TF1260 80Mhz Limit
Posted: 12 Aug 2021 09:33
Ok I have pretty much conclusively found why the 80Mhz limit is there on the TF1260 (at the moment). Its pretty simple really and most of you can measure it if you have one.
Basically the TF1260 existing firmware produces a 2x clock which is used to sample the 14mhz clock to recreate it.. I wont go into the details of why its got to be so fast but Nyquist etc.. The faster the sample the more faithful the clock reproduction will be.
Anyways. The CPLDs we have just crap out past 160Mhz and we stop getting a clock on the CPU line. There is a small resistor between the CPU and the CPLD on the top side of the board. You can measure CPU speed there. To test if this is happening simply probe that resistor after you set a high frequency and get a lock up. I usually 8 out of 10 times get no CPU clock.
The rest of the design works well past 80Mhz because ive had the TF4060 up at 100Mhz for extended periods. You dont need terminations or anything stupid like that because at 100Mhz everything is a terminator and being terminated fast. I actually do maths on stuff so i dont have to overengineer... but i usually measure for where there is a problem first.
Anyways the new TF1260 firmware will not have this limitation. It will not use a doubled clock and that in turn will give us more frequency steps.
We should now be able to get these steps
MULTIPLIER_4 // 100 Mhz
MULTIPLIER_3_75 // 93 mhz
MULTIPLIER_3_333 = //83 mhz
MULTIPLIER_3_2 // 80 mhz
MULTIPLIER_3_125 // 78Mhz
MULTIPLIER_3 // 75Mhz
MULTIPLIER_2_666 // 66Mhz
MULTIPLIER_2_5 // 62.6 Mhz
MULTIPLIER_2_4 // 60 Mhz
MULTIPLIER_2_333 // 58Mhz
MULTIPLIER_2 // 50mhz
Feel free to go measure the clocks at different speeds to see whats happening.
Basically the TF1260 existing firmware produces a 2x clock which is used to sample the 14mhz clock to recreate it.. I wont go into the details of why its got to be so fast but Nyquist etc.. The faster the sample the more faithful the clock reproduction will be.
Anyways. The CPLDs we have just crap out past 160Mhz and we stop getting a clock on the CPU line. There is a small resistor between the CPU and the CPLD on the top side of the board. You can measure CPU speed there. To test if this is happening simply probe that resistor after you set a high frequency and get a lock up. I usually 8 out of 10 times get no CPU clock.
The rest of the design works well past 80Mhz because ive had the TF4060 up at 100Mhz for extended periods. You dont need terminations or anything stupid like that because at 100Mhz everything is a terminator and being terminated fast. I actually do maths on stuff so i dont have to overengineer... but i usually measure for where there is a problem first.
Anyways the new TF1260 firmware will not have this limitation. It will not use a doubled clock and that in turn will give us more frequency steps.
We should now be able to get these steps
MULTIPLIER_4 // 100 Mhz
MULTIPLIER_3_75 // 93 mhz
MULTIPLIER_3_333 = //83 mhz
MULTIPLIER_3_2 // 80 mhz
MULTIPLIER_3_125 // 78Mhz
MULTIPLIER_3 // 75Mhz
MULTIPLIER_2_666 // 66Mhz
MULTIPLIER_2_5 // 62.6 Mhz
MULTIPLIER_2_4 // 60 Mhz
MULTIPLIER_2_333 // 58Mhz
MULTIPLIER_2 // 50mhz
Feel free to go measure the clocks at different speeds to see whats happening.