Philips' Dynamic Noise Limiter

This circuit was developed by Philips in the late sixties or early seventies to improve the subjective signal-to-noise ratio of cassette reproduction. Unlike the various Dolby systems, which require encoding and decoding, the DNL works on playback only and therefore can be used with any source material. Indeed, it can improve the sound of not only tape playback but any source where broadband noise (hiss) is troublesome - noisy FM and AM radio signals are examples.

Philips made the circuit available on a royalty-free basis. The circuit diagram here is based on two published versions; Wireless World (December 1971 pp 585 to 586) and Elektor (January 1984 pp 1-36 to 1-41). There are some differences between the two circuits - mainly component values and part numbers. To accommodate the two designs using one diagram I have listed the component values in a separate table. (WARNING - DO NOT MIX THE TWO DESIGNS - USE ONLY THE COMPONENT VALUES FROM ONE COLUMN OF THE TABLE). The major difference between the two circuits is the limiter made up from D1, D2 and C6 which are absent from the Elektor design. The addition of C12 in the Elektor design reduces the high-frequency gain of T4, possibly to improve stability.

I have built the Wireless World version and it works well. I used BC108 transistors (the BC148 versions have identical electrical specs but are in 'Lockfit' packages). I also used 1N4148 diodes instead of the BA217s which are hard-to-get (obsolete); almost any other type of small-signal silicon diode should be suitable (but don't mix types).

The principle of operation is straightforward - it is only during quiet passages that tape noise becomes intrusive, by reducing the gain for the higher frequencies during these periods an improvement in signal-to-noise ratio is achieved. This has only a minor effect on the higher frequency components of the signal since these are effectively masked by the noise anyway. The effect is progressive with the high-frequency attenuation increasing as the signal level falls. A 0 dBm (775 mV) signal passes through the circuit virtually unaffected (with the frequency response being almost flat); however, as the input falls to (say) 2 mV the attenuation rises to 10 dB at 7.5 kHz and 20 dB at 10 kHz. The effective roll-off is then about 18 dB/octave.

DNL circuit diagram

Because the amount of high-frequency attenuation depends on the input signal level, it is important that the input level is correct. The driving circuit should generate 'line level' at the input - i.e. the loudest portions need to be at 775 mV (RMS). The best way to check this is with a sinusoidal source and a 'scope. For tape use, record a few minutes of a 1 kHz sine wave at 100% modulation and check the level at the input of the DNL, a potential divider (total resistance of around 25 kΩ) can be used to reduce the input amplitude to the correct level.

Closing the switch disables the noise limiting effect giving a flat response (and thus allowing the DNL to be left permanently in circuit).

Component List

Component Wireless World Elektor Comment
R1 68k 270k  
R2 56k 150k  
R3 680 1.5k  
R4 1k 5.6k  
R5 680 1.5k  
R6 8.2k 15k  
R7 1k 2.2k  
R8 330k 680k  
R9 100k 180k  
R10 1.8k 3.9k  
R11 330k 330k  
R12 22k 22k  
R13 680 680  
R14 4.7k 5.6k  
R15 330k 120k  
R16 22k 22k  
R17 120k 220k  
R18 680 680  
R19 4.7k 6.8k On WW version: variable 0 to 4.7k preset
C1 4.7µF 4.7µF  
C2 22nF 4.7nF  
C3 3.3nF 1.8nF  
C4 470pF 270pF  
C5 1.8nF 1.5nF  
C6 22nF not fitted forms limiter with D1 & D2 in WW version
C7 330pF 680pF  
C8 4.7nF 22nF  
C9 4.7nF 22nF  
C10 10nF 4.7nF  
C11 4.7µF 4.7µF (16V)  
C12 not fitted 2.2nF  
C13 not fitted 10µF (16V) Not shown on WW circuit but should be used
T1 BC148B/BC108B BC547B  
T2 BC148B/BC108B BC547B  
T3 BC148C/BC108C BC547B  
T4 BC148B/BC108B BC547B  
D1-D2 2 x BA217 not fitted forms limiter with C6 in WW version
D3-D6 4 x BA217 4 x 1N4148  
VCC +14V +12V  

I haven't included anything like a detailed description of the circuit operation as yet - please let me know if you feel this would be useful.

Last updated: 8 August 2003;   © Lawrence Mayes, 2003