EMRFD Message Archive 775

Message Date From Subject
775 2007-05-20 13:44:12 Kevin Purcell SM5BSZ measurements on diode switching IMD and HD
SM5BSZ has made measurements on silicon diode switching performance
measuring IMD and HD for a range of diodes.

I found this whilst looking for IMD measurements of 1N4148 PN
switching diodes in mixers. But I still haven't found anyone who's
made that measurement.

But these are interesting measurements.

In a diode ring mixer there are two contributions to IMD and HD: that
generated during diode switching when the diodes are non-linear
(where I think schottky diodes have the edge with three orders of
magnitude switching time especially as the frequency increases) and
that generated when the diodes are switched on. The SM5BSZ
measurements numbers be applied to the latter case.

In a +7dBm mixer there is something like 4mA through a pair of diodes
when they're turned on (if I've got my numbers right). Unfortunately
he doesn't measure IMD with diode currents that low in a 1N4148
switches because his mixer IP3 is much higher so they're of no
interest to him.

One observation for the RF experimentalist is the IP3 of the BC457B
base-collector diode exceeds that of the 1N4148 by at least +10dB (at
40mA and even more at lower current). Perhaps one might get
interesting results building a "diode" ring mixer from BC457B base-
collector diodes and measuring their performance. How does that
compare to a commutative mixer in which the transistors switch? How
does it vary with frequency?

BC457B base-collector diodes also look useful for RX switching in
portable receivers with diode ring mixers as their IMD exceeds that
of the +7dBm mixer at reasonable currents.

First a little side track ...

<http://www.nitehawk.com/sm5bsz/linuxdsp/rxhfa/rfamp_hfa.htm>

> A hf receiver may be operated from a low efficiency antenna that
> does not produce much signal. In such situations a low noise figure
> is needed. A good receiver needs a dynamic range of about 100dB
> according to a study by Peter E. Chadwick, G3RZP HF Receiver
> dynamic Range: How Much Do We Need? which appeared in the May/June
> 2002 issue of QEX. The problem is that the third order intercept
> point has to be +33 dBm according to this study and that can not
> easily be combined with a low noise figure. It is not necessary
> however. At times when IP3 has to be around 30 dB, it is ok for the
> noise figure to be relatively high.

Some might argue with this (in fact they have) see KF6DX

<http://www.doug-smith.net/moredynamics.htm>

> Switches are used to switch in and out attenuators and amplifiers
> in order to shift the noise floor and the point of saturation up or
> down to fit the demands of the situation. A 10 dB attenuator will
> degrade the noise figure by 10 dB but it will also shift IP3 10 dB
> upwards for an unchanged dynamic range.

So back to the diodes ....

> An ordinary diode such as the 1N4148 can be used as an RF switch
> with low distortion as long as the current through the diode does
> not vary noticeably over the RF cycle. For use in receiver front
> ends we might want to allow power levels (1 dB saturation) in the
> order of +20 dBm which means that the RF current in a 50 ohm system
> is about 45 mA RMS or 63 mA peak. To allow such signals the DC
> current through the diode has to be about 100mA with a variation
> from 37 to 163 mA.

Yikes. I never actually run the numbers for this scenario. And this
is an RX situation (not a low level driver).

> By using PIN diodes rather than ordinary diodes, one can operate
> the diode switch at a much lower current. The diode switches, like
> everything else is a source of intermodulation. [...] PIN diodes
> are not easily available, the base-collector diode of a transistor
> [BC547B] can be used instead. The data is from 1.6 MHz, the
> advantage of the PIN diode decreases with frequency [...] It is
> immediately obvious from the figure that the transistor diode needs
> much less current to provide low enough third order intermodulation
> [than a 1N4148 silicon switching diode].

> The third order intercept points change with diode current as shown
> [below]. At 40 mA the [BC547B] transistor is 10 dB better than the
> diode and the difference is larger at lower currents. The frequency
> of the measurements is 1.6 MHz, the difference is larger at higher
> frequencies because the carrier lifetime that causes the difference
> will be more significant.

