Designer’s Guide Community

taotaohai

New Member

Offline

Posts: 5

Please help me expert analysis of this spectrum
Jan 10^th, 2009, 6:53am

Please help me expert analysis of this spectrum, the low-frequency noise sources.

First of all, my circuit is a multi-stage cascade, with a number of feedback loop of the limiting amplifier. Design

goal is to achieve transfer rates of 20Gb / s data rate.

In the tape-test process, found before the frequency of 15GHz , there is a lot �noise , a lot of clutter. The

pseudo-random signal to 20Gb / s accession �, the eye diagram poor.

The chart is the circuit output frequency spectrum only with power, not yet to join the ac signals .And the spectrum

is an average of all frequencies on the map after.

I guess there may be two reasons: 1, multi-ring structure of the self-excited, resulting in a lot of points on the

frequency noise spectrum; 2,problem in the layout, substrate noise is the main reason, there are a lot of noise

acoustic coupling, or incoming power supply noise coupling.

But this is only conjecture, please understand the spectrum of friends to help me analyze.

Thank you very much! ! Want to know more about the circuit, my email: taohaizhou@gmail.com
Please!

Back to top

SCREN003.GIF

IP Logged

pancho_hideboo

Senior Fellow

Offline

Posts: 1424
Real Homeless

Re: Please help me expert analysis of this spectrum
Reply #1 - Jan 10^th, 2009, 7:53pm

Why do you post your question to "The Designer's Guide Community Forum ≫ Design ≫ Mixed-Signal Design" ?
Your question should be posted to "The Designer's Guide Community Forum ≫ Design ≫ High-Speed I/O Design or RF Design"

taotaohai wrote on Jan 10^th, 2009, 6:53am:

First of all, my circuit is a multi-stage cascade, with a number of feedback loop of the limiting amplifier.
Design goal is to achieve transfer rates of 20Gb/s data rate.
In the tape-test process, found before the frequency of 15GHz, there is a lot �noise, a lot of clutter.
The pseudo-random signal to 20Gb/s accession, the eye diagram poor.

taotaohai wrote on Jan 10^th, 2009, 6:53am:

I guess there may be two reasons:
� 1, multi-ring structure of the self-excited, resulting in a lot of points on the frequency noise spectrum;
� 2, problem in the layout, substrate noise is the main reason,
there are a lot of noise acoustic coupling, or incoming power supply noise coupling.

But this is only conjecture, please understand the spectrum of friends to help me analyze.

I can't understand what you mean by "the tape-test process". But I think your problem is following.

Your limiting amplifier for 20Gbps transmission is noisy especially for under 15GHz frequency, so eye diagram is dirty.
Here you want to know this cause.
Is this correct ?

taotaohai wrote on Jan 10^th, 2009, 6:53am:

Please help me expert analysis of this spectrum, the low-frequency noise sources.

What do you mean ?

taotaohai wrote on Jan 10^th, 2009, 6:53am:

The chart is the circuit output frequency spectrum only with power, not yet to join the ac signals.
And the spectrum is an average of all frequencies on the map after.

What do you mean by "chart" and "map" ?
I think your spectrum analyzer is Agilent PSA or ESA Series Spectrum Analyzer.
You use detector mode of "sample" and RBW=3MHz.
What do you mean by "an average of all frequencies on the map after." ?

Your spectrum is Center=13.26GHz and Span=26.49GHz.
Is there any remarkable spectrum dip for more low frequency ?

taotaohai wrote on Jan 10^th, 2009, 6:53am:

Want to know more about the circuit

I think main cause is oscillation due to feedback.

What number of stages do you use for limiting amplifier ?
What value is total gain ?
What value is one unit gain ?
What value do you expect as cutoff frequency of limiting amplifier ?

With no input signal, none of amplifiers is saturated ?
If gain of limiting amplifier is very high, posterior amplifiers will be saturated even though input signal is not fed.

What is purpose of your feedback ?
Only broaden bandwidth ?
For canceling offset ?

