferred for smooth startup as in case of LMK00304.
What is "smooth startup"?

So I did some more searching on the TI forum and found something similar.
https://e2e.ti.com/support/clock-timing-group/clock-and-timing/f/clock-timing-forum/394455/lmk00306---lvds-ac-coupling-termination-method

The main point of the answer is "The LVDS driver mode of LMK0030x device needs a DC current path between Out_P and Out_N to insure proper output DC biasing and avoid in-phase clock waveforms upon output startup. ".
"insure proper output DC biasing and avoid in-phase clock waveforms upon output startup" What is this?

Both your answer and the forum I linked to explain that the LMK0030x series has problems at startup when using TYPE A.

Therefore, if the user wants to use TYPE A, it is said that it should be a double termination structure.
In order to avoid signal attenuation while using TYPE A, it is said that a large resistance about 1K should be put in the source output.

When using TYPE A, please explain in detail why it is a problem when starting the IC.
And I would like to ask if this phenomenon (improper output DC biasing and in-phase clock at startup) also applies to the LMK0030x as well as other common LVDS ICs.

Best regards

Lee HG

0 Asim Khan over 2 years ago in reply to Lee HongGyo

TI__Genius 9250 points

Hi Lee,

Smooth start up refers to "correct LVDS output clock phase at startup with settling time". Below is an explanation for it.

Situation A:

When the AC coupling caps are between the LVDS output driver and termination resistor (TYPE A) with strong bias (e.g GND, VCC, or a voltage source) the LVDS output will start in "Push pull" mode with normal output phase and expected settling time delay due to coupling cap.

Situation B:

When the AC coupling caps are between the LVDS output driver and termination resistor (TYPE A) and Vbias is a “weak” bias voltage (e.g. resistor divider), the LVDS output can start-up in “push-push” mode with differential output in-phase. This takes some time to eventually settle to correct clock with longer settling time like 1-1.5 us. You wait longer to get the correct clock output.

So if correct LVDS output clock with minimum latency is required then use DC termination with AC coupling (TYPE B).

LMK003XX and LMK1D12XX would be similar for both situations. But LMK1D12XX performs better than LMK0030XX in terms of settling time.

Duty Cycle distortion due to offset on P/N:

For LMK1D12XX Family

Duty cycle distortion doesn't happen with either configuration TYPE A or TYPE B since the architecture is different and there is no weak fixed offset on the inputs. I have verified this in the lab as well on couple of frequency as well. I couldn't do really high frequency measurement due to my scope bandwidth since all the high speed scope are highly desired by everyone :) , below are few screenshots for it.

Below setup was in cascaded mode. LMK1D12XX OUTX to CLKINX of LMK0030XX/LMK1D12XX devices with both termination schemes.

I will also try to verify using sig gen case but it would take time to update these.

TYPE A : 100 MHz

TYPE A : 200 MHz

TYPE B : 100 MHz

TYPE B : 200 MHz

I didn't see significant changes for LMK1D12XX family.

For LMK0030XX Family

For LMK0030XX family, I am verifying with different setups on the TYPE B because it seems the TYPE A is more robust than TYPE B in this case with respect to duty cycle as compared to LMK1D12XX devices. I need to verify the issue with a better scope and differential probe as well to find conclusive results.

Best,

Asim

0 Lee HongGyo over 2 years ago in reply to Asim Khan

Prodigy 220 points

Hi Asim,

Thank you again for your kind explanation and deep interest.
The LMK1D12XX Family is explained by the experimental results that there is no difference whether it is Type A or Type B.
The results of this experiment are general and good.
Thank you for your efforts.

I have an additional question about LMK1D12XX Family.
I wonder how the LMK1D12XX Family responds to Type A and Type B at low frequencies.
If possible, please provide results for approximately 1MHz and 100KHz.

We look forward to hearing from you about the LMK003xx series.

The LMK003xx series also hopes to include low frequency results.

Best regards

Lee HG

0 Asim Khan over 2 years ago in reply to Lee HongGyo

TI__Genius 9250 points

Hi Lee, Just an update. I will have results for LMK003XX by tomorrow.

Best,

Asim

0 Asim Khan over 2 years ago in reply to Asim Khan

TI__Genius 9250 points

Hi Lee,

I was able to capture few plots for LMK003XX family as well.

I don't see major duty change from 100MHz to 200MHz but TYPE A is a little better than TYPE B as you can see below. Note these terminations are on the receiver side to analyze the offset at input stage.

