ADS5232: Pulse sampling

P.S. an update:

We changed the clock source the observed the following for a 10us pulse. Please notice the strange fluctuation within a pulse and the again the strange behaviour during the intra pulse period. 

To further investigate, we captured a single tone with the original clock source and got the following (some sort of amplitude variation over time)

With the new clock source, the following waveform was observed:

raw data capture files for all three are attached. 

Sorry for the data overload but I hope it will assist in understanding the problem:

Channel_1_10us_EB.txtChannel_1_15MHzIF_CWsingletone_EA.txtChannel_1_15MHzIF_CWsingletone_EB.txt

3566.Channel_1_15MHzIF_CWsingletone_EA.txt

0361.Channel_1_10us_EB.txt

Thanks again,

Hi Jim,

     Inadvertently I clicked the wrong button. 

We have a 200MHz/2GSps scope with single ended probes. I tried looking at the output of the transformer but could't get anything conclusive. We do not have differential probes. 

Regarding the schematic, I will have to check with our dept for that. 

I think though the problem is probably not just with pulses. If you see my second post the long term (64K samples) capture shows the CW signal also showing some problems (you can have a look at the raw data as well). 

Thanks for your quick reply,

SM

SM,

I just wanted to see how all of the ADC pins were connected. I do not need the entire schematic.

Jim

Hi Jim,

    Sincere apologies, I will need to check with the concerned dept, we are not allowed share any schematic/section without consent. 

May I ask if you would like to know something specific about the connections then I can answer them. 

Regards,
SM.

Hi Jim,

      I think we made some progress. There are two ADC ICs on the board and we turned OFF the clock to one of the ADCs and the output looks much cleaner now. 

This is the capture of 20us pulses:

This is the capture of a 150us pulse:

There is a question here: why is the shape at the starting and after ending of the pulse like the way it is? Again, the pulse does not show this behaviour (when captured using a vector signal analyzer). 

2) This improvement is by turning off the clock to the second ADC chip (both were being fed 60MHz). In order to proceed further,

1) Is it because the clocks are "messing" with each other directly (some sort of coupling)

2) or some sort of ADC noise getting coupled

in other words, how do we fix it :)

Thanks again for looking into our matter,

The data files are also attached in case you want to have a closer look:

Channel_1_166us_15MHzIF_chBclkOFF.txtChannel_1_20usON_40usOFF_15MHzIF_60MSps_chBclkOFF.txt

SM,

I was wondering how you had the following pins connected:

REFt, REFb, SEL, INT/EXT

How are the internal registers programmed? If not using the SPI, how are the SPI pins connected?

Can you lower the pulse amplitude and check the results? Just want to make sure the ADC is not saturated.

Regards,

Jim

Hi Jim,

        We are using the internal reference with REFt/b connected as per datasheet i.e., series resistor and capacitor to ground.

The device is being programmed through SPI from a microcontroller. 

I think this might be a very good point. The SPI lines for both the ICs are connected in a bus configuration from a microcontroller and it is quite possible that there might be some interference through them. 

I will try putting some ferrites in series on these lines and check. 

We decreased the pulse amplitude by 6dB but the shape remains the same. 

I might be too naive about the shape but doesn't it look like a capacitor is getting charged and discharged by the pulse (especially the trailing end of the pulse).

SM>

SM,

After consulting with a coworker, we think the balun and the series 0.1uf cap are causing the envelope issue.

We would suggest using a much large series cap, around 10uF, and a differential amplifier, such as a LMH3401 or LMH5401 in place of the balun/transformer. This may not be an easy option for you, but we do not see any other way around this due to you input.

You could test this by using an amplifier EVM and routing the outputs to the ADC by connecting wires from the EVM to the pads of the transformer after it has been removed.

Regards,

Jim 

Thanks for the input Jim. We will try that. 

Meanwhile we introduced the ferrites as discussed and now the "cross talk" issue is resolved. Thank you for that. 

We are trying out sampling inputs signals at two IFs; they are 15MHz and 75MHz. 

The following is the data sampled 60MSps:

15MHz IF, CW (can seem some amplitude taper)

15MHz IF, 20us pulses (some amplitude taper can be noticed)

75MHz CW:

75MHz IF, 20us pulses: 

Please advise. (raw data files are attached), We initially that this might be because of sub-sampling but we generated the "same" signal in Matlab and it does exhibit this behaviour. 

Also, you will notice the dc-offset during the pulse (when there is no pulse the dc-offset is very little). 

15thDec2020_taper.zip

Did you remove the transformers? We did not suggest anything regarding ferrite beads. Confused by your reply.

No, we haven't removed the transformers. The ferrite beads were for the cross tall between the two ADCs.

Please remove them as we think this is your problem.

Are you referring to the problem I mentioned in my post today or a previous post. Please have a look at my post today. Previous problems are resolved.

Hi Jim,

     Awaiting your response. Thanks for your effort,

SM

We still think this is somehow related to the transformers. Try sending the input using an amplifier. 

Could you please this on one if your eval boards. I doubt if the amplitude taper is because of the transformer, especially in the cw case.

SM,

Please try this with the transformers removed.

Regards,

Jim