Back to top

« Last Edit: Jan 10^th, 2009, 9:55pm by pancho_hideboo »

Kita━━━━━━(ﾟ∀ﾟ)━━━━━━ !!!!!
http://www7.plala.or.jp/ungeromeppa/flash/kita.html
http://www.youtube.com/watch?v=mjIxGh55bMM&feature=related

IP Logged

taotaohai

New Member

Offline

Posts: 5

Re: Please help me expert analysis of this spectrum
Reply #2 - Jan 10^th, 2009, 10:10pm

pancho_hideboo wrote on Jan 10^th, 2009, 7:53pm:

Why do you post your question to "The Designer's Guide Community Forum ≫ Design ≫ Mixed-Signal Design" ?
Your question should be posted to "The Designer's Guide Community Forum ≫ Design ≫ High-Speed I/O Design or RF Design"

taotaohai wrote on Jan 10^th, 2009, 6:53am:

First of all, my circuit is a multi-stage cascade, with a number of feedback loop of the limiting amplifier.
Design goal is to achieve transfer rates of 20Gb/s data rate.
In the tape-test process, found before the frequency of 15GHz, there is a lot �noise, a lot of clutter.
The pseudo-random signal to 20Gb/s accession, the eye diagram poor.

taotaohai wrote on Jan 10^th, 2009, 6:53am:

I guess there may be two reasons:
� 1, multi-ring structure of the self-excited, resulting in a lot of points on the frequency noise spectrum;
� 2, problem in the layout, substrate noise is the main reason,
there are a lot of noise acoustic coupling, or incoming power supply noise coupling.

But this is only conjecture, please understand the spectrum of friends to help me analyze.

I can't understand what you mean by "the tape-test process". But I think your problem is following.

Your limiting amplifier for 20Gbps transmission is noisy especially for under 15GHz frequency, so eye diagram is dirty.
Here you want to know this cause.
Is this correct ?

taotaohai wrote on Jan 10^th, 2009, 6:53am:

Please help me expert analysis of this spectrum, the low-frequency noise sources.

What do you mean ?

taotaohai wrote on Jan 10^th, 2009, 6:53am:

The chart is the circuit output frequency spectrum only with power, not yet to join the ac signals.
And the spectrum is an average of all frequencies on the map after.

What do you mean by "chart" and "map" ?
I think your spectrum analyzer is Agilent PSA or ESA Series Spectrum Analyzer.
You use detector mode of "sample" and RBW=3MHz.
What do you mean by "an average of all frequencies on the map after." ?

Your spectrum is Center=13.26GHz and Span=26.49GHz.
Is there any remarkable spectrum dip for more low frequency ?

taotaohai wrote on Jan 10^th, 2009, 6:53am:

Want to know more about the circuit

I think main cause is oscillation due to feedback.

What number of stages do you use for limiting amplifier ?
What value is total gain ?
What value is one unit gain ?
What value do you expect as cutoff frequency of limiting amplifier ?

With no input signal, none of amplifiers is saturated ?
If gain of limiting amplifier is very high, posterior amplifiers will be saturated even though input signal is not fed.

What is purpose of your feedback ?
Only broaden bandwidth ?
For canceling offset ?

For your questions,I will express to you one by one.
"the tape-test process"is means that,in the process of testing of the chip.
YES! U R right! There is really much noise under 15GHz,which makes my eye-diagram is dirty!

Please help me expert analysis of this spectrum:that's may be a mistake.I just want someone who understands could help me find the reasons for this spectrum.

Just as you said,my spectrum analyzer is Agilent PSA or ESA Series Spectrum Analyzer.
I used detector mode of "sample" and RBW=3MHz.my spectrum is Center=13.26GHz and Span=26.49GHz.
"an average of all frequencies on the map after." ,in frequency domain,in any point,the spectrum is not stable.So I get the average of every point ,which seems relatively stable just shows in the chart.

I guess the two reasons. For my circuit,there are 12 stages,8 loops. the total gain is 49dB,every stage is about 4--5dB.Before tape-out,simulation shows that the cutoff frequency(you mean the GBW?) of the LA is about 36GHz,and -3dB BW is 17.5GHz.
"With no input signal, none of amplifiers is saturated ?" Do U mean the input noise makes the amp saturated?may be.
The purpose of my feedback is two facets:one of the loops is for canceling offset,the others are for broaden BW.

