High Frequency Diff to SE Conversion

[revky]

Hi,

I have a question about the best way to combine to short pulses with small amplitude to obtain a monocycle pulse.

I am using a fast CML Comparator to generate short pulses (roughly 0.5 - 1nS) which i differentiate both complementary outputs to remove the DC part the and the positive edge i delay (using TL) by the pulse width in order to offset them. What is are some of the ways to recombine these pulses as I need to feed them into a MMIC amplifier?

Amplitude of these pulses is relatively small at this stage around 200-300mV

Many Thanks

 

[BigBoss]

I don't know the frequency but a Differential Pair and Current Sink/Source Output stage will combine your signals.
In fact, a Combiner Circuit that is realized with OPAMPs may serve you well.
However, the frequency is pretty important.

 

[revky]

Hi,

It has to be passive, the pulsewidth is 500ps with the rising/falling edge 80-90ps. No op-amp will do that so it has to be some rf technique

 

[dick_freebird]

Balun?

 

[BigBoss]

Hi,

It has to be passive, the pulsewidth is 500ps with the rising/falling edge 80-90ps. No op-amp will do that so it has to be some rf technique

So I guess those signals are in GHz range, right ? If they are so, your task is a little bit difficult because Phases are very important here and it's not possible to expect 100% performance with passive circuits such as Balun. Because they have Phase Error and as well as Amplitude Error.
You can realize such circuit with discrete component using with RF/Microwave Transistors but this needs highly Design Experiences.
I believe a L-C Balun will serve you better. But it should be carefully designed and the layout must be perfect. Be careful about the components that you'll use.

 

[D.A.(Tony)Stewart]

Hi,

I have a question about the best way to combine to short pulses with small amplitude to obtain a monocycle pulse.

I am using a fast CML Comparator to generate short pulses (roughly 0.5 - 1nS) which i differentiate both complementary outputs to remove the DC part the and the positive edge i delay (using TL) by the pulse width in order to offset them. What is are some of the ways to recombine these pulses as I need to feed them into a MMIC amplifier?

Amplitude of these pulses is relatively small at this stage around 200-300mV

Many Thanks

Can you define the function more clearly rather than what you did?

Vin = 200~300 mVpp , Vdc offset= ?
PW50 = {min, typ, max or Nom. +/-%} (50% amplitude width)
Purpose: 1-shot 750 ps +/- 250 ( i.e. monocycle Is that like a unicycle ? Does it repeat ? What's it for?
trailing or leading edge ??

Latency = ??
Rise/Fall time = ??
Voltage or current source?

Output: CML ?? or single or differential? ? or define logic levels like 1V @ 50 Ohms or
Supplies avail??
Operating ambient temp range?

Fairchild was the first to make sub-ns ECL one-shots in the 70's when I started work.
Today 1 picosecond 1-shots are used in lasers.

 

[revky]

Can you define the function more clearly rather than what you did?

Vin = 200~300 mVpp , Vdc offset= ?
PW50 = {min, typ, max or Nom. +/-%} (50% amplitude width)
Purpose: 1-shot 750 ps +/- 250 ( i.e. monocycle Is that like a unicycle ? Does it repeat ? What's it for?
trailing or leading edge ??

Latency = ??
Rise/Fall time = ??
Voltage or current source?

Output: CML ?? or single or differential? ? or define logic levels like 1V @ 50 Ohms or
Supplies avail??
Operating ambient temp range?

Fairchild was the first to make sub-ns ECL one-shots in the 70's when I started work.
Today 1 picosecond 1-shots are used in lasers.

I take the output of a fast CML Comparator ( ADCMP572) and I get a 200mV pulse in 50Ohm load. Width FWHM of 500ps. Yes it is a repeating monocycle with a frequency of 100MHz. Rise fall times already mentioned around 60-80pS and since its CML the source really is current steering logic.

As mentioned, my problem is not generating the pulses is combining them. as I have these pulses on parallel circuits, and I want to put them into a single MMIC

 

[revky]

So I guess those signals are in GHz range, right ? If they are so, your task is a little bit difficult because Phases are very important here and it's not possible to expect 100% performance with passive circuits such as Balun. Because they have Phase Error and as well as Amplitude Error.
You can realize such circuit with discrete component using with RF/Microwave Transistors but this needs highly Design Experiences.
I believe a L-C Balun will serve you better. But it should be carefully designed and the layout must be perfect. Be careful about the components that you'll use.

My understanding was baluns convert signals that are 180 degree out of phase into SE. Here I just need to superimpose the two pulses into the same amplifier input

 

[BigBoss]

My understanding was baluns convert signals that are 180 degree out of phase into SE. Here I just need to superimpose the two pulses into the same amplifier input

OK, understood. So you will work in Time Domain.
But which Logical Operation you intend to ? I mean you will either ADD these signal or SUBSTRACT them.
A passive Circuit does not serve this purpose. BalUn's are Frequency Domain components and they ADD the signals in Frequency Domain by 180° Degree Phase Shift.
A Frequency/Phase discriminator may serve as a Phase Comparator @100 MHz but it's not Passive.
There is no passive elements that will serve your purpose to my knowledge. Someone may know better.
Another chance is to use Power Combiner but not sure in Time Domain.

 

[D.A.(Tony)Stewart]

Balun - Wikiwand

A balun is an electrical device that allows balanced and unbalanced lines to be interfaced without disturbing the impedance arrangement of either line. A balun ...

www.wikiwand.com

This is what I understand.

