[time-nuts] Wavecrest DTS-2077 Teardown

Ed Palmer ed_palmer at sasktel.net
Thu Aug 22 02:52:51 EDT 2013


Sounds like I need to do some experiments.

Thanks for the advice and idea, Bruce. :)

Ed

On 8/22/2013 12:15 AM, Bruce Griffiths wrote:
> That will work to some extent however you need to tailor the stage 
> gain and bandwidth distribution to suit for optimum performance.
> Its somewhat difficult to control the gain and bandwidth of an off the 
> shelf CMOS inverter and you also need to know its noise parameters.
> Maybe adding resistors in series with the power supply leads of the 
> CMOS inveters and adding some output capacitance will suffice to 
> adjust the gain and bandwidth.
>
> Bruce
>
> Ed Palmer wrote:
>> Since you're looking for rise times in the low or sub nanosecond 
>> range, why wouldn't you include any logic gates where such rise times 
>> are inherent?  I was thinking of maybe a chain of faster and faster 
>> logic gates.  For example, Potato Semiconductor - no, I'm not making 
>> that up - PO74G04A has a risetime of < 1 ns and, if you can keep the 
>> load capacitance low enough, a maximum input frequency of > 1 GHz.
>>
>> Always trying to learn....
>>
>> Ed
>>
>> On 8/20/2013 11:28 PM, Bruce Griffiths wrote:
>>> The same analysis applies however one would probably use something 
>>> like cascaded longtailed pairs with well defined gain (series 
>>> emitter feedback) and the low pass filter cap connected between the 
>>> collectors rather than opamps.
>>>
>>> Bruce
>>>
>>> Ed Palmer wrote:
>>>> Does anyone know if this situation would benefit from doing 
>>>> something similar to a 'Collins Hard Limiter' i.e. instead of 
>>>> squaring the signal in one stage, use maybe two or three cascaded 
>>>> stages with increasing bandwidths? Normally, Collins limiters are 
>>>> used with beat frequencies of less than 1 KHz, but maybe there's 
>>>> value in doing at typical time-nuts frequencies.
>>>>
>>>> Any thoughts?
>>>>
>>>> Ed
>>>>
>>>>
>>>> On 8/20/2013 10:02 PM, Said Jackson wrote:
>>>>> Hi Ed,
>>>>>
>>>>> For anything up to about 150MHz try the NC74SZ04 types from 
>>>>> National if you can find them NOS. they stopped making these years 
>>>>> ago.. Fairchild is ok too but not as fast from what I have seen.
>>>>>
>>>>> Forgot I wrote about it in 2009. Oh boy -age kicking in.
>>>>>
>>>>> Bye,
>>>>> Said
>>>>>
>>>>> Sent From iPhone
>>>>>
>>>>> On Aug 20, 2013, at 20:17, Ed Palmer <ed_palmer at sasktel.net> wrote:
>>>>>
>>>>>> Hi Said,
>>>>>>
>>>>>> Yes, I saw your message from 2009 where you warned about the sine 
>>>>>> waves.  That's why I was watching for it. Thanks for the 
>>>>>> warning.  I also realized that a DC Block and a 10 db attenuator 
>>>>>> makes a very nice TTL or CMOS to Wavecrest converter for anything 
>>>>>> except 1 PPS which would need about 15 db.  I tried an old 
>>>>>> circuit that uses an MC10116 ecl line receiver - it's actually a 
>>>>>> dead Racal Dana 1992 counter where I'm using the processing on 
>>>>>> the external reference input to square up the signal.  It gives 
>>>>>> me a slew rate equivalent to about a 50 MHz sine wave.  It helped 
>>>>>> a lot, but not enough. I'll try a 74AC04 and a BRS2G Differential 
>>>>>> Line Receiver (risetime < 3ns, 400Mbps throughput).  Both are in 
>>>>>> my junkbox.
>>>>>>
>>>>>> Ed
>>>>>>
>>>>>>
>>>>>> On 8/20/2013 8:12 PM, Said Jackson wrote:
>>>>>>> Guys,
>>>>>>>
>>>>>>> The dts needs to be driven by square waves, driving them with 
>>>>>>> sine waves gives jitter values that are displayed significantly 
>>>>>>> too high due to trigger noise.
>>>>>>>
>>>>>>> Holzworth makes a small sine wave to square wave converter that 
>>>>>>> can drive 50 ohms. Use a DC block and an attenuator on the cmos 
>>>>>>> output to avoid damaging the dts inputs. You can make your own 
>>>>>>> converter using a single fast cmos gate, resistor, and blocking 
>>>>>>> cap. By using hand-selected gates I was able to achieve less 
>>>>>>> jitter with that circuit than what the Holzworth box was able to 
>>>>>>> achieve.
>>>>>>>
>>>>>>> Doing that conversion can bring down the measured rms jitter on 
>>>>>>> a very good 10MHz sine wave source from 10ps+ to less than 2ps - 
>>>>>>> basically at or below the noise floor of the dts.. Once you run 
>>>>>>> at the units' noise floor, you know your source is quite good..
>>>>>>>
>>>>>>> Bye,
>>>>>>> Said
>>>>>>>
>>>>>>> Sent From iPhone
>>>>>>>
>>>>>>> On Aug 20, 2013, at 18:51, Ed Palmer <ed_palmer at sasktel.net> wrote:
>>>>>>>
>>>>>>>> Adrian,
>>>>>>>>
>>>>>>>> I used Timelab to assess the reaction of the DTS-2077 to 
>>>>>>>> different sine wave inputs.  The differences in the noise floor 
>>>>>>>> are surprising.  The attached picture was made by taking the 
>>>>>>>> output of an HP 8647A Synthesized Generator through a splitter, 
>>>>>>>> and then through different lengths of cables to the inputs of 
>>>>>>>> the DTS-2077.  The combination of splitter and cable loss meant 
>>>>>>>> I couldn't get +7 dbm @ 1 GHz. If I could have, the 1 GHz line 
>>>>>>>> might have been lower than it was.
>>>>>>>>
>>>>>>>> Ed
>>



More information about the time-nuts mailing list