Overclocking to 128 MHz

When Justin used a 128 MHz clock for the ADC  I simply thought that it was not possible. I based this to the FX3014 specs that state the upper limit of GPIFII to 100MHz.  

He was right and this kind of overclocking is possible, maybe because we don't use the FX3 in an extended temperature range or maybe because the GIPIFII state machine we use is simple.

The transfer rate of the USB stream raises to 128 Msps with 16 bit samples i.e. 256 MBs. This requires a quite fast PC and it may raise some more problems in the transfer. 

Nevertheless it runs on my PC (I7-3770)  and offers another advantage other than wider bandwidth.

Howard Su noticed that the noise level when running in 128M mode decreases by some 3dB while the signals remain with the same amplitude. 

What's happening is that we have a 3dB decimation gain as explained in  https://www.analog.com/en/technical-articles/analog-tips-decimation-for-adcs.html :

" ...The SNR calculation of an ideal ADC must account for the processing gain of the filtered noise. Using a perfect digital filter, the processing gain due to the filtered noise will increase by 3 dB for every power-of-two reduction in bandwidth. "

This gain is related to the ratio between ADC bandwidth and demodulation bandwidth; it is not related to the sample rate IF bandwidth.

Bjarne Mjelde  reported the similar result during MW listening.

Please notice that the IF filter is wider in the 128M mode when the 2 MHz bandwidth is used.
The comparison is valid in the center of the picture from 300 kHz to 1700 kHz, outside of it there are fft/filtering artefacts.  Different filtering in the 2M band with 128M mode has to be tested.

Update 29/12/2020

On https://github.com/ik1xpv/ExtIO_sddc/releases/tag/v1.1.0 you can find the sources and the binary of the dll.
We tried to fix some of the bugs you reported of v.1.1 RC1. 
Thank you for the support and please continue to signal the bugs you find. We know they are there.
You can open an issue in https://github.com/ik1xpv/ExtIO_sddc/issues 
or post it in https://groups.io/g/NextGenSDRs or email to me ik1xpv[at]gmail.com,

Possibly we will have no new release coming before 2021 :-))
Best Season's Greetings ,
Oscar  





Comments

Post a Comment

Popular posts from this blog

A look into the RX-888 mkII

Image
  The RX-888 mkII designed by Justin Peng and Howard Su  arrived!   It is an evolution of the first RX888. The curiosity won out and I opened it. Be careful if you decide to open your device because two heat pads are used to carry the heat from the R828D and the LTC2208 to the aluminium case. I managed to slide them out without damaging them by moving the PCB very gently. A plastic wide stick may help to push out the pads and after to reassemble the radio. The PCB layout is similar to RX888 In HF the input attenuator is a  PE4312   covering a 31.5 dB attenuation range. This replaces the 0, -10 ,-20 dB attenuator of RX888.  It is followed by the low pass filter at 64 MHz.  Then two switches PE4259   select the HF input or the VHF tuner R828D produced by Rafael Micro that replaces the R820T2. The added transformer connects in a balanced way to the R828D IF outputs. An AD8370  digitally controlled variable gain amplifier (VGA) provides precision gain control, high IP3, and low noise figur
Read more

RX888

Image
Yesterday I received the RX888 designed by Justin Peng and Howard Su.  The first impression in o pening the shipping package  is that it is very small.    It is as big as the Cypress fx3 superspeed explorer kit ! Thanks to Guy Atkins'  SWLing post  I was able to resist the desire to open it right away to see the interior.   I started looking at HF. I connected it to my old I7-3770 pc and to the 10 meters wire antenna that I have on the balcony of my house.   My city is Turin and here the local transmitter in MW has frequency 999000 Hz synchronized to a high precision reference. I tuned this station and without calibration of the internal oscillator of R888 I measured an accuracy better than 1 ppm. This is not obvious after a transport of more than 10000 Km. In case there is the internal trimmer that allows you to calibrate the reference oscillator when you have a reliable reference. Frequency calibration test The ExtIO_RX888.dll v0.4 has the gain scale calibrated without subtractin
Read more

HF103 rc1

Image
HF103 is a HF only prototype of BBRF103  that started at the end of 2019. Thanks to Andrew G4XZL for ideas, test reports and advice.  Here after the modifications. Hardware: - It is designed for LW-MW-HF band only: 0-30MHz. - A Low Noise Amplifies has been added to improve NF of receiver. - The Clock reference used is a low noise xtal oscillator with lower phase noise and spurs than Si5351a.     - The RF attenuator of HF103 uses  DAT-31-SP+   - The PCB has been made larger 100x100 mm to increase spatial separation between components' groups (COVID style)  Here the  scheme and PCB layout . Some pictures of hardware Case  ADC and RF layout                                                                       LNA, filter, attenuator and xtal oscillator Software (still in evolution) : - The reference clock oscillator, that is the ADC sampling frequency has accuracy error. The error is compensated with a software trimming value into the ExtIO.dll's user interface dialog. - The ExtI
Read more