linrad support: Anti-alias filter for Delta44
(April 13 2002)

Purpose of the anti-alias filter

This filter is part of the 2.5MHz to audio converter.

The Delta44 has (of course) its own built-in anti alias filter. Since the board is desined for music aliasing, near 48 kHz is not an issue - the filter is not designed for an extreme cut-off. It is designed for amplitude flatness and phase response at frequencies well below 48kHz.

One could perfectly well use the signal directly from the mixer to feed the Delta44, but then it will not be possible to tolerate signals above the level that saturates Delta44 in the region from the passband edge to as far outside (50kHz or so) as required for the RF filters to attenuate the signals. Without anti-alias filter the useful bandwidth is limited to about 43kHz because of aliasing.

The anti-alias filter incorporated in the 2.5MHz to audio converter has some gain below 48kHz and attenuation above. The mixer saturates 17dB above the Delta44 which means that the filter extends the dynamic range by 17dB for frequencies more than 1kHz outside the visible passband. (In the frequency range 48 to 49 kHz saturation of the Delta44 is the limiting factor) The attenuation at 47kHz is about 6dB so the useful passband is extended to 47kHz by this filter.

Schematic diagram and frequency response

The schematic diagram is shown in fig. 1 and the frequency response in fig. 2.

Fig.1. Schematic diagram of the anti-alias filter.

Fig.2. Frequency response of the anti-alias filter from input to output.

Individual stages

The first stage is a notch filter followed by an integrator. There is some feedback/feedforward to the second stage to improve the attenuation between the two notch frequencies and to make the corner at 48kHz sharper.

The frequency response at the output of the first stage is shown in fig. 3. Frequencies above 48kHz are attenuated by 6dB or more so there is no risk for signals above 48kHz to saturate this stage which is capable of the same voltage swing as the input stage.

In the schematic diagram the notch filter coils are marked 2mH and 20mH respectively. The details of these coils are as follows:

The "2mH" coil is 100.5 turns of 0.20mm enameled wire on an RM6 core of material N48 with an Al-value of 160. The core is not critical, other materials and Al values will also work fine. The coil has to have a parallel resonance at 55kHz when parallel with 4.7nF.

The "20mH" coil has to have a series resonance with 470pF at 49.2kHz. It is wound with 340 turns of 0.10mm wire, also on an RM6 core with Al=160 made from N48 material. Also for this core the Al-value and material is not critical.

Fig.3. Frequency response at the output of the first stage.
The second stage is a a third order low-pass filter of Sallen/Key type. The frequency response at the output of the second stage is shown in fig. 4.

Fig.4. Frequency response at the output of the second stage.
The third stage is similar to the second stage and it is added for better suppression of spurs caused by extremely strong signals in the range 20 to 100kHz from the visible spectrum edge.

The last stage is a differential amplifier which is needed to reference the output signal to the analog ground of the Delta44. The analog ground must not be current loaded so it can not be connected to the ground of the 2.5 to audio converter.

Spur suppression for extremely strong signals outside the passband

A very strong signal from an old vacuum tube signal generator was swept from 1 kHz outside the passband, 2.549kHz and upwards. The signal level was +1.4dBm, 15dB above Delta44 saturation or about 160dBc/Hz.

The level of the alias spur was measured and is listed in table 1.

Freq    alias freq     level
(Mhz)     (kHz)         (dBc)
2.549      47           -30
2.550      46           -49
2.551      45           -66
2.552      44           -76
2.553      43          -108
2.554      42          -115
2.555      41         <-140
2.556      40          -132
2.557      39          -130
2.558      38          -133
2.559      37          -118
2.560      36          -120
2.561      35          -120
2.562      34          -116
2.563      33          -115
2.562      32          -132
2.563      31          -125
2.564      30          -132
2.565      29         <-140
2.566      28          -125
2.567      27          -134
2.566      26          -114
Table 1 Suppression of alias signal.
The waterfall diagram in fig. 5 shows the signal swept from 2.546 to about 2.090MHz.

Fig.5. Waterfall graph while a very strong signal (level=140dB on dB scale) is swept from 2.046 to about 2.090MHz. The white area at the bottom is due to Delta44 saturation below 2.048kHz. The white line near the top is when the signal generator is switched off.

Fig.6. Channel 1 Q. (For component numbers.)

Fig.7. Channel 2 I. (For component numbers.)

Fig.8. Channel 2 Q. (For component numbers.)