Keywords: Transformer Saturation/Non-Linearity in RND Red Silk Texture Control Neve Shelford/Portico 511 Channel
The Portico 511 is an API-500 format mic pre-amp. I purchased some to use in a small recording studio. I was intrigued by the so-called "Silk Red Texture" control and so had a look at PCB and traced out the circuit.
NE 5534 op-amps are used throughout, with external compensation capacitors fitted where used in low-gain configurations, which is most of the time. The signal chain is singled-ended through the HPF and Silk circuit with the output transformer restoring differential operation.
The multi-position main gain control switch alters the resistance between the input transistor emitters.
The input amplifier is fully differential and uses PNP input transistors buffered by Texas Instruments NE 5534A op-amps (4nV/rtHz). PNP transistors generally offer lower base spreading resistance. This base resistance is commonly the dominating source of thermal noise in such amplifiers (that are not transformer-coupled on the input) and the choice of transistors affects the overall noise floor of the unit. The ones in this unit are small SOT-23 devices whereas in many similar product much larger power transistors are commonly used owing to their correspondingly lower base resistance.
The main, switched, gain control adjusts the coupling between the emitters of the input transistors in the classical long-tailed pair killer configuration. For high gain, the emitters are tightly-coupled and so each sees a low impedance from the other.
The differential output from the input amplifier is passed to the phase cross-over switch and then to the single-ended convertor that uses a further two NE 5534s. This section also includes the 6dB gain trim control. At low input gain settings, owing to the long-tail kill, the common-mode rejection of the input amplifier is defeated and so it is here in the single-ended convertor that rejection occurs.
The sweepable high-pass filter uses an NE 5534 and a double-ganged pot in a straightforward Sallen and Key configuration. The bypass switch isolates and removes this component entirely from the signal chain.
The 8-LED level meter uses a pair of quad op-amp/comparators to generate the segment control.
The output stage consists of an NE5534 whose output is buffered by a totem pole of NPN power transistors. The upper one operates as an emitter follower and the lower one as a constant current stage. This is directly coupled to the output transformer primary.
The output transformer is the large unit labelled 5052. The other, smaller inductor in the picture is the HF common-mode rejecting balun on the input side.
The output transformer has two separate secondaries: one provides the output from the unit and the other provides negative feedback to the output driver.
The silk RED texture control, as turned up, applies an increasing DC bias offset of up to 0.6 volts to the primary of the output transformer. The primary is two windings in parallel with a resistance of 38 ohms, so the maximum standing current is 16 mA. This generates an even harmonic distortion arising from asymmetric core saturation. The second gang of the texture control also reduces the negative feedback around the output stage and provides treble boost. Lower frequencies will saturate earlier when content-energised, but a standing DC current will effect all frequencies.
The datasheet says the maximum harmonic distortion, when this control is fully engaged, is 2 percent.
The small, unpolarised capacitor values are not readily reverse engineered without desoldering, although perhaps I could measure the transient response a little with the scope and a square-wave input. I measured 10 nF value for C65 and C64 while in circuit. Combined with 2k2 in the Sallen/Key low-pass configuration, this would give a cut-off that limits the output stage to a bandwidth of about 7.2 kHz. However, bench measurements confirm the published data that the response does not start to roll of until just above 100 kHz. If there were a lot of open-loop gain, the negative feedback from the output transformer would tend to extend the bandwidth a little, but we have a low-gain configuration. I expect I can work this out, but if you are reading this Rupert, perhaps you can explain why the bandwidth extends so much.
There seems to be a little bit of uninformed discussion on Gearslutz and other sites regarding the Silk and Tape Emulation RND circuits. This posting may help dispel some of it by revealing some of the truth for this transformer saturation design at least!
DJ Greaves, June 2017.