AudioLogs

03/09/2026

RT TPA3116D2, Class D with BD modulation measurements, which you probably haven't seen yet, prove that the chip is good, it actually works as written in the datasheet. Class D amplifier with BD modulation, for short speaker cables does not need an LC filter. I added the filter for signal measurements, for Class D. Voltage is 13.5V, output load (dummy Load) is 4E.
-picture 1. we see the PWM generator, connection to gnd and one output, i.e. one of the jumpers on the output, before the filter (probe 20x)
-picture 2. we see the PWM generator, connection to gnd and one output, i.e. one of the jumpers on the output, before the filter, more finely (probe 1x)
-picture 3. noise on the output, when there is no signal, TPA3116D2 is on mute, noise is 20mV Vpp, AC ripple
-picture 4. we see the BD modulation clearly, i.e. a small phase shift and a dignal in it, this is before the filter, above is + half, below - half of the signal
-picture 5. at a small input signal, before the filter, modulation is taking place, these peaks are seen as a break during the BD modulation phase
-picture 6. at a larger signal we already see a 1kHz sine, a test signal in modulation, before the filter
-picture 7. 1kHz after the filter, the filter cuts off the high-frequency PWM modulation D class
-picture 8. output filter alloy coil 10mH two pieces in series and capacitor 470nF in parallel, load is 4E
-picture 8. TPA3116D2 smallest amplifier module
-picture 10. output filter on measurement
-picture 11. oscilloscope calibration at 1khz, (probe 20x)

24/08/2026

How to make a static image more dynamic with a little extra. Back when we still used cassette tapes, this was a great way to show that something was happening or PLAYing instead of spinning the cassettes.

22/08/2026

How to build a filter for a class D amplifier for a pro version of the TPA3116D2 when we have long speaker connection cables.

Although the TPA3116D2 has a fixed frequency (e.g. set to 400 kHz), this only means that each switching cycle starts at exactly the same time (every 2.5 microseconds). However, within this fixed time, the length (width) of the pulse is constantly changing.
In BD-modulation (MODSEL = GND), the chip exploits this fixed frequency in a very clever way by using two separate PWM signals that are phase-shifted from each other.
Here is a detailed description of how the fixed frequency chip controls the OUTP (positive) and OUTN (negative) outputs:
1. Idle state (Silent) — Switching without current
In the silent state, the chip switches both outputs with the same fixed frequency and the same duty cycle (50%), but with a small phase shift (in BD modulation, the switchings do not occur simultaneously).

When OUTP is turned on at 12.5 V, OUTN is already at 12.5 V. The difference between them is 0 V.

When OUTP drops to 0 V, OUTN is already at 0 V. The difference between them is 0 V.

Since the difference between the two outputs is almost always 0 V throughout the entire cycle, almost no current flows through the speaker. The fixed frequency runs in the background, and the speaker is idle.

2. Positive part of the music wave — Changing the width within a fixed time
When the music requires the speaker diaphragm to move forward, the chip does not change the frequency (it still remains, for example, 400 kHz), but changes the pulse width (duty cycle) within this fixed window:

OUTP gets a wider pulse (on for, for example, 75% of the fixed cycle time).
OUTN gets a narrower pulse (on for, for example, 25% of the fixed cycle time).
Since OUTP stays on longer than OUTN, a voltage difference is created between them (+12.5 V). Current flows through the speaker and moves the diaphragm forward.

3. Negative part of the music wave
When the music wave goes down, the opposite happens, still at the same fixed frequency:

OUTP gets narrower (25% of the time).
OUTN gets wider (75% of the time).
The voltage difference is reversed, the current flows in the opposite direction and pulls the membrane back.

The essence of BD-modulation (Why doesn't it need a filter?)
In classic modulation (AD-modulation), the outputs switch in completely opposite directions (when one is 12.5 V, the other is 0 V), which means that a huge amount of high-frequency noise constantly flows through the speaker, which must be stopped by the LC filter.
In BD-modulation, however, the chip, by changing the pulse width, achieves that the voltage difference between the outputs only appears when it is really necessary for the music [slaa903]. Since in silence or during transitions the pulses on both outputs "cancel out" each other, the speaker itself (with its coil) easily smooths out the remaining small switchings. The fixed frequency of 400 kHz thus serves only as a stable "rhythm" according to which the chip constantly corrects the pulse width according to the shape of the sound.

