[time-nuts] Soundcard sampling Re: Picking a good HP 10811

[Bruce Griffiths]

Stanley Reynolds wrote:
> <snip>
> Thus
> for a 1kHz input and 100dB SNR the input clock jitter should be < 18ns rms.
>
> for a 10kHz input and 100dB SNR the input clock jitter should be < 1.8ns
> rms.
>
> It is very likely that the differential sampling clock jitter
> contribution of a high end sound card to the ADEV noise floor will in
> around 1E-15 or less at Tau = 100sec.
>
> Bruce
>
> So offset each DUT to prevent injection lock within the 20 Hz range, and get relaxed spec on mixers and buffers. The retail was about 12 US$ for the surface mount and 50 US$ for the BNC mini-circuit mixers. Home made buffer amps and mixers sounds possible for me.
>
>   
2N3904's are more than adequate for very high reverse isolation buffer
amplifiers at 10MHz.
You should use a design similar to the NIST 5 and 10MHz distribution
amplifiers not the 100MHz design.
Opamps have considerably higher phase noise.
MMICs have too much gain to be useful.
> Use a 10Mhz +-10Khz offset common oscillator to put beat in the sound card range, just because jitter of 1.8ns sounds better than 18ns.
>   
No, 1 kHz or so is probably better, the above calculation was used to
infer a likely maximum for the sound card sampling clock jitter from the
specifications.
It is probably much better as other effects contribute to the SNR.
> Slave two cards to one oscillator for 4 channels run test, post data / wave files.
>
>   
Making the wave files available would indeed be useful in that they
would be useful for algorithm development.
Matlab, for example, can read wave files.
The same data can be analysed using various algorithms ranging from
calculating the zero crossing time stamps to a implementing a dual
Channel Costas receiver.

However use an application that actually saves 24 bit data.
A lot of applications only save 16 bit data even if you have a 24 bit
sound card.
You often need to use a hex editor to examine the files to find this out.
Start with a simple test using 2 mixers and one sound card.
Drive the RF port of each mixer from the same OCXO.
The resultant data can be used to establish the system noise level
provided the offset oscillator and the OCXO don't injection lock.
Then do a test using to separate OCXOs.
Finally when all appears to be working well test the 2 sound card 4
mixer system.
You will need to use high end sound cards like the M-Audio AP192 for
good performance.
This particular one has balanced 4VrmsFS inputs which will be useful in
reducing the effect of low frequency ground loops.
If necessary (to break LF ground loops) the mixer preamp could drive an
Audio output transformer such as the Lundahl LL2811 with mumetal
magnetic screens and internal shielding.
Hybrid feedback can be used to reduce the distortion and extend the low
frequency response.
> Any reason to upgrade the sound card sample clock, I have a clock-block ? 
>
>   
Don't use that it will probably have too much jitter.
> What would be a reasonable common LO oscillator, sine wave XO module ?
>
>   
Dont forget to have an individual isolation amplifiers between each LO
port and the LO itself.
The easiest method is to use a DDS system to generate the required
offset LO frequency.
However bandpass filtering of the output is necessary and care must be
taken to select the DDS clock frequency to ensure that the principal
phase truncation spur as well as other spurs are sufficiently far from
the desired output that they can easily be filtered out. Use the DDS
tools on the Analog Devices site to help with this.
The DDS clock itself needs to have relatively low noise either divide
10MHz by 2 and then multiply by 9 to get 45MHz which appears appropriate
for the AD9850 series chips, or build a low noise discrete 45MHz LO.
> Stanley
>
> I have several HP-8708a NOS that are full of parts they have a 20Mhz XO that would phase lock to the lab standard and provide a voltage of 1 volt per Mhz for an input signal that was the result of another mixer from the Ghz range which I do not have. Thinking of using them one for each DUT. They use the same connector for the backplane as the 10811 but each board is in a smaller compartment, could take three of the six slots but it would require much work and the front part of the case has space. 
>
> Does anyone have any info on this 8709A option HU8 Synchronizer ? Think it was made about 1987.
> _______________________________________________
> 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.
>
>   

Bruce