Moode Forum

Full Version: My second hybrid amp, tube + class D
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
I present to you my second construction - a hybrid tube amplifier plus a class D power stage. (First, prototype construction basen on nuvistor 6C51N from Soviet Union and Philips magic eye DM70).

The design assumptions are the use of the TPA3116 integrated circuit, the use of readily available and cheap lamps. The design is maximally compact. Powering from the external impulse power supply. PCB board designed so that it can be closed in a housing adapted to its dimensions and in a larger housing with an attached mains supply, a switchable input block. Of course, it is possible to use the amplifier without a case, without the required parts of the structure connected with cables or with delicate connectors. Construction is designed for a novaday low impedance output audio sources (<10kΩ), both with commercial output level and high, pro output level.

As for the operating parameters, I assumed a frequency response not worse than 20Hz - 20kHz with a decrease of 0.5dB.

In addition, I decided to switch the sensitivity ranges of the magic eye so that it was alive while listening to music quietly.

Switching on the power is done with a relay with a special contact design, called pre-make tungsten contact, because the low resultant ESR of the capacitors will cause the contacts of a typical relay or a built-in switch to be glued in the volume control potentiometer.

In addition, the mute system is active after switching on the power supply, eliminating two knocks, the first resulting from the smooth start of the anode voltage converter of the lamps, and the second when the cathodes of the lamps reach a sufficiently high emission during their heating.

The next (4th) picture shows the schematic diagram of the amplifier.

The circuit, apart from the audio path, includes two DC-DC converters. The first buck delivers a tube filament voltage of 6.3V. The second boost provides anode voltage to the tubes equal to 52V. With this inverter, you will use the voltage multiplier trick and you will get a second voltage of 206V for powering the magic eye.

The boost converter has one unusual solution, an additional resistor R44 simultaneously loading the converter and forcing a continuous flow of current through the coil. The first ones are required at low load power and, at the same time, with low ripple voltage at the output of the converter, it may work unstably and fall into oscillations in the range of the acoustic band. The second is desirable to minimize electrical noise.

The aforementioned power-on mute circuit was realized almost forcefully with the use of one transistor T2 and the RL2 relay.

The magic eye works in an special arrangement where the triode of this tube is used twice. The gain is to increase the sensitivity of this indicator and obtain a characteristic close to the decibel scale. Additionally, a relatively low time constant for the rising level (attack) was obtained with a higher time constant of the falling of level indication (release).

The zener diodes D1 and D2 protect the TPA3116 inputs against overvoltages below -0.7V and above + 6.2V. Another overvoltage protection (overvoltage from normal and defective magic eye circuit) of the TPA3116 integrated circuit is the D3 diode together with the R23-R26 resistors.

The resistors in the signal path are of the metal-film type (MELF and miniMELF), the remaining ones are thick-film ones.

The coupling capacitors in the signal path are polyester capacitors (red visible in the photo of the amplifier). The other capacitors in the signal path (also Zobel and snubber on the TPA3116 output) are ceramic MLCC C0G (white visible in the photo of the amplifier). Also the two decoupling capacitors 1n0 C31 and C32 TPA3116 are of the C0G type. The others are X7R (beige, visible in the photo of the amplifier). This type of dielectric used in TPA3116 bootstrap capacitors has no effect on sound quality - C15-C18). The electrolytic capacitors used for the filtration of the anode voltages are of the hybrid type (OSCON) offering low ESR ensuring high suppression of the supply voltage ripple. Capacitors C35-C40 are aluminum of higher quality. As they are connected in parallel their resultant ESR is very low.

Magic-type eye sensitivity switch offering the same height above the button PCB as for the axis of the volume control potentiometer.

TPA3116 working with 2MHz switching clock frequency.

The next pictures (5 and 6) show the PCB design of the main PCB and the small one that is the base for the magic eye, which is placed horizontally.

The pattern of the tracks is important to achieve a low level of self-noise and distortion. I separated several grounds, one basic one related to the TPA3116 circuit, two grounds for each tube preamplifier and the last one for DC-DC converters.

The last four pictures show the PCB rendering, each on both sides. Both plates are double-sided, FR4 laminate 1.55mm thick, copper thickness 35um.

Basic amplifier measurements were made using the Focusrite Scarlett 2i2 2gen audio interface. 24 bit, 192kHz sampling was used. Laptop audio interface. Audiotester, ARTA, STEPS and RMAA programs were used in the Windows 10 21H1 64-bit environment.

The load of the amplifier were power cemet resistors with a resultant resistance of 8.3Ω. Parallel to these resistors are resistive voltage dividers with an RC filter suppressing the residual frequency of the power stage keying.

The waveforms obtained are the result of measuring both the amplifier and the audio interface itself. I do not include measurements of the audio device itself. Ultimately, the real performance of the amplifier is slightly better than given below.

The first graph is the sweep frequency response.

The second and third graphs are the harmonics distribution for 0dB and -20db levels (blue line only).

The fourth graph shows the inter-channel crosstalk.

The fifth graph shows the noise and distortion level relative to the 0dB level. Note - the tested amplifier was not enclosed in a scanned housing, the constant noise 50Hz, 100Hz, 150Hz ...

The sixth plot shows the intermodulation distortion at full undistorted power output.

The seventh graph shows the measurement of the passband using the multitone method.

The eighth last graph is the frequency response of nonlinear distortions with additional breakdown into second to sixth harmonics. Measurement for -3dB level.

For passband measurements - output impedance os signal source is less than 1 kilohm!. 

The supply voltage at full output was 23.9V.

The RMS output voltage of the sine for the adopted maximum power point without distortions, i.e. the conventional 0dB, is 15.2V (28W on 8Ω). It has an RMS voltage of 130mV at the input. For the commercial level, i.e. 316mV, this is a 6dB additional margin of the input sensitivity.