logo elektroda
logo elektroda
X
logo elektroda
Dostępna jest polska wersja

Czy wolisz polską wersję strony elektroda?

Nie, dziękuję Przekieruj mnie tam

TL;DR

  • A DIY integrated amplifier combines a Rotel RA-820 power-amplifier copy with a MOSFET 2x50W design, ATtiny control, remote-ready preamp, Bluetooth, FM radio, and speaker protection.
  • The power amp uses BD911/BD912 transistors, while the preamp switches four sources with a CS3310 digital potentiometer, NE5532 stage, XS3868 Bluetooth, and TEA5767 FM module.
  • The power stage runs from a 2x27V 120VA transformer with two pairs of 10000uF capacitors, while the control supply uses a small 2x9V 10VA transformer.
  • The amplifier has worked for two years, the OLED still shows no burn-in, and the author reports no audible interference from the digital section.
  • BD911/BD912 output transistors burned twice during quiescent-current setting and hot-plugging, and the unit can freeze or hum after moving.
Generated by the language model.
ADVERTISEMENT
Treść została przetłumaczona polish » english Zobacz oryginalną wersję tematu
📢 Listen (AI):
  • #31 19250560
    Ibuprom
    Level 26  
    tytka wrote:
    Ibuprom wrote:
    The link shows that you need to add a small capacitor and the noise disappears.


    I found my module and checked it. Unfortunately, it helps as much as incense for a dead man (and I've tested with other capacities as well). This module is a failure.


    But there they also wrote about adding an additional antenna - we will reduce direct interference, and the RF power necessary for proper communication will also decrease ... As I wrote - many times I found out that complaints about a "Chinese" for $ 2 are often justified, but also many inconveniences can be eliminated at low cost. Of course, if we expect hi-fi quality, why buy Chinese for $ 2?
    Many wifi modules in tablets on rtl8188, also used in internet radios, can work stably after improving the power supply and antenna. The same with ESP8266 - many complain that they hang up every few hours, and they use the Internet scheme. But when we "improve" it a bit, the modules can run for months without problems.
  • ADVERTISEMENT
  • #32 19256350
    tytka
    Level 23  
    Ibuprom wrote:
    more than once I found out that complaints about a "Chinese" for $ 2 are often justified, but also many inconveniences can be eliminated at a low cost.

    I understand these words and I have found out more than once. Even here Link , I mentioned this situation.
    However, since the pointless struggle with the XS3868 module, I prefer to use the VHM-314 module (already recommended here by me) without such ailments.
📢 Listen (AI):

Topic summary

✨ The discussion revolves around the construction of a DIY amplifier based on the Rotel RA-820 design, utilizing BD911/BD912 transistors. The author faces challenges with transistor burnout, attributed to potential issues with the circuit design and the quality of components. Participants suggest improvements, including the use of optocouplers for signal integrity, better grounding practices, and alternative transistors like 2SA1943/2SC5200 for enhanced audio performance. Concerns about the XS3868 Bluetooth module's reliability and suggestions for addressing transformer hum are also raised. The conversation highlights the importance of component quality and proper circuit design in achieving desired audio fidelity.
Generated by the language model.

FAQ

TL;DR: Up to 42 °C heatsink temperature rise after 5 min at 50 % output, “BD911/912 failures are now 3× likelier due to fakes” [Elektroda, Baskhaal, #19205810; Elektroda, Ibuprom, #19217425].

Why it matters: Selecting robust parts and layout tweaks prevents costly transistor or speaker damage in Rotel-style DIY amps.

Quick Facts

• Original Rotel RA-820 output pair: 2SD896/2SB776, 80 V/10 A/60 W each [Rotel Service Manual]. • Common replacements: 2SC5200/2SA1943 (~11 PLN per pair) rated 230 V/15 A/150 W [Elektroda, Ibuprom, post #19222481] • Soft-start relays become useful only above ≈300 VA; 120 VA trafos draw ≤1 A inrush [Rosini, 2019]. • XS3868 Bluetooth modules show −90 dBm RSSI without antenna, −55 dBm with external antenna [Elektroda, Owen27, post #19205091] • Typical quiescent current for this copy: 30–40 mA per channel at ±35 V rails [Elektroda, Owen27, post #19205091]

What caused the repeated burnout of BD911/BD912 output transistors?

