299

ASPECTS OF DIRECT CONVERSION RECEIVER DESIGN

N C

Hamilton

Radio Society of Great Britain

UK

I " This paper describes various symptoms that can be shown by direct conversion receivers. It explains their causes, and describes some cures. Broadband sub-harmonic mixers are then considered. Finally, a design for a prototype receiver will be described. An attempt to build a direct conversion transceiver for the 144 Wz amateur band was successful. Because the transceiver was intended t o be reproducible, the deaign went through many iterations. The knowledge gained from this experience has been successfully applied to a number of prototype circuits intended for HF use. These circuits are more complex than those normally used by radio amateurs ( 1 ) . but offer improved performance. This has been achieved by the use .of broadband RF circuits after the preselector. and by reducing LO leakage. Sub-octave preselection, gain and phase matching of I and Q channels, and dc drift are all problems that do not concern the radio amateur, who is quite prepared to build receivers that operate over narrow frequency bands. and is also prepared to tolerate the iuage frequency reception. These problems will not be considered further.

due to LO p o w e r - l i n a . Apart from direct pickup of hum by the early stages of the audio amplifier, hum can be due to LO radiation (2). The LO is first radiated. The radiation induces current in wiring which has an impedance to ground which varies at 50 Hz (eg. rectifier wiring in a power supply). This modulates the induced LO current and the result is then re-radiated. Some of this energy reaches the antenna or some other sensitive part of the receiver, and the hum is demodulated.

AF I n s w i t v due to

strength in the receiver case, any metallic part can be microphonic. including the case itself. %cause most amateur DC receivers suffer from microphony. headphones are aluost invariably used in preference to a loudspeaker: this breaks the mechanical feedback path. the Receiver. In any tion insreceiver, reciprocal mixing problems must be avoided, so it is necessary to minimise the phase noise of the LO. To do this, the inductors of the oscillator must operate with a high stored energy, and this implies large tuned circuit currents and voltages. If the oscillator is running at the receive frequency, this makes satisfactory screening very difficult. The problems of LO radiation and conduction have lead to the consideration of dual be conversion schemes (3). These should approached with caution. They increase both the complexity and the number of spurious receiver responses. Also, with passive ring mixers. leakage from one mixer mixes in the second mixer. and this re-generates the unwanted leakage at the receive frequency. LO . Without amplification prior to the mixer, direct conversion receivers usually radiate an unacceptably high LO power level from the antenna. The radiated power might be -36 dBa. The signal levels in the receiver that give rise to this are given in Figure 2s. Also. the microphony of the tuned circuits in the preselector would be quite high due to LO leakage.

To reduce the LO radiation from the antenna, it is common practice to use an amplifier, as shown in Figure 2b. This is successful in reducing the LO radiation by 2668 due to the amplifier's S,*. But, as the figure shows. this provides only 6dB reduction of the microphony (and the dc drift due to antenna VSWR changes). because the reflected signal is amplified again by 20dB before reaching the mixer.

Lp. This is due to the same mechanism as power-line hum, the difference being that the LO leakage is reradiated from af wiring with af sidebands. See Figure 1. In a receiver with ADC this causes a characteristic "plopping" noise.

To reduce the LO radiation still more, and to give some protection against microphony. an amplifier with an unusually low S . . i s required. Neutralisation could be tried, or circuits using directional feedback (4).

.due to LQ. Most microphony in amateur radio direct conversion receivers is due to the LO feed-through described by Bollinger and Vollenweider (3). Mechanical disturbances which induce a change in the voltage of the LO at the mixer's RF port will produce an audible output from the receiver. For this reason. any tuned circuit in the receive path will give microphonic effects i f it has a high level of LO feedthrough. Indeed, if there is a high LO field

Because of the high natural noise level on the HF band, gain prior to the mixer usually reduces the receiver's dynamic range without improving the S/N ratio. The amplifier's gain of 20dB in Figure 2b can be removed by placing a 20dB attenuator between the amplifier and the mixer in order to give no net forward gain. This is shown in Figure 2c. The LO radiation is reduced by an extra 20dB. and the microphony from the preselector (and dc drift) is reduced by 40dB.

mi I

300

F M i c r o D W . Some microphony is inevitable :ue to the mechanical properties of the af filter components prior to the first af amplifier stage. This has become more apparent using the new operational amplifiers in which the internally generated noise is nearly negligible. Moulded polycarbonate capacitors and some ferrite pot cores have given the least microphony of those tested. Regrettably, all af transformers tried so far have been too microphonic to use directly on mixer outputs without prior sisnal amplification. Spurious Demo-. This problem is (5). Strong discussed well by Gosling signals at the input to the mixer tend to be demodulated to give an af output from the mixer, irrespective of the LO frequency; this is a particularly unpleasant form of interference. It can be minimised. but not eliminated by good design. When used on air, the first sign of spurious demodulation by a direct conversion receiver with poor af filtering is a SkHz whistle. This is Caused by the 5kHz channel spacing used on the HF broadcast bands.