LMK0030XX

TYPE A: 100 MHz

TYPE A: 200 MHz

TYPE B: 100 MHz

TYPE B: 200 MHz

At lower frequency both termination scheme (at the input) have similar results as well both devices looks similar. Below are some 1MHz plots for you.

LMK1D12XX

TYPE A :

TYPE B :

LMK0030XX

TYPE A:

TYPE B:

Bottom line is, for output driver side if the good settling time is required then the termination type B is good. For receiver side, it doesn't matter in case of LMK1DXX family but may be good to use the type A in case of LMK003XX family.

Best,

Asim

0 Lee HongGyo over 2 years ago in reply to Asim Khan

Prodigy 220 points

Hi Asim,

Thanks for your experiment.

Datasheets and forums related to LMK003xx recommended using Type B, but the test result was Type A.

Why didn't the following phenomenon appear for Type A?
"improper output DC biasing and in-phase clock at startup", "smooth startup"

If it's not a typo, it looks like the experiment needs to be reviewed again.

I would like to ask about your experimental environment.
I think you experimented with the following structure.
Driver --> Termination --> Receiver --> Load
What driver did you use? for example si

I would like to inquire about the order of the experiment.
For example, power-on --> clock-on
Or clock-on-->power-on, or at the same time

I'm going to test the experiment with the environment you've been experimenting with.

Best regards

Lee HG

0 Asim Khan over 2 years ago in reply to Lee HongGyo

TI__Genius 9250 points

Hi Lee,

Why didn't the following phenomenon appear for Type A?
"improper output DC biasing and in-phase clock at startup", "smooth startup"

I don't know if you are referring this for TYPE A or TYPE B. This has been explained in one of my previous replies for both of them for the start up. This has been shown by design and a known thing previously, hence some older thread recommending the termination types for the start up.

For the receiver side, experiment is to get insights about the impact of offset on the inputs for LMK0030XX.

As I mentioned above in one of the replies, the experiment follows Sig Gen > LMK1D12XX> termination at receiver > LMK1D12XX/LMK0030XX> load.

For the second experiment, the power up routine doesn't matter since we are just checking different frequencies affect on the receiver of LMK1D12XX/LMK0030XX.

Best,

Asim

0 Lee HongGyo over 2 years ago in reply to Asim Khan

Prodigy 220 points

Hi Asim,

Regarding your inquiry, I will respond as follows.

Why didn't the following phenomenon appear for Type A?
"improper output DC biasing and in-phase clock at startup", "smooth startup"
I don't know if you are referring this for TYPE B or TYPE B.

--> This is what TI answered in various forums that this phenomenon can occur when using TYPE A.

This has been explained in one of my previous replies for both of them for the start up.
This has been shown by design and a known thing previously, hence some older thread recommending the termination types for the start up.

--> The results of this experiment came out contrary to the previous forum answers, or there seemed to be no difference in the use of TypeA / TypeB.

As I explained in the beginning, for Type A and Type B, various data sheet descriptions and actual applications are used interchangeably, and the explanation is not clear, so I asked a question about LVDS AC coupling.

In various places (I linked to it in several places above), TI replied that the LMK0030x series had the following symptoms, so TYPE B was recommended.
"improper output DC biasing and in-phase clock at startup", "smooth startup"

However, there is no difference between TYPE A / TYPE B in your experimental results.
You are recommending TYPE A. Other forums recommend using TYPE B.
Why is there a difference between the answers in other forums and the results of this experiment?
I'm curious about this part.
Therefore, I wonder how the experiment was conducted this time.
Please explain the experimental environment and experimental conditions that you conducted this time.

Best regards

Lee HG

0 Asim Khan over 2 years ago in reply to Lee HongGyo

TI__Genius 9250 points

Hi Lee,

Yes, there are multiple version of termination types in the datasheet and different forums suggested.

And the suggested TYPE B termination is usually for settling time. Its also recommended to add a 1K or 500 ohm resistor before the decoupling caps with TYPE A termination to get the added advantage of TYPE B . This is on the driver output side for both LMK00304 and LMK1D12XX devices.

For your question about the duty cycle distortion was related to the receiver side termination. And we experimented for that and LMK1DXXXX family showed no change with either termination type on the receiver side of these devices. While its better to use TYPE A with LMK0030XX family on the receiver since its showing better results for duty cycle.

This was conducted as I listed in one my previous comments.

Hope this makes it clear. Let me know if you have any further questions.

Best,

Asim

Clock & timing

Clock & timing forum

LMK00308: Question about LVDS AC coupling