Lecroy incorporates 12 bit ADC in new line of scopes

JW schrieb:

Hello,

I did not find a specification if it is possible to get real 12 bit resolution and 4 GHz bandwidth and 40 Gigasamples/s together. Maybe the full 12 bit resolution is possible only with much slower samplerates, with full samplerate the resolution is only 8 or 10 bits.

If I did not find the respective lines in the specs, it is my fault. ;-)

Bye

..

LeCroy collaborated with SP Devices on the product line, uses SP Devices digitizer technology, the sample rate is real:

Hmmm, high speed acquisition... but 10mn to not even load the page!

--
Thanks,
Fred.

The claim of 12 bits at 4GHz only applies to DC and low frequency.

ADC's are jitter limited above a few hundred MHz. For example, 12 bits requires less than 100 femtoseconds above 709 MHz. See page 7 of "

"

Walt Kester of Analog Devices discusses this a bit more in "Aperture Time, Aperture Jitter, Aperture Delay Time...", at

Maxim has a jitter calculator at

To get 12 bits at 4GHz would require less than 9.71 femtoseconds jitter. No readily available clock oscillators can reach that. Even if they could, the jitter in the signal being tested would far exceed it.

The only way to improve the situation is by averaging. However, the 9.71 femtosecond requirement applies also to drift. The system would likely drift more than that during the measurement, which makes the result invalid.

The only way to obtain that kind of amplitude resolution at these freequencies is with Binary Sampling. This has the unique property of rejecting amplitude and timing noise in the signal, and it can complete the measurement several orders of magnitude faster than with conventional averaging.

I used to have a fairly complete description on my web site, but I have since discovered some very important applications that are now under development. In the meantime, I have removed most of the information until I can get the patent applications filed.

Regards,

Mike Monett

Link is broken due to wrap. Here is the tinyurl link

Regards,

Mike Monett

Their gear still runs Windoze. The company is headed by a two-faced asswipe. Ask Larkin.

35

They're doing the 10Gs/s per channel regardless of the measurement signal BW. The idea of mux'ing multiple phase clocked ADCs is not new- the meat is in their patented 'reconstructor'. IIRC Tek has been doing similar since the early 90s or late 80s.

Oh- you mean he is an EE ...?

It doesn't matter how they sample the data and put it back together. They are still limited by the jitter in the sample and signal clocks. The figures of 12 bits and 4GHz they claim are not attainable.

Most low cost digitizing scopes interleave multiple samples. For example, the Instek GDS-2204 uses four Analog Devices AD9288 Dual 8-Bit A/D Converters running at 125MSPS, which is 25% above the maximum spec. Eight converters are interleaved to give 1GS/sec for a single channel. More info here:

This raises serious questions about the waveform sampling accuracy and long-term stability due to the extreme requirements of interleaving at these frequencies. For example, see the following article in the August

2003 issue of Analog Dialogue, "Advanced Digital Post-Processing Techniques Enhance Performance in Time-Interleaved ADC Systems" at

The Rigol DS5052E interleaves 10 AD9288's clocked at 100MHz to give

1GS/sec. See "EEVblog #37 - Rigol DS1052E Oscilloscope Teardown", at

Lecroy also uses a different technique of breaking the frequency spectrum into sections, then recombining them. I hate to think of the problems they will encounter trying to keep that mess stable over long periods, let alone the errors caused by clock and signal jitter.

I also have discovered a disrepancy between the ENOB values stated in the Maxim jitter calculator at

and the Reinhardt article at

The Reinhardt article appears to do the calculations at 10 dB below the maximum amplitude the ADC is capable of. The numbers in the article appear to be off by a factor of two, but the ratio is not constant for different frequencies and ENOB. I'll check further, but at the moment, the Maxim calculation appears to be correct.

Regards,

Mike

Not so sure about that. I was once told that we could never get 4x the sample rate out of four ganged 12bit ADCs without the ENOB taking a big hit. And then we did it.

I have that scope and must say that it is rather stable. What does drift off a bit is the offset in the most sensitive mV/div settings. Timing-wise or during fast aquisition, no complaints. I also found it to be a bit better in noise than some "name brand" scopes.

I have designed an interleaved ADC systems. One trick I found out is not to derive the input to corrective measures from FFT but to do it strictly in time domain. Otherwise it takes forever to get good results. Or if you want super good results it might not converge and then you have to do some auto-backoff or you are screwed. One must adjust for offset, amplitude and aperture delay. Servo in those adjusters so they become linear also helps a great deal in speeding this up.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/