I

itivitv. It has been said that a DC receiver with no RF amplification has an inherently high noise figure, because the mixer spuriously demodulates its own LO noise. This is not so. Well balanced mixers with resistive terminations have very low so are not discriminator constants, sensitive to phase noise. A standard Well terminated +7dBm ring mixer was measured. The discrimination constant was only 1wV/MHz change in LO frequency. On the other hand. the same ring mixer gave a detection constant of l~V/O.OldBchange in LO level. is clearly much more serious. This Fortunately, this problem can be eliminated by using suitable amplitude limiting of the Lo prior to the mixer.

The sub-harmonic mixer has two advantages for the radio amateur. First, the LO drive is at half the RF frequency, so that the "LO leakage" is reduced by the amount of second harmonic suppression in the generator. Second, it is not too prone to spurious demodulation. Also, the LO drive requirement is slightly less than that of the standard ring mixer. Figure 3 shows an example of a broad-band sub-harmonic mixer. This design offers low conversion loss between 0.5 and 500MHz. It also avoids the use of the tuned circuits used in earlier designs, and is thus much less prone to microphony. Figure 4 shows how a standard ring mixer may be used as a sub-harmonic mixer. Unfortunately, all these mixers use their diodes in the linear mode. and so have a third order input intermodulation intercept point of about OdBm/tone.

B PROTOTYPE 7WJ2 DC RECEIVER From the discussion above, it is clear that the most important task of the designer of the direct conversion receiver is control of LO leakage. Figure 5 shows the block diagram of the direct conversion receiver which is to be discussed in detail, and which was designed to minimise leakage.

Lpcal Oscil-. The master oscillator runs at a multiple of the LO frequency. in order to eliminate radiation at the receiver frequency from the oscillator COmPartwnt. The Output is then passed through a well screened box containing a broad-band frequency divider. The divider ensures good LO amplitude limiting to reduce mixer noise. and provides an opportunity for the provision of accurate I and Q LO outputs.

.-

If the divider, mixer and isolator are in separate compartments. they should be connected using a coaxial cable with a solid outer conductor. High quality threaded connectors must be used. Even at HF, the screening provided by BNC connectors is too low.

.-

The circuit diagram of the signal path from the antenna to the second shown in Figure 6. The af stage is preselector provides filtering of the broadband input. A parallel resonant tuned circuit is followed by a series resonant circuit. This rejects multiples and submultiples of the receive frequency that could cause spurious receiver responses.

-.

Instead of amplifying the receive signal and then attenuating it, as Stated above, the isolator amplifies the signal and then attenuates it due to the mismatch between the high drain impedance of the SFET and the load. The two stage isolator provides broadband operation between 0.5 and 300MHz. It has no forward gain, but a reverse isolation of 80dB. It has the additional advantage of providing a good match to 50 P at both ports.

w.

The sub-harmonic mixer has not been used in this receiver, because the conventional double balanced diode ring mixer with a dc coupled IF port provides higher 3".order intercept points. It also allows the LO to run at a harmonic of the received frequency, as described above. Additional transformers have been added at the LO and RF ports of the mixer. These provide a balanced drive to the mixer, and this, by improving the balance of the mixer. reduces the LO leakage.

.-

The diplexers used here are pairs of HP/LP filters whose inputs are connected in parallel. The input impedance is Z . at all frequencies, when the outputs are terminated in Z.. Diplexer values can be found from the tables ( 6 ) using back-to-back Butterworth HP and LP T filters for Z d O Q . This is a very useful Circuit seldom found in standard filter design books. Two diplexers are connected in series. This presents a broad-band 50P termination to the mixer's IF Port from 0 to over 100MHz. The first diplexer has a crossover point of 1 MHz, and gives a good match to RF. The second diplexer not only provides a good match up to 1 MHz. but provides two poles of af filtering Prior to the first gain block. In order to fit the pole pattern, the values of the components in the second diplexer have been adjusted so that the poles fall at approximately -1 Lf40' on the complex frequency plane normalised to 3kHz. The diplexer still has a worst-case input return loss of 20dB. Only two poles are used. because these components are prone to af microphony.

301

-ier/Filw. The first af amplifier has a gain of 3068. and also has a real pole, such that the amplifier output has an af response that approximates to 3- order Butterworth. The first resistor of the second amplifier/filter stage is placed inside the mixer compartment. This allows the af output to be brought out of the box via a feedthrough capacitor, as a capacitance to ground at this point is required by the af filter circuit. This reduces LO leakage. The subsequent amplifiers are active filters with poles designed so that the overall af response is that of a ghorder Butterworth LP filter. The subsequent AGC amplifier is followed by a 3- order Cauer filter to remove high frequency amplifier hiss when the AGC stage is operating at maximum gain.

