[time-nuts] Advice on 10 MHz isolation/distribution (Clay)

Bob Camp lists at cq.nu
Sat Feb 13 04:31:39 UTC 2010


Hi

There's also the "throw everything at it" approach. 

Use something like common base stages for the input and op amps for the outputs. Boost the level into the op amps and pad it at the outputs. You might get what you need. More parts than a pure op amp design, more current. Likely easier to get running.

Lots easier to do with a couple transformers in there.

Bob

 
On Feb 12, 2010, at 11:02 PM, Bruce Griffiths wrote:

> The only data available seems to be John Ackermann's measurements on the TADD-1 distribution amp.
> Unfortunately the opamp used is now obsolete or about to be.
> Most recent discrete designs (not the HP5087 amplifiers) that I have seen phase noise data for, have significantly lower flicker phase noise and phase noise floors, particularly at 10MHz.
> 
> Bruce
> 
> Bob Camp wrote:
>> Hi
>> 
>> I have no data, but I believe that in the real application, the phase noise would not be degraded by a good low noise RF op amp / buffer amp. About all you can do for flicker noise data is to look at what they do supply and make an guess based on how the noise rolls up over the range they do show.
>> 
>> An op amp circuit would certainly would take fewer parts, and likely more current. No free lunch ....
>> 
>> Bob
>> 
>> 
>> On Feb 12, 2010, at 10:11 PM, Bruce Griffiths wrote:
>> 
>>   
>>> In the later version the input amplifier has a gain of 2x and the output amplifiers have unity gain.
>>> 
>>> Whilst the reverse isolation (and output impedance) can be improved by using a complementary symmetry emitter follower output stage, one has to ask at that point is the performance gain worth it?
>>> 
>>> One has then in effect built a high open loop gain discrete current feedback opamp that has a somewhat lower input noise than a wide band IC opamp but it0 uses more components.
>>> 
>>> The problem with wide bandwidth opamps is there is very little data available on their RF flicker noise.
>>> 
>>> The measurement data I have seen for an isolation amp using a 2N5179 and a 2N3904 in a Sziklair pair configuration as the input stage indicates that it doesn't seem to noticeably degrade the phase noise of a 10811A. However no residual phase noise measurements have been made.
>>> 
>>> Bruce
>>> 
>>> Bob Camp wrote:
>>>     
>>>> Hi
>>>> 
>>>> Since it's the input stage, it's likely the point most impacted by a higher flicker noise part. That might make one want to look at alternatives.
>>>> 
>>>> Of course, it's not real clear that a super low noise amp is needed in this case.
>>>> 
>>>> Bob
>>>> 
>>>> On Feb 12, 2010, at 8:46 PM, Bruce Griffiths wrote:
>>>> 
>>>> 
>>>>       
>>>>> The series RC to ground keeps the high frequency impedance seen by Q1 and Q7 low so that the base current noise which increases significantly as the frequency approaches the ft of these transistors.
>>>>> However such a series RC network does little to suppress the the rise due to gain peaking.
>>>>> A shunt capacitor from the output stage collectors to the output stage bases is much more effective for the 2x gain stage.
>>>>> 
>>>>> Such a capacitor increases the noise for the 1x gain White emitter follower.
>>>>> Using an input transistor with higher bandwidth is more effective in this case.
>>>>> 
>>>>> Bruce
>>>>> 
>>>>> Bob Camp wrote:
>>>>> 
>>>>>         
>>>>>> Hi
>>>>>> 
>>>>>> I suspect your noise spike can be cured by a series R-C to ground from the junction of Q1 base, Q7 base and all the other stuff. Something is going to have to set a high frequency roll off. With no coils some combo of R and C is going to have to do it.
>>>>>> 
>>>>>> You might also try returning all of the upper emitter resistor bypasses to ground rather than B+. Another alternative would be emitter to emitter bypass as shown on the JPL schematic. I'm guessing both would improve isolation in a real world circuit.
>>>>>> 
>>>>>> Bob
>>>>>> 
>>>>>> 
>>>>>> On Feb 11, 2010, at 8:34 PM, Bruce Griffiths wrote:
>>>>>> 
>>>>>> 
>>>>>> 
>>>>>>           
>>>>>>> life speed wrote:
>>>>>>> 
>>>>>>> 
>>>>>>>             
>>>>>>>> Message: 2
>>>>>>>> Date: Fri, 12 Feb 2010 12:12:29 +1300
>>>>>>>> From: Bruce Griffiths<bruce.griffiths at xtra.co.nz>