Combine two consecutive 100 MHz rep rate pulses 200 mV (0-pk) x 500 ps @ 100 mV with DC offset
- to remove the leading edge so you can see the difference in the trailing edge of consecutive pulses.
- you have differential outputs are passive differentiated to remove DC (and leading edge ?)

Other important variables may include;
DL path loss, amplitude variation, PW variation, jitter, amplitude matching, sync error one leading edge, pulse jitter due to AM, spectral density of source and essential BW.

You have a differential Comparator so you can sum analog signals by the difference of of complementary signals one being delayed -inverted but you want to cancel the lead edge so delayd non-inverted.

You can use a splitter and balun in either direction. The Balun raises the CM impedance much higher than the Zo so the differential out is more balanced or you can convert single-ended to differential and visa versa.

So you can combine using Differential inputs or use Balun or use a 3 dB splitter as a combiner. rated from say 10% to 10x

We once used cascaded LC all pass delay lines in 2 ns increments to measure window margin instead of eye pattern for 10 MHz data in the 80's in an automated test set.

Maybe you can draw some signals in a block diagram to show the measurement you want out with resolution & tolerances with the above variables that matter..

 

[revky]

Balun - Wikiwand

A balun is an electrical device that allows balanced and unbalanced lines to be interfaced without disturbing the impedance arrangement of either line. A balun ...

www.wikiwand.com

This is what I understand.

Combine two consecutive 100 MHz rep rate pulses 200 mV (0-pk) x 500 ps @ 100 mV with DC offset
- to remove the leading edge so you can see the difference in the trailing edge of consecutive pulses.
- you have differential outputs are passive differentiated to remove DC (and leading edge ?)

Other important variables may include;
DL path loss, amplitude variation, PW variation, jitter, amplitude matching, sync error one leading edge, pulse jitter due to AM, spectral density of source and essential BW.

You have a differential Comparator so you can sum analog signals by the difference of of complementary signals one being delayed -inverted but you want to cancel the lead edge so delayd non-inverted.

You can use a splitter and balun in either direction. The Balun raises the CM impedance much higher than the Zo so the differential out is more balanced or you can convert single-ended to differential and visa versa.

So you can combine using Differential inputs or use Balun or use a 3 dB splitter as a combiner. rated from say 10% to 10x

We once used cascaded LC all pass delay lines in 2 ns increments to measure window margin instead of eye pattern for 10 MHz data in the 80's in an automated test set.

Maybe you can draw some signals in a block diagram to show the measurement you want out with resolution & tolerances with the above variables that matter..

Hi,

I drew a diagram. Basically I have two short pulse on different tracks (pink and green in attached pic) I delay one of them and I want to combine them into a single ended signal (blue) to feed them into the same MMIC amplifier. I am unsure of solution for the combiner ( i was thinking there is perhaps some pin diode rf mixer/autoswitch topology)

 

[D.A.(Tony)Stewart]

I suppose you may have a splitter for your delay line. It looks like this on EBAY. OR Amazon. then you can combine the same at the other end. but you lose 3.5dB on each with 20 dB isolation.

Otherwise if you need gain, use the differential amplifier , AC if there is DC offset and use the inverting input with 4x gain with two input Rin- resistors to combine.

I am thinking of a solution that works for 3.3 or 5V as +/-V. You might want to use a gain of 2 or 4 for the flattest f. response. Normally you might use a differential ADC, but if single-end use a balun. " To interface with single-ended test equipment, external baluns are required between the EVM output ports and the test equipment."

The LMH5401 is a fully-differential amplifier (FDA) configurable with external resistors for noise gain greater than 2 V/V or 6 dB (GBP = 8 GHz).

But one cannot design this without knowing your s-parameters (source and load spectral impedance) and what result you expect to measure.

 

[revky]

An LMH won't work as the output is still differential and the bandwidth is far too low. I don't need gain that's what my MMIC amplifier is for. I guess there is no better solution than a balun or power splitter.

 

[D.A.(Tony)Stewart]

What is your rise time spec?

0.35/ 2 GHz has half power at 175 ps 10~90% risetime.

At these speeds you ought to be using differential to avoid CM or unbalanced threshold errors.

 

[revky]

What is your rise time spec?

0.35/ 2 GHz has half power at 175 ps 10~90% risetime.

At these speeds you ought to be using differential to avoid CM or unbalanced threshold errors.

I need single ended for the MMIC amplifier which cannot accept differential.

 

[biff44]

i am not following your circuit.
if you differentiate a positive going square wave leading edge, you get a positive going spike. If you differentiate the negative going square wave leading edge, you get a negative going spike. If you delay the first and recombine the two you get two spikes exactly canceling each other out! i.e. an all-stop filter.

maybe explain WHAT your system is intended to do, rather than immediately talk about the hardware implementation.....

 

[revky]

i am not following your circuit.
if you differentiate a positive going square wave leading edge, you get a positive going spike. If you differentiate the negative going square wave leading edge, you get a negative going spike. If you delay the first and recombine the two you get two spikes exactly canceling each other out! i.e. an all-stop filter.

maybe explain WHAT your system is intended to do, rather than immediately talk about the hardware implementation.....

This is exactly from my initial post:

I am using a fast CML Comparator to generate short pulses (roughly 0.5 - 1nS) which i differentiate both complementary outputs to remove the DC part the and the positive edge i delay (using TL) by the pulse width in order to offset them. What is are some of the ways to recombine these pulses as I need to feed them into a MMIC amplifier?

I mention there i delay one positive edge which means it doesn't cancel out.

I have also provided a drawing on what I am trying to achieve on my post on January 15 as a drawing attached: I am trying to generate a bipolar pulse from the complementary pulses of a cml gate by differentiating both outputs and delaying one.