Designer’s Guide Community

neoflash

Community Fellow

Offline

Mixed-Signal
Designer

Posts: 397

ADC T/H Linearity and No. of Bits
Jul 10^th, 2007, 8:12am

Some literature will state their T/H circuits are designed for 8-bit linearity. Which simulation they run and which spec they judge whether it is 8-bit linearity?

In my opinion, since T/H circuit input and output are all analog circuits. No INL/DNL spec can be put there. Are they measuring SDR and compare it with SNR requirement of certain No. of bits?

Back to top

IP Logged

boe

Community Fellow

Offline

Posts: 615

Re: ADC T/H Linearity and No. of Bits
Reply #1 - Jul 10^th, 2007, 8:36am

Hi neoflash,
neoflash wrote on Jul 10^th, 2007, 8:12am:

In my opinion, since T/H circuit input and output are all analog circuits. No INL/DNL spec can be put there.

Why not? Measure the DC characteristics and compare the linearity error with the LSB of the 8-bit ADC which follows it.

neoflash wrote on Jul 10^th, 2007, 8:12am:

Are they measuring SDR and compare it with SNR requirement of certain No. of bits?

Also possible, but not really an INL/DNL spec, is it?

BOE.

Back to top

IP Logged

Sid Community Member Offline Posts: 35	Re: ADC T/H Linearity and No. of Bits Reply #2 - Jul 10^th, 2007, 2:28pm Hi Neoflash, You can also run an FFT (pick a settled point in your analog output) and see if the SFDR compares with an ideal N-bit quantizer. SFDR-ideal = 9N (approx) (See "Spectral Spurs due to Quantization in Nyquist ADCs" by Pan and Abidi) -Sid
Back to top	IP Logged

neoflash Community Fellow Offline Mixed-Signal Designer Posts: 397	Re: ADC T/H Linearity and No. of Bits Reply #3 - Jul 11^th, 2007, 5:00am boe wrote on Jul 10^th, 2007, 8:36am: Hi neoflash, Why not? Measure the DC characteristics and compare the linearity error with the LSB of the 8-bit ADC which follows it. BOE. I do not think there is any DC characteristics meaningful for T/H circuit, which is essentially a dynamic circuit.
Back to top	IP Logged

neoflash

Community Fellow

Offline

Mixed-Signal
Designer

Posts: 397

Re: ADC T/H Linearity and No. of Bits
Reply #4 - Jul 11^th, 2007, 5:02am

Sid wrote on Jul 10^th, 2007, 2:28pm:

Hi Neoflash,

You can also run an FFT (pick a settled point in your analog output) and see if the SFDR compares with an ideal N-bit quantizer.

SFDR-ideal = 9N (approx)
(See "Spectral Spurs due to Quantization in Nyquist ADCs" by Pan and Abidi)

-Sid

I do not think it is very practical way, since transient simulation does not show random noise.
However, seperate calculation is possible. Any literature doing this analysis for us? thanks.

Back to top

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: ADC T/H Linearity and No. of Bits
Reply #5 - Jul 11^th, 2007, 6:05am

Hi Neo,

Since you are simply interested in linearity of the S&H, it is possible to simulate this using transient analysis, followed by FFT.
There are 2 components to this problem: DC nonlinearity & AC nonlinearity.

While the S&H is a dynamic circuit and it does not make much sense to do a DC nonlinearity measurement, it is sometimes useful
to do and it is certainly possible to do so:

1. Apply a DC input and sample this with your S&H. Plot transfer curve to see any DC nonlinearity in your S&H, for instance due
to nonlinearity in the S&H amplifier if you use one. If the design is not done well, then even this simulation may show you something.

2. Apply fullscale AC input at maximum frequency of interest and look at the harmonic distortion from the transient analysis. You
need to setup your simulation carefully for this, using suitable strobing, or else you will only get garbage. Two-tone or mult-tone
IMD can also be done this way.

Regards
Vivek

Back to top

IP Logged

boe

Community Fellow

Offline

Posts: 615

Re: ADC T/H Linearity and No. of Bits
Reply #6 - Jul 11^th, 2007, 6:49am

OK, I'm biased on this: I simply don't like to measure time domain or DC parameters in the frequency domain unless it can't be avoided.
So if I want to see dynamic effects, I'd rather try to get a "DC transfer function" from a transient simulation than do something in the frequency domain. But of course it also depends on what's relevant for your application...

BOE

Back to top

IP Logged

neoflash

Community Fellow

Offline

Mixed-Signal
Designer

Posts: 397

Re: ADC T/H Linearity and No. of Bits
Reply #7 - Jul 11^th, 2007, 6:55am

vivkr wrote on Jul 11^th, 2007, 6:05am:

Hi Neo,

2. Apply fullscale AC input at maximum frequency of interest and look at the harmonic distortion from the transient analysis. You
need to setup your simulation carefully for this, using suitable strobing, or else you will only get garbage. Two-tone or mult-tone
IMD can also be done this way.

Regards
Vivek

Vivek, thanks for the reply. By the way, what do u mean by suitable strobing?

Back to top

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: ADC T/H Linearity and No. of Bits
Reply #8 - Jul 12^th, 2007, 3:00am

Hi Neo,

I just meant that you need to strobe the outputs at the point where they are settled, using the "strobeperiod" feature
in Spectre. Using coherent sampling is also advisable as it improves the resolution of the FFT. With this approach,
you often can get good results even with very few samples (128-256 points are mostly enough).

After this, I usually look at both the input and output spectra.

Regards
Vivek

Back to top

IP Logged

HdrChopper

Community Fellow

Offline

Posts: 493

Re: ADC T/H Linearity and No. of Bits
Reply #9 - Jul 14^th, 2007, 9:54pm

Hi Neoflash,

A probably faster method to what Vivek is proposing is to run a pss analysis. The harmonic distortion can be readily obtain as a result of this analysis. If I�m correct, with this approach you do not even have to worry about strobing appropriately the signal in order to get accurate results.
The distortion will give you the ENOB for the S&H linearity.

Tosei

Back to top

Keep it simple

IP Logged