[time-nuts] OT: Levelled sine wave generator

[Magnus Danielson]

Bruce Griffiths skrev:
> David C. Partridge wrote:
>> Sort of related, but only just - however the signal to noise ratio here is
>> so good that I feel impelled to ask.
>>
>> For 'scope calibration I'm considering building a levelled sine wave
>> generator.
>>
>> Ideally the specs I'm looking for are:
>>
>>  o Close to DC (10kHz or 100kHz would be fine) up to at least 1GHz.
>>  	more would be better but not critical
>>
>>  o Output levels from 0.5Vp-p(-2dBm) to at least 4Vp-p(+16dBm) into 50R
>> 	 (up to >6Vp-p(say +20dBm) would be better)
>>
>>  o Output flatness levelled within 2% of desired output level (+/- 0.086dB)
>> 	across the entire frequency range at the final connector to the DUT
>> 	This will almost certainly mean an external levelling head.
>>
>>  o Modulation - not critical, FM or AM might be useful.
>>
>>  o A logarithmic sweep capability might be nice, but isn't necessary.
>>
>>  o Frequency display - nice to have but output to external counter is OK.
>>
>> Generating the basic signal is probably just a case of using something like
>> an HP VTO-8200, mixing it with 2GHz (Mini-Circuits RMS30?), low pass filter,
>> an AGC stage (see below) and then amplify probably using an MMIC like the
>> Mini-Circuits ERA-2SM followed by an additional stage to get the extra few
>> dB.   For more accurate frequency control some sort of synthesiser locked to
>> a reference might be in order (I had to get a time-nuts hook in here
>> somehow).
>>
>> The question is what should go in the sensor head?
>>
>> Logically I need to sample a proportion of the signal delivered to the
>> output connector, compare the output of the sensor against a DC reference
>> level telling it the desired output level, and feed back a voltage to a
>> wideband AGC stage (any suggestions for this?) in the main unit.   I also
>> need to be able to detect that output is not levelled.
>>
>> Or should I just forget the whole idea and go talk to R&S with a large
>> cheque in hand?
>>
>> Cheers
>> Dave
>>   
> The phase noise wont be particularly low especially for low output
> frequencies.

I think this is acceptable for the intended application.

> A diode double balanced mixer with dc current applied to the IF port is
> useful as a wideband current controlled AGC device.

Indeed. BTW, I used that trick when using my network analyzer for 
sweeping the Caesium-beam. I unplugged the 12,6 MHz and inserted my 
output port which was split to also go into a mixer. I picked up the 
detector voltage and feed it into the IF port and the resulting 
modulated sine was sent to the receiver port. Worked like a charm and 
provided me with a nice user-interface.

> NB mixer IF response must extend to dc.

It should not be too hard to find a suitable mixer that matches that 
description thought.

As for sensor-head, an old technique is to use a pair of diodes 
thermically connected, where one is fed a 100 kHz square wave and the 
other is used as a detector. The reference diode is bias adjusted and 
the reading from the other is compensated with the same amount and thus 
allowing for a linearization. I don't know if there is any modern ways 
which is more suitable. I think it could be a bit of a challenge for 
detecting it all the way down. I'm sure Bruce can elaborate some on that.

One possible issue would be that input impedance could mess things up. 
So maybe one should consider making quadrature readings in the head such 
that with some processing the propper level can be given considering the 
impedance mismatch. For the intended repeatability this might be 
something to consider.

Cheers,
Magnus