EumEKEs 1.

Hawker. P., 1978 "KEEP IT SIMPLE: Direct-conversion hf receivers", Proc. IERE Conf. on Radio Receivers and Associated Systems.

2.

"The ARRL Handbook Amateur" 1990 p 12.8.

3.

Bollinger, U . . and Vollenweider. W.. 1990. "Some Experiments on Directconversion Receivers". Proc. IEE 5Conf. on Radio Receivers and Associated systems.

6.

Koren. v.. 1989 "A new negative feedback amplifier". RF, Feb., pp 54-60.

5.

Gosling. w., 1986 "Radio Receivers" ILE Telecouunications Series 15. p 177.

6.

Besults The frequency coverage of the receiver is controlled only by the LO and preselector. 0y changing these, the basic design gives uniform performance from 1.5MHz to over 15OMHz. The LO screening in this receiver has proved sufficient to reduce the microphony to the level where a loudspeaker is useable. The microphony that remains is due to the AF microphony discussed above.

for

the

Radio

Hamilton. N.C.. 1991 "Improving Direct W Conversion Receiver Design". Vol. 67 to be published in April.

.-

An

input signal of -1176gives an output of 666. This is much more sensitive than is required for the HF band. The spurious demodulation is inaudible for a 95% An signal at -30dBm. The LO signal reaching the antenna socket after two isolator stages is about -100dBm.

S+N/N

CONCLUSION The designer of direct conversion receivers must regard the minimisation of LO leakage as the top priority. The LO leakage can be either by radiation or conduction. Radiation can be reduced by careful screening and by running the master oscillator at a harmonic of the receiver frequency. and then dividing down.

CONSTANT LO LEAKAGE

I NTO HEADPHONE CABLE LEAKAGE MODULATED BY AF OUTPUT

Conduction can be reduced by the use of a broadband isolator circuit in the signal path between the preselector and the mixer. The isolator should have a small forward gain (or loss), and a high reverse isolation. The mixer, frequency divider and isolator should all use broadband circuits, in order to avoid microphony. By using these techniques, conversion receiver can have good performance figures.

the direct surprisingly

MODULATED LO RE-RAD I ATES FROM HEADPHONE LEADS INTO PREAMP.

Figure 1 AF instability due to LO leakage

BI I

I 302

-36dEm RADIATED

\/

UP TO l00uV dc OFFSET DUE TO PRESELECTOR REFLECT ION

PRESELECTOR: l0dE RETURN LOSS 3dE INSERTION LOSS

-43dEm INTO MIXER ___)

\y/

~e

1

PRESELECTOR: l0dE RETURN LOSS 3dB I N S E N Loss

-

AMPLIFIER GAIN -%2

- 6 9 a m LEAKAGE

~m

PRESELECTOR: I0dE RETURN LOSS 3dE INSERTION LOSS

La-

c>

-

UP TO 50uV dc OFFSET DUE TO PRESELECTOR REFLECT ION

4

-49dEm INTO MIXER c--

-59dEm LEAKAGE

b>

\y/

-62dEm RAOIATEO

-26dE REVERSE ISOLAT ION

-33dBn LEAKAGE mE@% - ISOLATION

-82dEn RAD IATE0

-

-89dEm LEAKAGE -79dEm LEAKAGE

NO NET FORNARD GAIN +20dE-20dB-

-

Lo '(InER

UP TO 0.5uV dc OFFSET W E TO PRESELECTOR REFLECT ION

-89dEn INTO MI XER

c26ds 46dE REVERSE ISOLAT ION

4 x 1

ISOLATION'

I Lu runtK

Figure 2 LO radiation and microphony suppression usins amplifier Stag-

RF IN

sz

2s

sz

zs

L ? T

4

18.2"RFC

AF OUT

1-

-+

AF OUT

g--+

Figure 4 SHM using conventional ring mixer

Figure 3 Broadband sub-harmonic mixer (SHM)

1 1 I

- - - - - - - _.I SRAiH, SBl I , e t c .

1

303

7-7.3 I SOLAT I ON MHz

0 I PLEXERS

NO I SE FILTER

AGC

AMPL IF IER

AMP

Figure 5 Block diagram of experimental direct conversion receiver for 7HHz

O/P TO MIXER

r-------------------------RF I N

M I XER COMPARTMENT

I k6

I L - _ _ - - _ - - - - - - - - _ - - - - - - - - - - - -

MIXER

0 I PLEXERS

I80R

I

30dB AMPL IF I ER

I0dB AMPL I F I ER

Figure 6 Circuit diagram of the signal path from the-antenna to the AF stages of figure 5

IIiI

-____

I I