Shop OBEX P1 Docs P2 Docs Learn Events
No longer 3.3Vpp when synthesizer over 100MHZ? — Parallax Forums

No longer 3.3Vpp when synthesizer over 100MHZ?

TransistorToasterTransistorToaster Posts: 149
edited 2007-10-17 22:29 in Propeller 1
Hello,
I measured on my oscilloscope the frequency synthesizer performance for frequencies above 100MHz and saw that the signal gives a reduced amplitude in the sense it doesn't go to neither Vcc or GND rails anymore. I tested on a few pins and got the same results. Considering that I am at the limits of my 100MHz scope with probes, I was wondering what other people may have measured. Did anybody get 3.3Vpp at the 128MHz limit? Are there any particular pins or cogs that would give better performance? Any quick fixes to get 3.3Vpp?
Thanks,
Frank

Comments

  • Mike GreenMike Green Posts: 23,101
    edited 2007-10-17 03:33
    You're not going to get 3.3Vpp at any speed. The I/O pins are simply not capable of that. There's a drop of maybe 0.4V through each of the MOSFETs in the output circuit ... to Vdd and to ground. Have a look at the datasheet for Vol and Voh specs. Maximum peak to peak voltage is 3.3-0.8 = 2.5V. If you have some kind of resonant circuit, it can increase the peak voltage (even to the point of damaging the I/O pin).
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2007-10-17 04:07
    It is true that the spec sheet says 0.4V and 2.85V but that is for a load of 10ma! Normally you will get to within 50mv of the supply rails for lightly loaded signals. Think of the mosfets as programmable resistors without any inherent diode barrier as in bipolar junction transistors. What you get on the output is akin to a voltage divider with the mosfets on-resistance equivalent to around 40ohms and you load resistance.

    However, and this is a big however, do not expect the scope to measure a 100Mhz digital signal faithfully, a 100Mhz sine wave maybe just. If you buffered the signal from the prop with a very high-speed buffer it may look a little better but never perfect. You can test this theory by selecting a lower bandwidth if your scope supports it or just by using a cheap scope.

    Now there is also the problem with driving signals over any distance with all that line inductance and capacitance and source impedance. What are you going to get? Exactly what the maths will tell you, and it isn't a nice square 3.3Vpp signal. Internal to the chip the signals will be fast and sharp but as soon as you connect some track to that pin it is going to affect it, as will any probe.

    *Peter*
  • TransistorToasterTransistorToaster Posts: 149
    edited 2007-10-17 15:01
    >There's a drop of maybe 0.4V through each of the MOSFETs in the output circuit.
    I never noticed a fixed drop like that anywhere.

    >You're not going to get 3.3Vpp at any speed.
    I get close to the rails easily for speeds below 60Mhz. I observed that the distance to the rails are not the same for the positive and negative supplies.



    >It is true that the spec sheet says 0.4V and 2.85V but that is for a load of 10ma
    That would tend to imply a load resistance of 330ohms (3.3V/0.010A). The capacitance on the BNC plus the adjustable one on the probe short the resistors to in that range for high frequencies. At 128MHz, I get 0.50V and 2.14V for the PDIP40 prop loaded with a 10X probe using cog 0 for the generation.

    >Now there is also the problem with driving signals over any distance with all that line inductance and capacitance and source impedance. What are you going to get? Exactly what the maths will tell you, and it isn't a nice square 3.3Vpp signal. Internal to the chip the signals will be fast and sharp but as soon as you connect some track to that pin it is going to affect it, as will any probe.

    Absolutely. The PDIP package surely doesn't help.

    Maybe cog 7 is better for high frequency output, considering that it is closest to the output in the daisy chain of the cog output pin driver circuits. I used cog 0 in my measurements.

    Edited for transcription mistake: (3.3V/0.010A) was written as (3.3V/0.10A).

    Post Edited (TransistorToaster) : 10/17/2007 5:46:48 PM GMT
  • deSilvadeSilva Posts: 2,967
    edited 2007-10-17 15:17
    TransistorToaster said...
    That would tend to imply a load resistance of 330ohms (3.3V/0.10A). The capacitance on the BNC plus the adjustable one on the probe short the resistors to in that range for high frequencies. At 128MHz, I get 0.50V and 2.14V for the PDIP40 prop loaded with a 10X probe

    R ~ 1/(f*C) with 30pF for the probe = 1/4mS = 250 Ohms - right!
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2007-10-17 22:29
    TransistorToaster said...

    >It is true that the spec sheet says 0.4V and 2.85V but that is for a load of 10ma
    That would tend to imply a load resistance of 330ohms (3.3V/0.010A). The capacitance on the BNC plus the adjustable one on the probe short the resistors to in that range for high frequencies. At 128MHz, I get 0.50V and 2.14V for the PDIP40 prop loaded with a 10X probe using cog 0 for the generation.

    A slight correction is that the spec sheet states 2.85V(min) with a 10ma load. R=V/I = 2.85/10ma = 285ohms not 330 which would be the total series resistance including the on -resistance of the mosfet.

    Note the use of min and max, it's all relative as a Voh of 2.85V min (@10ma) means it will not be lower than this whereas a Vol of 0.4V max (@10ma) means it won't be any higher than this. So at 10ma load we will get a guaranteed swing of 2.45V min although typically it will be closer to 3V (but I have yet to measure it). In fact of course this is over the full temperature range of -40'C to +125'C with the worst case chips in the batch, you might never see it get to these extremes.

    *Peter*
Sign In or Register to comment.