[time-nuts] On low-voltage TAC/TDCs for a GPSDO

[Bruce Griffiths]

J.D. Bakker wrote:
> Thank you for your comments. Replies inline:
>
> At 11:37 +1200 13-08-2010, Bruce Griffiths wrote:
>> 1) Use a 74AHC05 for Q1 and Q2.
>
> I was looking at the 74LVC1G07. It's a single gate, so probably less 
> package parasitics to worry about. Output capacitance is actually 
> specified (5pF typ); charge injection might be a bit higher than the 
> AHC part. As its output is 5V tolerant, it might be worthwhile running 
> it at a lower voltage to reduce injection (but maybe that's premature 
> optimization at this stage).
>
>> 3) Replace the current source with a resistor.
>> The resultant nonlinearity is well defined and software correction 
>> should be relatively easy.
>
> (and, from a later message):
>
> At 17:59 +1200 13-08-2010, Bruce Griffiths wrote:
>> Yet another option is to sample the output of a simple 1us time 
>> constant RC low pass filter and fit an exponential to the sampled 
>> data and calculate the threshold crossing from this.
>
> I've considered this, and haven't entirely rejected it yet. The 
> hardware simplicity is appealing. On the other hand, a linear ramp 
> gives equal exposure to the entire code range of the (linear) ADC, 
> which is better for averaging out noise and ADC nonlinearities. Even 
> with the dual ramp configuration and the more complex current mirror 
> the increase in BOM cost is less than a dollar, all in parts that have 
> multiple sources and are easy to get.
>
>> 4) If the ADC(s) have a sufficiently wide full power bandwidth then 
>> one could just sample a pair of quadrature phased 250kHz sinewaves.
>
> As someone who's used to thinking in I/Q I must say I've always liked 
> the elegance of this approach. Trouble is that I don't see a 
> cheap/easy way to generate quadrature sines with low enough 
> distortion/noise.
Distortion isnt a great problem if its relatively small and stable as it 
can always be measured as part of the calibration process and its effect 
may then be compensated in software.

>
>> Measuring negative time intervals should not be necessary as the TAC 
>> (or other TDC) should be used merely to measure the delay of a 
>> synchroniser the output of which is used to synchronously sample a 
>> counter clocked with the same clock as the synchroniser.
>
> I want to have a GPS-synchronized PPS output. I believe that the 
> uncorrected PPS from some GPS modules (including the M12?) can be both 
> early and late, so I need some way to account for this. While I can of 
> course generate an early PPS for the TAC to provide an offset of a few 
> hundred ns, I'd prefer to tap the actual (synthesized) PPS output so 
> that the output driver's delay drift is inside the loop. Naturally I 
> have to take care that loading on this output doesn't end up pulling 
> the GPSDO in that scenario.
>
> (I'm not sure why I'd want to use a synchronizer in this path. The way 
> I see it the TAC operates as a linear phase detector, with the GPS PPS 
> and the synthesized PPS as inputs. The microcontroller then applies 
> the sawtooth correction to the measured time offset, and uses the 
> result in a DPLL. There will, of course, be a synchronizer in the 
> input line from the GPS PPS to the microcontroller, but that's only 
> used for the FLL and for rough synchronization).
>
Using a synchroniser allows the TAC output range to be combined with the 
coarse timestamp derived by sampling a counter clocked by the same clock 
as the synchroniser.
This allows the TAC range to be reduced to 2 clock periods (with a 2 
stage synchroniser) and doesn't require measuring a negative time 
interval with the TAC.
However this requires one TAC per measured channel.
If, however the regenerated PPS is synchronous with the synchroniser 
clock a TAC isnt necessary for time stamping this signal.
If you want to include the driver delay then a TAC is required however 
you then need to calibrate the offset between the 2 TACs.
This isn't trivial and using reed relays (or equivalent ) to ensure well 
matched propagation delays for the various cal signals driving each TAC 
may be required.

>>> - Circuit 3 expands on this approach by having dual ramp generators, 
>>> and having the ADC measure the voltage difference between the two.
>>
>> Not a good idea, as this requires accurate matching of the gains of 
>> the 2 TACs.
>
> Why?
>
> At that points they're not TACs yet, just ramp generators. Circuit 3 
> uses the difference between these ramps, and I believe it need not be 
> constant.
>
> Assume there's a 1% difference in ramp rates; say C3 charges at 1V/us 
> and C4 charges at 1.01V/us. When the ramps are started simultaneously, 
> the difference between ramp voltages (ignoring saturation) is simply 
> 0.01t (in us). When one ramp is started 1ns earlier than the other, 
> the difference (during the time that both ramps are active) is 0.001 + 
> 0.01t. I believe that this linear factor is easy to calibrate out with 
> regular early/on-time/late calibration.
>
Since NP0/C0G caps are only available in 5% and 10% tolerance at best, 
matching gains to 1% will require using selected parts (adjust current 
source to compensate) or trimming.
Simulation indicates nonlinearity of the order of 1% or so in the ramp 
generator. This is largely due to the Early effect and semiconductor 
output capacitance modulation.

> I am somewhat concerned that in this scenario it may be harder for 
> software to detect the point that both ramps are running from sampled 
> values alone, and nonlinearities in the current source are harder to 
> calibrate out. Still I believe the basic scheme should work fine. Am I 
> missing something?
>
The effect (non linear) of sampling with direct ADC connection if you 
continuously sample the ramp.

>> You will also need to ensure that the current source recovers 
>> sufficiently quickly from saturation.
>
> In the most intensive calibration regime I'm looking at now 
> (early/on-time/late every second) I'd have four measurements per 
> second, so the current source has the better part of 250ms to recover. 
> That won't be a problem for the four-transistor mirror; I'd have to do 
> the math with actual part values to be sure with the simple version.
>
Thermal effects may have longer time constants than that.
If you use an opamp (with rail to rail outputs and inputs) to compensate 
for current source Vbe tempco then the settling time when the required 
decoupling is included (required unless you use a very fast opamp) will 
be quite long. Diode clamping the ramp output to avoid saturation is 
advisable.

The output compliance of your four transistor current mirror is limited 
to around 1.3V or so before the onset of saturation or gross nonlinearity.
Unless the transistors are closely thermally coupled some resistance in 
the emitters of the CE transistors may be required to avoid thermal runaway.

>> Another issue is to limit the discharge current flowing in the 
>> discharge switch.
>
> I know, I left that out for clarity. A small drop across the discharge 
> switch/limiting resistor needn't be a problem, given that many 
> single-supply converters and buffers are happier if their input is 
> ~100mV above ground.
>

It may be necessary to use a somewhat larger series resistor value than 
100 ohms to limit the initial discharge current.

> Thanks again,
>
> JDB.
Bruce