Discrete SLIC review and recommendations

Review:

1. In an integrated SLIC, the "talking battery" is typically about -24 VDC and the "ringing battery" is in the neighborhood of -70 to -100 VDC. Notice that both of these levels are negative with respect to ground. For the discrete SLIC design, the ringing circuitry is connected between +30 VDC and -10 VDC. In a telephone, the ringer is AC coupled so for a very short loop application with only one or maybe two phones connected, this polarity difference should not matter. Also, the ringing driver transistors are connected in a bridged H-switch configuration so the ringing signal appears to the load a balanced AC signal. When the phone is off the hook, generally the "talking battery" supply of -24 VDC is used with the voice signals superimposed on top. In the discrete design, the 2-4 wire hybrid function (this converts the two wire full duplex current loop to/from the user's telephone to separate transmit and receive signals for the CODEC) is done with a dual op-amp and some transistors. The signal from the CODEC to the 2-4 wire hybrid modulates a transistor which is connected through a diode to TEL2 and and a transistor to the +30 VDC supply. The other side is eventually connected to ground. The signal from the hybrid to the CODEC is AC coupled. As a result, the battery voltages are mostly positive rather than negative as is customarily done.

2. The ringing signal is a square wave. A discrete astable multivibrator is used to make the ringing signal. Generally, for a multivibrator generating 50% duty cycle square waves, the two timing resistors are equal and the two timing caps are equal. The usual formula for the square wave period is approximately T = 1.38RC. In the circuit, R is 22K and C is 1uF so T=(1.38)(22000)(1x10-6) = 30msec. F=1/T so F=1/.03 = 33Hz. In the US, the standard ringing frequency is 20 Hz while in Europe it is 25 Hz. Both of these frequencies are close enough to the calculated 33 Hz so that the actual frequency produced by the device should be verified to be close to 20 Hz. In other words, component variances and the fact that the formula is only an approximation may mean that it does oscillate at a frequency closer to 20 or 25 Hz than 33 Hz and it should be confirmed.

3. The ringer in the telephone presents a load that is an impedance made up of a resistive part and a capacitive part. This impedance is called the Ringer Equivalence Number or REN. In the US, 1 REN = 6930 ohms in series with 8uF. The resistive part is from the coil in the ringer and the capacitive part is from the blocking capacitor. If two 1 REN phones are both connected to the same line, together they present a load of 2 REN which is 6930 ohms / 2 and 8uF * 2 or 2 REN = 3465 ohms + 16 uF. In Europe, 1 REN is defined differently as 1800 ohms + 1uF. As well, the ringing frequency is 25 Hz instead of 20 Hz. US Bellcore specs call for the central office (CO) line card to be able to drive 5 REN (= 1386 ohms + 40uF) to at least 40 Vrms for ringing. This is the key here, the telephone handset requires a minimum of 40 Vrms to ring. A central office may be miles from the user's phone so there could be as much as 1500 ohms of wire resistance in series. This means that the CO must be able to drive over 90 Vrms at the line card to be able to get 40 Vrms at the phone. For this very short loop application, the series resistance of the wire will be much less and thus the circuit needs to drive much less.

4. The above described multivibrator has the collectors and timing resistors tied to ground instead of VCC. The transistors are NPN. The emitters are tied to the collector of NPN Q25 which has its emitter tied to the -10VDC supply. Through a PNP wired as an inverter, a logic 0 asserts the input F_DRING_N which turns Q19 on which turns Q25 on which pulls the emitters of the multivibrator to -10V, thus starting the oscillation. The NPNs and PNPs and the associated diodes and resistors are arranged in a bridged H-switch configuration. The emitters of the PNPs are tied through 22 ohms to +30V and the emitters of the NPNs are tied in common to the emitters of the multivibrator and the collector of the pull down transistor, which is connected to -10V. The collectors of each side's NPN-PNP pair are connected though a diode together and the load (the user's telephone) is connected between each side. The bases of the lower two NPNs are connected to the collectors of the multivibrator. The bases of the upper two PNPs are connected through resistors to the opposite NPNs collectors. Thus when the multivibrator flips one way, one side's NPN turns on and connects the load to to -10V while the other side's PNP turns on and connects the other side of the load to +30V. Steering diodes and a transistor ensure that when F_DRING_N is deasserted and brought back high, the bridge is always flipped the same way so that the transistor is on and the other is off and the transistor which connects to the 2-4 hybrid is correctly biased. When ringing, assuming no voltage drops across the transistors, diodes, or resistors, 40V is passed one way through the load. When the multivibrator flips the other way, 40 V is passed the other way through the load. Therefore the square wave looks to the load (the ringer) as an 80 V peak to peak signal.

5. In reality there are voltage drops across the devices. When the upper right side is on, the lower left side is on and current flows from TEL2 through the ringer to TEL1. (TEL1 and TEL2 are connected through a relay to pins 3 and 4 respectively of the RJ-11 connector to the user's telephone.) From the +30V supply, there will be a drop across the 22 ohm series resistor, then a Vce drop across the PNP, then a P-N drop across the diode, then a Vce drop across the NPN and finally another Vce drop across the NPN to the -10V supply. Assuming all the Vce's and the diode drops are .6V each, then .6 * 4 = 2.4V total drop. With a 1 REN load, the 22 ohm resistor drops about 0.1V and with a 5 REN load it drops about 0.6V. In the worst case, the square wave that the ringer sees is about 37V peak = 37 Vrms and not 40 Vrms.

