Home
The reason for HW modification is because we are using the devices for something other than the intentions they were designed for.
The CX card was designed to capture CVBS signals, so probably somewhere around 8mhz max frequency expectations, and the input filtering is designed around that. That's a fine frequency for VHS, but it is way too low for laserdisc, and possibly for some other tape formats (like SVHS).
The pre assembled amplifier boards like the opa657 and ad8367 i have been using, it isn't 100% clear what they are assembled for and used for, i think maybe SDR application? If so, then they are designed to work optimally at much higher frequencies than the RF we are capturing.
The information below, is just what i've done, found, discovered, thus far. I make no claim that it is perfection, and actually, welcome and encourage others to take what has been found here, and perform your own experiments. Do your own modifications and tweaks, to move the progress forward. I probably have done something wrong or sub-optimal. 😄
One issue here, is that some of the parts, the cost of shipping is high, but is the same for a large number of parts. It would make economical sense, if someone were to buy in quantity, assemble, and sell fully made kits. The crystals and low pass filters fall into this category. At around $1 for the crystal, and $14 for the filter, it's pretty inexpensive. The shipping (USA to USA) is like $8 for the crystals and $12 for the filter!
The reason this is all based around the PCIe CX cards, is because they are all relatively the same schematically, and they are readily available to purchase new from various places. The PCI cards, have too much variation to adequately instruct everyone on how to modify it.
At a certain point, things start to be subjective and relative to personal preference. I've seen this echoed in the community, some people prefer a sharp image, which tends to show the noise a little more (which results in a slightly lower ld-analyse SNR), and some people prefer a softer image with less noise (which results in a higher ld-analyse SNR). The placement of the components in the filter setup can change the result, i will document both ways, where i can. NOTE the soft/sharp filtering configuration is different per card + amp setup, and some cases, even OPPOSITE of each other.
For the daring: it is possible to mod CX card with a 54mhz crystal. Heatsink required, and fan blowing in the "general direction" suggested. At 54msps rate, it seems to be able to work with laserdisc RF with no low pass filter (this is good because the low pass filter is the most cumbersome item to obtain):
- with SXLP-13+
/getsnr FULL_4300D_rfTap_100InOut100itOPA657_13+lpf_GRN54c2_bi50t_dxvx2_FAN_50k150ciCXlvladj_gain+1_ggv1069.r8.tbc.json Total Fields: 50437 Averaged White SNR: 39.894 Averaged Black SNR: 40.622 - with no LPF
ggv1069DbS1_4300d_RFtap_mod100iotPA657_NOlpf_smOrngBNC_modBNCg54c2@8x54_Bit50_DXVX2_bgfanhs_cxlvlcav_pgwaydown.r8.tbc.json 50407 ,39.703 ,40.465
With no filter, essentially only the "sharper" configuration is available though.
You can increase the raw sampling rate of the cx raw capture, by upgrading the hardware timing crystal. 40mhz is the easiest (same form factor) and matches the *-decode tool chain. This will be part of the mod procedures documented later on. For informational purposes, other speeds might be useful for other applications (54msps seems to be enough to capture MUSE LD for instance). Here is a list of crystals tried and known to work: XTAL List
Note that capturing at speeds slower or faster than 40msps, engages the resampler function of the *-decode tools, and will take slightly longer, and also the seek functionality precision ( -S ) is reduced.
Using anything other than the RMS amp, you'll also have to use the CAV capture script included with the cxadc driver. This is to account for the signal strength change over the course of the disc. The procedure has a few steps. The RMS board reduces the complexity of CAV capture.
When it comes to the AD8367 vs OPA675 setups, when modding the CX card you remove the same onboard components . If using external termination, then technically modding the card is the same, however you'd have to make a custom terminator to use with AD8367 (of about ~170ohms). The opa657 works with 50ohm load. The ad8367 needs just above 200 ohm load. In addition to the ~170 ohm termination you use with the AD8367, you also need a ~50ohm resistor in series with the signal path, either before or after the low pass filter. If doing onboard card mod, removing old components, and replacing with new ones to meet the termination requirements, then it's one 50ohm and one 0ohm for the opa657 setup, and two 330ohm and one 0ohm for the AD8367 setup (you can actually use a single ~170ohm resistor, but that's a non standard value, hard to find).
Note that using the opa657 board requires a more complex power source, +5v and -5v, whereas the AD8367 works on a single power source with a wide tolerance.
Using opa657 and non RMS AD8367 would also require the use of the special CAV capture script. The RMS board build does not.
"That'll do pig, that'll do..." (or if you're younger: "That'll do donkey, that'll do...")
I'd note that the vectorscope readout fromm ggv1069 color bars comes closer to the right level with the sharper filter setup, and this is also the setup i personally prefer:
Here are a few comparison shots (same decode and ld-analyse settings):
All my shots in the above links are from the capture i uploaded to IA: https://archive.org/details/4300-d-rf-tap-100-in-out-100it-opa-657-13lpf-grn-54c-2-8x-54-bi-50t-dxvx-2-fan-5 using opa657 amplifier, sxlp-13+ low pass filter, and the gain changing capture script.
(for historical reference, this older document is provided: https://github.com/tandersn/cxadc-hw-mod/blob/main/The_Historical_Documents/CX_(cxadc)_card_notes.pdf)