I have managed to replace both buck DC-DC converters with AMS1117 3.3 &
lm317 linear regulators and I successfully got rid of the mentioned 1 MHz
buck converter clock to obtain as pure a signal as possible. Do not expect
miracles as the DAC's are 8-bit and have a best case SNR of -48dB, but
getting the sideband modulation out of the signal surely improves the
resulting spectrum a lot.
I have performed the mod on 4 fl2k adapters.
Two of them came with a 1.2V AMS1117 for the 1.2V and a buck psu for the
3.3V. Another two came with 2 buck psu's
Using 2 LM317 is a bit complex but can be done, as my proof of concept
testbed has demostrated. One could actually fit them inside the plastic
box. Mind you, linear regulators can get quite hot. Still not hot enough to
inflict any damage after a 8-hour operation of the device.
Another notable effect was the total elimination of "cb status 1" message
on the testbed. The testbed suffered a lot of USB connection drops due to
poor power delivery to the chip, which is also the plague of those devices
that is causing bad reviews along with the poor assembly of the cable from
those that use the device in regular vga output mode. The workaround is
extremely simple; we use the onboard capacitors for filtering the output of
the linear regulators, I install them inverted and add metal pins from 1/4W
resistors to patch the power from the pin of the buck chip to the input of
the linear regulator. Same with ground.
Of course we are using USB3 interface that can provide more than 2.5W
(500mA) of output power. I haven't thorougly tested device operation and
stability under USB2.0 which has the aforementioned power limiting. The
device looks stable though.
There are implementations utilizing a pair of AMS1117 from factory.
[image: 20190125_194224_001.jpg]