>>>>>>>> The output (collectors of Q5, Q6 emitter of Q4) of the input amplifier
>>>>>>>> sets the dc voltage at the inputs ( Q1 base, Q7 base respectively) of
>>>>>>>> the output amplifiers.
>>>>>>>> 
>>>>>>>> The circuit consists of a unity gain input amplifier (Q4, Q5, Q6) that
>>>>>>>> drives a pair of output amplifiers (Q1, Q2, Q3 and Q7, Q8, Q9
>>>>>>>> respectively) each with a gain of 2x (6dB).
>>>>>>>> The input amplifier is essentially a white emitter follower with a
>>>>>>>> complementary symmetry output stage (shown in transistor electronics
>>>>>>>> books from the 1960's) where an input CE transistor drives a
>>>>>>>> complementary pair of CE transistors with feedback from the common
>>>>>>>> collectors of the 2 output transistors to the input transistor emitter.
>>>>>>>> In effect its merely a very simple unity gain opamp. Its usually best to
>>>>>>>> ensure that the CE output stage pair provide the dominant open loop
>>>>>>>> pole. Using a higher ft (2 to 3x)  input transistor than the output pair
>>>>>>>> is the usual way of ensuring this.
>>>>>>>> 
>>>>>>>> Well, it is so obvious now that you explained it.  I had forgot about the need for one of the stages to set the dominant pole.
>>>>>>>> 
>>>>>>>> Thanks Bruce and Bob for sharing your obsession with frequency controls.  I'll simulate this further, and have a prototype PCB built within the next few weeks.  I did notice the resistor at the base of Q2,5,8 is responsible for significant noise.  I'll have to be careful with the bias circuit.
>>>>>>>> 
>>>>>>>> Have to get busy for now, but I will report back with results.
>>>>>>>> 
>>>>>>>> Best regards,
>>>>>>>> 
>>>>>>>> Clay
>>>>>>>> 
>>>>>>>> 
>>>>>>>> 
>>>>>>>> 
>>>>>>>>               
>>>>>>> Clay
>>>>>>> 
>>>>>>> One can always use a smaller resistor in series with an RF choke that has no resonances in the region of interest.
>>>>>>> 
>>>>>>> The attached circuit schematic illustrates one method of biasing for which the emitter current of the input transistor can be largely sourced via a resistor rather than from the collector current of the npn output transistor.
>>>>>>> 
>>>>>>> My simulations indicate if that one uses 2N3904's as the input device rather than the 2N5179's shown that there is an enormous peak in the output noise spectrum at around 150-200MHz or so.
>>>>>>> When the 2N5179 is used this noise peak is much smaller and broader.
>>>>>>> 
>>>>>>> Use the same bias divider bypassing techniques that NIST used including the use of electrolytic caps (they used tantalum caps) to reduce the low frequency noise from the power supply. The ceramic bypass caps ensure sufficient isolation between stages.
>>>>>>> Simulating the reverse isolation with realistic component parasitics is always informative/useful.
>>>>>>> 
>>>>>>> Bruce
>>>>>>> <Transformerless_10MHz_disA.gif>_______________________________________________
>>>>>>> time-nuts mailing list -- time-nuts at febo.com
>>>>>>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>>>>>>> and follow the instructions there.
>>>>>>> 
>>>>>>> 
>>>>>>>             
>>>>>> _______________________________________________
>>>>>> time-nuts mailing list -- time-nuts at febo.com
>>>>>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>>>>>> and follow the instructions there.
>>>>>> 
>>>>>> 
>>>>>> 
>>>>>>           
>>>>> 
>>>>> _______________________________________________
>>>>> time-nuts mailing list -- time-nuts at febo.com
>>>>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>>>>> and follow the instructions there.
>>>>> 
>>>>> 
>>>>>         
>>>> _______________________________________________
>>>> time-nuts mailing list -- time-nuts at febo.com
>>>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>>>> and follow the instructions there.
>>>> 
>>>> 
>>>>       
>>> 
>>> 
>>> _______________________________________________
>>> time-nuts mailing list -- time-nuts at febo.com
>>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>>> and follow the instructions there.
>>> 
>>>     
>> 
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at febo.com
>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> and follow the instructions there.
>> 
>>   
> 
> 
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
> 




More information about the time-nuts mailing list