> Table 3 Third order intercept point of a switch with a single diode
> at 1.6 MHz.
>
> IP3 with IP3 with
> Current 1N4148 BC457B
> (mA) (dBm) (dBm)
> 2.5 - +30
> 5 - +36
> 10 +26 +42
> 20 +33 +46
> 40 +40 +50

N7WIM: Using the BC457B even at low current has IP3 numbers well
above the IP3 of a +7dBm diode ring mixer. This sort of switching
could even be used in portable radios.

> The development in semiconductor technology goes towards smaller
> structures and less silicon. The data [above] were obtained with
> BC547B devices from the junk box. When buying new devices in
> December 2003, one gets a slightly "better" transistor which means
> that it is not quite as good to use as a PIN diode. [The table
> below] shows the performance of a single diode of various types
> operated as a switch with a forward current in the on state of 20
> mA and with a reverse bias of 5V in the off state.

> Table 4 Third order intercept point, isolation and insertion loss
> for a single diode operated as switch with 20mA DC current.
>
> Diode IP3(1.6MHz) Isol.(14MHz) Loss(14MHz)
> type (dBm) (dB) (dB)
> BAQ800 +52 -34 0.17
> BC547B(old),b-c +47 -32 0.08
> BC547B(old),b-e +47 -23 0.11
> BC547B(new),b-c +42 -33 0.13
> BC547B(new),b-e +42 -25 0.17
> BF240,b-c +40 -37 0.21
> BF240,b-e +33 -32 0.21
> J310,g-s +43 -30 0.19
> J310,g-d +43 -30 0.19
> BAT42 +38 -28 0.31
> 1N4148 +34 -37 0.31

Note for the J310 he's just interesting in that gate diode. Not the
FET action. Thats thinking different!

> On the basis of table 4 one would like to use the BAQ800 AM PIN
> diode, but that would be a mistake for the first group of diodes at
> the antenna input. Diodes have a capacitance that varies with the
> voltage. Because of that, they generate second order
> intermodulation. The front end of the RXHFA unit uses five filters
> so there are four diodes that produce second order intermodulation.
> Table 5 gives the level of second order intermodulation with
> different diodes used for the four off-state switches with a wire
> replacing the on-state diode.

> Table 5 Second order intercept point for the RXHFA input selection
> switch with the on-state diode replaced by a wire and with four
> identical diodes for the off-state.
>
>
> Diode IP2(1.6MHz)
> type (dBm)
> BAQ800 +64
> BC547B(old),b-c +58
> BC547B(old),b-e +57
> BC547B(new),b-c +61
> BC547B(new),b-e +54
> BF240,b-c +73
> BF240,b-e +69
> J310,g-s +54
> J310,g-d +55
> BAT42 +51
> 1N4148 +68
> None >90

> An inspection of tables 4 and 5 shows that the base-collector diode
> of BF240 transistors would be a reasonable choice for the input
> switch and provide something like IP2=73 dBm and IP3=44 dBm at a
> current of about 50mA. The RXHFA has two channels so the current
> needed for the input switch will be larger than conveniently
> supplied from a 74HC4053. To avoid this problem and to reduce the
> number of inductors needed, the input switch is four J310 in
> parallel. This gives IP2 = +82dBm. IP2 drops rapidly if fewer
> transistors are used. As an alternative a single J108 could be
> used, but that would require a larger voltage swing for the control
> voltage so the total complexity would not differ much.

N7WIM: It's interesting to note that although IP2 is always higher
than IP3 the ranking of HD and IMD isn't always the same for
different devices. The IP2 in the 1N4148 (+68dBm) is higher than the
BC547B "diodes" (54 to 61dBm) but the IP3 is higher in the in the
1N4148 (+34dBm) than the BC547B "diodes" (42 to 47dBm).

It's useful to have real measurements to work from.

Plenty of other interesting items on his site, Linrad especially.

<http://www.sm5bsz.com/>
<http://nitehawk.com/sm5bsz/index.htm> US mirror
--
Kevin Purcell
kevinpurcell@pobox.com