<azonenberg>
you can't actually decode the PRBS because the sample rate is too low. and of course there's a ton of loss in the frontend, i cheated by cranking drive strength to the max and putting a whopping 12 dB of emphasis on the signal
Degi has quit [Ping timeout: 260 seconds]
Degi has joined #scopehal
<Degi>
Huh, even at 20x oversampling?
<azonenberg>
20x? lol
<Degi>
100 MHz at 2 GS/s
<azonenberg>
the PRBS is 1.25 Gbps :D
<Degi>
I mean the SDS2104X+
<Degi>
Ooh, is that a different scope?
<azonenberg>
No
<azonenberg>
This is 1.25 Gbps into a SDS2104X+
<azonenberg>
no tricks
<azonenberg>
just 12 dB of emphasis and 1200 mV swing
<Degi>
Yeah, uhh that sounds a bit high for 2 GS/s. I wonder if a Rigol MSO5000 could decode it with 350 MHz BW and some de-embedding
<azonenberg>
whats the sample rate on the mso5000
<Degi>
8 GS/s single channel
<Degi>
Hmm, whats the bandwidth set to on the SDS2104?
<azonenberg>
100 MHz
<azonenberg>
Lol
<azonenberg>
This data rate into an MSO5000 wouldn't even be a challenge
<Degi>
Ooh, I thought it was set higher since you can buy better options. Hmm yeah, I guess 100 MHz + slowish rolloff + 12 dB yields something
<Degi>
The rigol can see like 1.9 GHz or so in a FFT tho
<Degi>
(And IIRC about 1 GHz by eye on the largest amplification with about 1 Vpp input)
<Degi>
Can you better recover the signal if you apply a sin(x)/x filter or so with 700 MHz cutoff?
<azonenberg>
i actually just did sinc upsampling
<azonenberg>
and the PRBS seems to decode ok
<azonenberg>
this is hilarious
<Degi>
Lol
<azonenberg>
gimme a minute
<azonenberg>
i might have to try actual 8b10b next
<Degi>
I mean you still have some oversampling relative to the data and have 3.2x oversampling relative to 625 MHz, which should theoretically be probably maybe enough to reconstruct the signal in some parts
<Degi>
Is this device any different from an FPGA SERDES? I could hook up an ECP5 to a Rigol MSO5000 to see how well it performs sometime in the next days
<azonenberg>
an fpga serdes would work fine, you'd just have to turn the emphasis on the driver up probably
<azonenberg>
idk what the ecp5 serdes have for emphasis
<Degi>
Hmm, IIRC you can set pre, main and post drivers in units of some hundred µA to some mA I think
<azonenberg>
yeah if you tune those right then it should be doable
<azonenberg>
basically you want the signal horribly overemphasized to compensate for losses in the scope lol
<Degi>
Hmm, I have 6 slices which I can assign to pre data, main data, post data and disabled. 4 can be set to 0.8, 1.6, 2.4 and 3.2 mA while the other 2 can be set to 100 µA to 800 µA in 100 µA steps
<Degi>
And I can invert pre and post emphasis
<azonenberg>
yeah i'm not sure what the best tap coefficients to use would be
<azonenberg>
you might hav to experiment
<Degi>
Not sure if the current equals output current either
<Degi>
I wonder what the difference between Reserved and Spare register bits is
<Degi>
Ooh, there is a PRBS error counter and a PRBS enable bit too, but they're mostly undocumented