EMRFD Message Archive 4586

Message Date From Subject
4586 2010-04-27 03:38:27 Heikki 12 MHz if for tx applications
Hello group,

Just a brief inquiry whether anyone has experience on 12 MHz IF for transmitting

Background: My current SSB generator has an if = 10.050 MHz which is not very suitable for conversion into the 30 mb. As I acquired a big bunch of 12.228 MHz crystals, I started on a new generator based on two 6-pole ladder filters, one for normal operation and another one for the rf clipper as explained e.g. in EMRFD. What worries me a bit is the if level (feedthrough) after the tx mixer (SBL-1 in my case), particularly in the 20 mb. Since I use 26.2 MHz LO the image frequency is not a problem, but the if less than 2 MHz away from the target frequency. Two pole LC-filter is what I am planning to apply in the ham bands I will include (80, 40, 30, 20, 17)but is this sufficient for 20 mb. If not, at least two possibilities exist in my opinion:

1) applying a multipole filter for 20 mb with steeper skirt selectivity
2) a notch filter for the if before all the bandpass filters

Personally I am in favor of the notch filter which is rarely seen in the current designs, primarily because most if stages are intended for transceivers where a notch filter should be disconnected when receiving. I am very much in favor of separate tx and rx, and do not have this complication.

The two pole bandpass filter for 20 mb seems to have abt 15 dB attenuation at the 12.2 MHz and the DBM should provide abt 30- 40 dB isolation (right ???) which add up into 45 - 55 dB (sufficient ???).

I would imagine that 30 mb which is also abt 2 MHz away from the if is not as problematic since the LPF after the PA would take care of the IF feedthrough. This, however, is not the case with 20 mb and since I anticipate to spend a major part of my operating time here, I have some doubts concerning my design.

Comments and suggestions are very welcome !

BR Heikki (OH2LZI)
4587 2010-04-27 06:43:24 ajparent1 Re: 12 MHz if for tx applications
4592 2010-04-28 00:12:40 Heikki Re: 12 MHz if for tx applications
Tnx for the comments! It is true that one should select an IF carefully and preferably prior to the design and construction phase but as I already explained the crystals (a few hundred of those) were there ...

Which gave me an idea of a crystal notch filter, but in a quick search I have not found any good references of their application and design. If I have understood correctly, at least W7ZOI has used one in his spectrum analyzer project.

After a preliminary testing of my SSB generator/transverter combination the lower bands 160 m - 40 m seem to be quite OK, even with a two.pole bandpass filter. 30 m is acceptable but 20 m just terrible. More filtering is obviously required, otherwise I need to limit my operation to the lower bands only !


Heikki (OH2LZI)
4593 2010-04-28 06:18:33 Tim Re: 12 MHz if for tx applications
Certainly in the "olden days" an IF trap was put on all radios where the IF was anywhere near the RF band, by default. For example, a 455kHz trap was de rigeur on a BCB-band (540 to 1600kHz) receiver - there the difficulty was having 455kHz near the low end of the BCB band. Traps and IF tweaking systems worked around 450 or 460kHz beat notes between two local stations. A very basic part of the alignment of any such receiver was to hook up a signal generator and tweak the traps as they were sharp tuned circuits and having them drift a little bit really did impact performance.

I don't think the LADPAC tools allow direct design of a notch or band-reject filter (although I'm sure some network inversi
4608 2010-04-29 13:55:29 w7zoi Re: 12 MHz if for tx applications
Hi Heikki, and gang,

In a recent post about your transceiver you commented that perhaps I had built some crystal notch filters. I did just that a couple of years ago and the results were then presented in QEX for July/August 2008, "Oscillator Noise Evaluation with a Crystal Notch Filter." The filters I built used from 3 to 5 crystals to obtain a notch that was from 75 to 100 dB deep and fairly narrow. The filters followed oscillators with the output applied to a spectrum analyzer or general coverage receiver. The oscillator was tuned to place the carrier within the notch. It was then easy to see the noise sidebands once the carrier was attenuated. Spurs in a DDS were also easily observed.

The filters were easily designed. Each crystal is a shunt element. Then there is a quarter wavelength of transmission line between each crystal. A real transmission line is not practical at the low frequencies of these experiments, so I used a synthetic line, which is just a pi network with each element having a reactance equal
to Zo. It was useful to transform up to a higher Z0 to get enhanced notch depth. Once away from the carrier by a kHz or two, there was minimal attenuation.

This is certainly not the ultimate way to measure oscillator noise. The major problem is that this is just a single frequency solution. However, the filters are cheap and easy to build and it is very practical to build them for many available frequencies. Mine are at 5, 7.37, 10, and 11.06 MHz. The other difficulty is that the observed noise is the sum of both phase and amplitude noise. As such, results are a worst case with respect to phase noise.

There is no design tool within LADPAC for notch filters. However, the filters can be "built" within LADBUILD and then analyzed with GPLA. That software is
4610 2010-04-29 16:35:27 ajparent1 Re: 12 MHz if for tx applications