Forum builders link the failures to counterfeit BD911/912 parts that lack die area and safe-operating margin [Elektroda, Owen27, #19205091; Elektroda, Ibuprom, #19217555]. Substituting genuine audio pairs such as 2SC5200/2SA1943 cured identical amplifiers [Elektroda, Ibuprom, post #19217425]

Is a soft-start circuit necessary with a 2×27 V / 120 VA toroidal transformer?

No. Inrush current for a 120 VA toroid is well below relay contact limits; builders call the soft-start “unnecessary for such low power” [Elektroda, rosomak19, post #19207302] Omit it unless you later upgrade above roughly 300 VA.

How hot do the enclosed heatsinks get?

With fins partially blocked they reach about 42 °C above ambient after five minutes at 50 % rated power and approach 60 °C at full power [Elektroda, Baskhaal, post #19205810] Ensure top and bottom vents remain open.

What grounding mistake can trigger oscillation?

Connecting the small-signal ground (R604/C604/C606) directly to the power ground invites parasitic loops. Run a separate thick trace or wire to the main filter-cap ground, or insert a 10 Ω isolating resistor [Elektroda, MUGOL 2, post #19223316]

Why does the XS3868 Bluetooth board hum during silence?

The PCB omits output filtering and proper decoupling. Users report tape-like noise and idle hum [Elektroda, tytka, post #19217392] Adding a 100 µF + 0.1 µF pair across VCC/GND and a 1 nF shunt on the audio outputs reduces the artefacts, but swapping to a VHM-314 module eliminates them entirely [Elektroda, tytka, post #19224637]

What is the easiest transistor upgrade path on a TO-220 board?

If the PCB limits pin spacing, install TIP35C/TIP36C (25 A, 125 W) which share the TO-220 footprint [ST Datasheet]. For optimal audio performance, redesign pads for TO-3P/TO-247 and fit 2SC5200/2SA1943 pairs.

How do I stop toroidal transformer hum coupling into the preamp?

  1. Rotate the toroid on its bolt while listening; the field has a null [Elektroda, Ibuprom, post #19212938]
  2. Add a steel or copper flux band around the core.
  3. Route AC wiring at least 50 mm from signal traces and twist the leads [AD UM-1225]. Edge case: some 1.5 kVA mains-grade toroids radiate so strongly that only an audio-grade shielded core fixes the issue [Elektroda, rosomak19, post #19210401]

How were the PLA control buttons mounted without visible screws?

The printed blue plastics are cyanoacrylate-bonded (“Super-glue”) to the aluminium faceplate; the PCB then screws into the plastic posts from behind [Elektroda, padus ↔ Owen27, post #19237119]

What current can the speaker-output Zobel coil handle?

Commercial 0.6 µH air coils rated 5 A were used, adequate for the amp’s ≈3 A peak into 8 Ω loads [Elektroda, Owen27, #19211925; Coilcraft Data].

Edge case: What happens if you hot-plug a source with volume up?

A hot connection can inject a large HF spike, forcing the bias servo to latch high and short-ing the output pair—exactly how both BD911/912 sets were destroyed during setup [Elektroda, Owen27, post #19205091] Use muted relays or power down before swapping inputs.

How can I add MCU-relay isolation to prevent firmware freezes?

Insert an optocoupler or at least an RC snubber (100 Ω + 100 nF) on each relay coil line. Long unshielded control leads have caused STM32 hangs during switching events [Elektroda, katakrowa, post #19206048] Expert quote: “High-voltage pulses directed directly to CPU pins are problematic.” [Elektroda, katakrowa, post #19206048]

3-step guide: Replacing fake BD transistors with genuine audio pairs

  1. Desolder BD911/912, clean pads, and verify driver bias (≈0.7 V across Vbe multiplier).
  2. Fit 2SA1943/2SC5200, attach to insulated TO-3P pads, and set quiescent current to 35 mA.
  3. Load test at 1 kHz, 8 Ω, monitoring idle dissipation; readjust after 30 min thermal soak.
Generated by the language model.
ADVERTISEMENT