6. If the square wave is perfectly square, it's crest factor (CF) is 1; a sinewave has a CF of 1.414. A 40 V peak square wave also is 40 Vrms while a 40 V peak sinewave is 40/1.414 = 28.3 Vrms. From the above it is seen that one gets more RMS voltage output the squarer the wave form. The Bellcore specs call for a CF of 1.35 to 1.45 while other specs are looser allowing a CF from 1.2 to 1.6. The Bellcore spec is tight because it is meant for CO to home/office uses where the cable run is long and in a common bundle with other lines. The higher harmonic content of the lower CF signal could couple into the other lines. For very short loop applications this is not a problem. If the RMS voltage is high enough, a square wave will do. The commercially available integrated SLICs generally use a trapezoidal ringing wave form which is actually just a square wave with slew limiting on the rise and fall which reduces the higher harmonic content. Thus they can meet the short loop ringing requirements without resorting to a very high battery voltage. If the crest factor is 1.25, 40 Vrms has a peak voltage of 50 V. If the circuit drops are 3V, this needs a 53V battery supply. Note that this does not take into account any additional wiring resistance. A sine wave has a CF of 1.414, resulting in 40 Vrms yielding 57 V peak. For the same drops, this needs a 60V battery supply. The commercial SLICs RC filter a logic level square wave and then use the tip and ring linear amplifiers to drive this trapezoidal ringing wave. The design uses transistors operating as saturated switches for this function so it would be very difficult to control the CF. There is a 1nf capacitor across the TEL1 and TEL2 terminals but that is not enough to have an appreciable effect on the CF. It is intended mainly for EMI filtering. A large enough cap to adjust the CF probably would affect the voice range frequency response adversely. On the other hand if only one or two phones are connected to it in a very short loop, the square wave that it drives out is probably OK as is, provided there is enough RMS voltage.

Design Recommendations:

• Change the -10 Vdc supply to -18 Vdc to increase the "ringing battery" to 48 Vdc

Assumptions:

1. a square wave ringing signal with CF = 1
2. 200 feet maximum of wire which is 20 ohms
3. 5 REN load (R=1386 ohms)
4. 40 Vrms minimum at the handset

The peak ringing loop current will be approximately .029 amps (1386 ohms with 40V). This means that the drop in the wire will be about .6V. The drops in the transistors, diodes, and resistors in the circuit contribute 3V of drop. So, at a bare minimum, the total supply voltage should be 40 + 3 + .6 or 43.6 Vdc. Increasing the voltage to 48 Vdc would probably provide just enough extra margin. This means that if the +30V supply stays the same, the -10V supply should be changed to -18Vdc.

• The -10V supply is derived from a set of windings on the switching power supply's transformer. It appears to be un-regulated as does the +30V supply which comes from another set of windings on the same transformer. This change would require a change in the windings of the transformer EFD20 (T1).

Items to check:

1. The transformer described in the data sheet for the On Semiconductor MC34166 switching regulator chip should be evaluated further because it may be larger than required. 42 turns of #16 AWG wire is a substatial amount of wire. It is possible that the current requirements are less for the application and a smaller version of this transformer could be used.

2. The -10V supply also supplies a 79L05 type -5V regulator. If the -10V were changed to -18V, the power dissipation in the 79L05 should be checked to make sure it can handle the increased overhead. The -5V appears to be connected to two opamps, the two CODECs and two CMOS switches.

3. Also to be checked is whether this change in -10V to -18V affects the ring trip detection and loop trip detection.
   • There only appears to be a circuit for loop detection (which is also called switch hook detection).
   • Ring trip detection checks for when the user picks up the phone while it is ringing. The phone presents a much lower impedance when it is off the hook and the high ringing voltage can cause too much current to flow. Commercial SLICs will signal when the user picks up the phone while it is ringing so that the processor can turn off the ringing signal and change the battery voltage to the lower talking battery which lowers power dissipation . This function is not present in this design. When the system is not ringing and on the hook, the phone presents a DC resistance of greater than 10K ohms. (Actually it should be an open but the specs allow for a minimum leakage equivalent to 10K.) The range is 10K to infinity. When the user picks up the phone, and it goes off-hook, the DC resistance is less than 500 ohms. This change is detected by measuring a voltage developed across a 100 ohm resistor connected to the emitter of the transistor. When the signal is deasserted high (ie not ringing), the transistor turns on, which in turn connects the TEL1 terminal to ground through a diode, andother transistor and resistor. When the user takes the phone off hook, a DC current begins to flow, which develops a voltage across a resistor, turning on a transistor which brings the signal low which tells the processor the loop is closed and the phone is off the hook. Since there is no ring trip detection, it should be checked to see if the increased ringing voltage could cause damage either to the user's telephone equipment (which should have it's own protection) or to the the circuitry. If it could cause damage, a ring trip detection scheme would need to be implemented.

4. Emprical measurements should be taken to ensure that the multivibrator actually oscillates at 20 Hz and not a higher frequency. If it is not 20 Hz, the discrete component values should be adjusted for 20 Hz operation.

5. The specifications of the transistors should be checked to ensure that increasing the voltage will not harm them. A quick glance at the General Semiconductor selection guide shows the BC846 and BC556 transistors to be 65 V and 100 mA devices. Assuming adequate overvoltage protection, this should be adequate.

6. Since this is all analog circuitry, all of the above changes should be checked and prototyped under real operating conditions. In particular, the switching regulator design must be breadboarded.

Adapticom Inc.
Raleigh, NC
919/870-0608