"Cheating" FCC compliance testing by using different USB cables?

[MicroBiggerThanMM]

I'm designing a product which will Tx/Rx at 2.4GHz, and I've thrown everything at it to guarantee I pass EMC first time (of course after pre-testing with an SDR). Max clock is 32MHz (ouch), but max signal speed is 7MHz, both traces are 5mm long and under solid RF shields on both sides. All 4 layers of the 0.8mm thick, <25mm diagonal board have ground pours via-stitched together across the perimeter of all 4 layers and every IC has at least 2, 0402 capacitors of different values to decouple the s**t out of the 10mW of total operating power of the device.
Obviously I will fail the EMI testing the first time, losing thousands despite everything I try to do.
I got cornered by design limitations and was forced to ground the micro-USB shield because it wouldn't fit otherwise. Obviously as soon as they plug in the USB cable to charge it, the thing will start screaming my mistake across the galaxy as the signals creep in (at least I decoupled all the data and power lines on the USB?) According to reply #1 on https://www.eevblog.com/forum/projects/usb-shield-ground-connection/ it's actually okay to ground the shield, but the topic of shielding seems to be a matter of opinion. Big device, isolate it, little fella, leave it grounded. It's not too late to change it, but can I just use a special isolated/shielded USB cables during the EMC testing to hide the non-conformity there, or will they use metrology grade micro USB-B connectors? Do they even test for conformity during charging? The charging circuit itself is just a linear circuit without switching, but without a rated PSRR like an LDO would have, I'm afraid that will just be a conduit for stray signals with the grounded shield making things worse.
Gosh I hope they just toss it on the rotary platform and measure the thing instead of actually testing it for its real world compliance. I could just disable operation while charging in software... or just toss in some LDO's to reject noise...
What else have you all done to narrowly get by the EMC tests?
Maybe my suffering with hand-soldering 0402's will lead to effective decoupling 

[AndyC_772]

Your product will be tested using whatever cables you supply with it. If you don't supply cables, it'll be tested with whatever cables you state must be used with it.

If you say "use a shielded cable", then you can test with whatever shieded cable you like - but that won't help if the shield is itself noisy. A really good foil & braid multi-layer shield is still an antenna, after all.

What noise are you expecting the unit to generate? Are you concerned about emissions related to your 2.4G radio, or about harmonics of your 7 MHz or 32 MHz clock & data?

[MicroBiggerThanMM]

Primarily, the 7 and 32 MHz signals since the radio has good integrated components which eliminate those issues. Can I just whack a cocktail of different decoupling cap values and sleep easy knowing all EM noise will be shunted? I'm gonna invest in some good cables.

[AndyC_772]

the radio has good integrated components which eliminate those issues

That's not an assumption I'd personally choose to make. A module with an integrated radio should - and I stress should - come with instructions for how it should be integrated into your product, ie. the requirements for spacing, grounding, shielding and so on. Check what the conditions are for the manufacturer's FCC and/or CE approval, and follow them to the letter.

Can I just whack a cocktail of different decoupling cap values and sleep easy knowing all EM noise will be shunted

No, sadly there's really no substitute for testing and experience - possibly not in that order.

IMHO the benefit of using lots of different capacitor values is overstated, and the topic is frequently misunderstood. I wrote an article on the subject here wihch you may find relevant.

[Neilm]

the radio has good integrated components which eliminate those issues

That's not an assumption I'd personally choose to make. A module with an integrated radio should - and I stress should - come with instructions for how it should be integrated into your product, ie. the requirements for spacing, grounding, shielding and so on. Check what the conditions are for the manufacturer's FCC and/or CE approval, and follow them to the letter.

Seconded
I have seen an integrated wireless module that had instructions to mount on the edge of the PCB with a defined area where there should be no conductors. This was not spotted on the board layout and it ended up that the range was barely outside the box and if there were leads connnected then it failed EMC compliance by a lot. We ended up shipping it with the module only enabled in certain modes where the leads should not be connected. (A "new model" with extra functionality was launched 8 months later which kept marketing happy)

[MicroBiggerThanMM]

Well, luckily I won't need to rely on a module's PCB antenna for production, I'm routing straight from the transceiver IC and setting up a matching network into a chip antenna, then replacing components until my SDR says the power is good-ish. There will be parasitic effects from the enclosure so it's worth adjusting with the matching network. On the antenna, I can whack on a bunch of integrated passive filters and do a lowpass to kill the 2.4GHz harmonics, up to the FCC regulated 24GHz.

What I'm terrified of is the main system clock creeping out through the ground plane. I would love to connect chassis ground and signal ground since the form factor is so small, but then I'd give noisy signals an antenna as they creep out through the ground plane. Can enough decoupling practically kill the dozen MHz clock noise? I'm benefitted by the fact that the small size of my board eliminates long elements to emit out of and instead of providing a quarter wavelength trace to leak out of, the best that evil clock will get is a hundredth of a wavelength element.

Is it possible to safely short chassis grounds and signal grounds without being noisy?
Can MHz signal harmonics be effectively killed, perhaps with a low impedance LC bandpass filter, or maybe an active circuit? Is that what people practically do to attain EMI compliance?

[JoeyG]

Can you do some cheap  RF pre screening with a spectrum analyser to see what is radiating.

Don't forget the harmonics of the fundermenta frequencies ,  they can also radiate.