Back to top

IP Logged

pancho_hideboo

Senior Fellow

Offline

Posts: 1424
Real Homeless

Re: Please help me expert analysis of this spectrum
Reply #3 - Jan 12^th, 2009, 3:30am

Don't use unfamiliar abbreviations such as "U R right", "Do U", "LA".
And use industrial standard words which everyone can understand.

taotaohai wrote on Jan 10^th, 2009, 10:10pm:

"the tape-test process" is means that,in the process of testing of the chip.

I don't think "the tape-test process" is a standard term which everyone can understand.

taotaohai wrote on Jan 10^th, 2009, 10:10pm:

Just as you said,my spectrum analyzer is Agilent PSA or ESA Series Spectrum Analyzer.

There is big difference between PSA Spectrum Analyzer and ESA Spectrum Analyzer.

taotaohai wrote on Jan 10^th, 2009, 10:10pm:

I used detector mode of "sample" and RBW=3MHz.my spectrum is Center=13.26GHz and Span=26.49GHz.

You should use "positive peak" detector mode.

taotaohai wrote on Jan 10^th, 2009, 10:10pm:

"an average of all frequencies on the map after." ,in frequency domain,in any point,the spectrum is not stable.
So I get the average of every point ,which seems relatively stable just shows in the chart.

Your spurious peaks in spectrum vary with time ?
Frequencies of spurious peaks are constant ?
Amplitudes of spurious peaks are constant ?

taotaohai wrote on Jan 10^th, 2009, 10:10pm:

I guess the two reasons.

There is no description about reason in the following sentences. You refer first your post ?

taotaohai wrote on Jan 10^th, 2009, 10:10pm:

For my circuit,there are 12 stages,8 loops.
the total gain is 49dB,every stage is about 4--5dB.
Before tape-out,simulation shows that the cutoff frequency(you mean the GBW?) of the LA is about 36GHz,and -3dB BW is 17.5GHz.

I mean cutoff frequency, fc as -3dB BW.
If I mean GBW, I refer it as transition frequency, fT.

In your simulation, how do you check stability of cascaded amplifier having multiple feedbacks ?

taotaohai wrote on Jan 10^th, 2009, 10:10pm:

"With no input signal, none of amplifiers is saturated ?"
Do U mean the input noise makes the amp saturated?may be.

Yes. But your total gain of limiting amplifier is not so large, so none of amplifiers could be saturated with no input signal.

taotaohai wrote on Jan 10^th, 2009, 10:10pm:

The purpose of my feedback is two facets:one of the loops is for canceling offset,the others are for broaden BW.

Very common.

taotaohai wrote on Jan 10^th, 2009, 10:10pm:

Please help me expert analysis of this spectrum:that's may be a mistake.I just want someone who understands could help me find the reasons for this spectrum.

Specrum you attached is fairly old, almost one year has passed after you measured spectrum.
Why have you left your problem unresolved ?

Try to vary power supply voltage and observe behavior of spurious peaks in spectrum if power supply of your amplifier is not regulated by on-chip regulator.
Frequencies of spurious peaks are constant ?
If you re-observe spectrum more preciously with peak detector mode, is there any periodicity in spectrum ?

Back to top

« Last Edit: Jan 12^th, 2009, 7:27am by pancho_hideboo »

Kita━━━━━━(ﾟ∀ﾟ)━━━━━━ !!!!!
http://www7.plala.or.jp/ungeromeppa/flash/kita.html
http://www.youtube.com/watch?v=mjIxGh55bMM&feature=related

IP Logged

taotaohai

New Member

Offline

Posts: 5

Re: Please help me expert analysis of this spectrum
Reply #4 - Jan 13^th, 2009, 7:02am

Thank you very much for your reminder.
In the process of designing the circuit,I simulated the phase margin of every loop,and almost every loop's PM >45�,so in the subsequent work,I did not spend much time in this problem.

Now,I guess the reason of the problem based on PM(phase margin).Just as I said above,the PM almost all were about 45�,if every loop oscillated,there must be many frequencies points of spurious peaks,so the spectrum.
By the way ,the time show in the chart is wrong.Maybe the instrument had not be seted. Because I test the chip just about 5 or 6 days ago.
I will test the chip again as your suggestion.
Thank you again!

Back to top

IP Logged