To Erik: Hi. The PNA block diagram I posted earlier had two errors. I corrected them in PNA V2.0, as shown in PMA2.PNG
The first error was a missing connection to the Phase-Frequency Detector (PFD). This was a trivial error and was due to issues with the image viewer dropping lines. Do not mess with the Irfanview viewer. I switched to the XNview instead, which gave good performance. So if you find problems with missing lines in the PNG files, try a different viewer. The second error was a missing 90 degree coupler to the Double Balanced Mixer (DBM), which you generate with discrete components. If you are interested, Minicircuits makes 90 degree hybrids for many different frequency ranges. For example, the https://www.minicircuits.com/pdfs/JSPQ-65W+.pdf is useful over the 5 MHz to 65 MHz range, and appears to be dead on 90 degrees at 10 MHz. When you get up to 100 MHz, the https://www.minicircuits.com/pdfs/PSCQ-2-120+.pdf gives about 0.5 degree error at 100 MHz. The advantage of a 90 degree hybrid is very low insertion loss and a 50 ohm environment. This eliminates the noise generated by resistors and op amps. Speaking of op amp noise, the lowest noise op amp I know of is the LT1028. The datasheet is shown at https://www.analog.com/media/en/technical-documentation/data-sheets/1028fd.pdf It has 0.85nV/vHz typ at 1kHz. However, if you look at the Voltage Noise vs Frequency curve, you find the noise increases drastically below 10 Hz. One way to get around this problem is a zero drift chopamp. An example is the ADA4523, shown at https://www.analog.com/media/en/technical-documentation/data-sheets/ada4523-1.pdf As shown in Fig. 2, the noise is below 5nV/(sqrtHz) from 0.02 Hz to 10 KHz. This might be useful in some designs. Another item that might be of interest is the PFD. The Hittite HMC984LP4E has -231 dBc/root(Hz) of noise, which is quite low. The datasheet is at https://www.analog.com/media/en/technical-documentation/data-sheets/hmc984.pdf I noticed in your phase noise plots numerous spikes starting at 50 Hz. These clearly show you are living in Europe. What a wonderful place. I lived in Metz, France for a number of years while I was with NATO. Thanks for the memories. However, the noise spikes show another problem. They may be due to noise on the VCC supply to the oscillators. This can be a serious problem, as John Ackermann showed in his page at https://www.febo.com/pages/oscillators/wenzel_uln/supply.html You can reduce this with a simple ripple filter, shown in 2N3906S.PNG. This reduces the noise by 80 dB at 10 Hz, and 120 dB at 100 Hz, as shown in 2N3906G.PNG. If you are interested in modifying the design, the schematic entry for LTspice is shown in 54E696BA.ASC. The corresponding waveform plot file is 54E696BA.PLT The advantage of this configuration is easy to filter negative voltages by switching to a NPN transistor, such as a 2N3904. The disadvantage is this filter is only capable of supplying limited current. It cannot supply the current needed by the OCXO heaters, so if this is a problem, you might consider switching to VCXO oscillators. I am searching for ways to increase the maximum current, but it is very difficult to keep the attenuation numbers. I have found a MOSFET that gives good attenuation, but it can only handle 1.4 Amps. It is the Onsemi FDC2512, and the datasheet is at https://www.onsemi.com/pdf/datasheet/fdc2512-d.pdf It might be necessary to use two separate filters, but this might be an advantage by keeping crosstalk low between the oscillators. If you would like to eliminate the problem of quadrature lock, the Hittite HMC984LP4E PFD might be of interest. The -231 dBc/Hz of noise is very low and might be hard to reach with a DBM. If you are interested in following up on phase-frequency detectors to eliminate the narrow lock range of double-balanced mixers, I can supply you with a wealth of information on the design, implementation, and testing. Just let me know if this would help. Mike
54E696BA.ASC
Description: Binary data
54E696BA.PLT
Description: Binary data
_______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe send an email to [email protected]