15/08/2026

AL pro Limited Edition in january 2027. https://www.bolha.com/harmonike/playback-harmonika-al-pro-limited-edition-oglas-12032174

23/07/2026

While preparing through-hole electronic components, the soldering work of SMD components can be done by a hot plate and low temperature solder paste. Since the circuit will not be in a vibration environment, we can use low temperature solder paste for soldering at 130-150°C.

19/07/2026

I have to write that I am still impressed with Tripath technology, DIGITAL POWER PROCESSING (DPP™) TECHNOLOGY. I still use Tripath chips, more than 15 years and I can't see any alternative at the moment. Although the results of other manufacturers show better results, I am not convinced.
I've already tried Maxim small chips, Infineon, IR and TI, but unfortunately they don't sound as good. I'm still using TA2021B for my products and slowly my supply is running low, I'm getting desperate :D. I'm sorry that this is exactly what happened, the big guys pushed Tripath out of the market because someone small can't be as good and they can continue to sell inferior products.
I'm currently using the red pcb because it's really ultimate and works great. I wind the coils by hand on a T60-2 core because it has the best results, very low THD level. One manufacturer has 10uH has excellent coils that I will use in my design later.

TA2021B red board:

The higher-order Delta-Sigma (ΔΣ) modulation in Tripath amplifiers acts as an advanced analog control system. Instead of the classic comparison of sound with a fixed sawtooth signal, Tripath used a dynamic feedback loop.

The process went through four key real-time steps:

1. Delta (Δ) – Error Search (Difference)
At the very beginning of the circuit, two signals meet: the input analog audio signal and the attenuated feedback signal coming directly from the amplifier output. The circuit calculates the difference (delta) between them. This difference represents the error or distortion that the amplifier introduced into the sound.

2. Sigma (Σ) – Error Accumulation (Integration)
This error signal is then sent to a sequence of analog integrators (in Tripath chips these were usually higher-order integrators, e.g. 3rd or 4th order). The integrator acts as an accumulator: it adds up (sigma) and accumulates the error over time.
If the output signal lags behind the input, the value in the integrator increases rapidly.
Higher-order integration allows for more accurate tracking of complex audio signals.

3. Quantization and Switching (Class T Specifics)
When the accumulated error in the integrator reaches a certain threshold, the comparator changes its state. This triggers the switching of the output power transistors (MOSFETs) between the "on" and "off" states.
Unlike classic Class D, where the switching frequency is fixed, in Tripath the frequency is variable (from 200 kHz to 1.5 MHz). The frequency adapts to the speed of change of the music:
For quiet, low tones, the switching is slower.
For demanding, high frequencies, the speed is radically increased to maintain accuracy.

4. Noise Shaping
This is the key trick of the whole process. Since transistor switching generates a lot of high-frequency noise, the higher-level Delta-Sigma modulation pushes this noise far away from the audible range (above 20 kHz) into the inaudible high-frequency regions.
At the end of the amplifier, a simple passive LC filter (coil and capacitor) completely removes this high-frequency noise, leaving only a clean, strong analog audio signal to the speakers.

T60-2 core for output filter:

Why is the "-2" material ideal for Tripath?
The "-2" designation means that it is a core made of iron powder of the No. 2 mixture (red/gray). This material has the following key properties:
-Low magnetic permeability (u_i = 10). Tripath has explicitly advised the use of low-permeability cores in its technical instructions (Application Notes). Since the permeability is only 10, the core behaves extremely linearly.

-Nearly impossible saturation (Saturation): This material cannot saturate brutally and suddenly like classic ferrites. Even at high currents, the coil will retain its inductance, which prevents sound distortion and protects your Tripath chip from destruction.

-Extremely low core losses at high frequencies: The "-2" material is designed to operate in the high frequency range (from 250 kHz to 20 MHz). Since Tripath's variable switching frequency jumps up to 1.5 MHz, this core will not generate too much heat during these switchings. The sound will remain clean and undistorted due to the absence of "magnetic memory" (hysteresis).

Open source project for DiyAudio forum with TA2021B, AMP225T link:

https://audiologs.com/AMP225T/AMP225T_project.zip

some open source projects if you want to do it yourself:

https://www.facebook.com/367076003311013/photos/a.566112266740718/1569587086393226/

https://www.facebook.com/367076003311013/photos/a.566112266740718/1094431307242142

17/07/2026

Dimpi on Ebay 😉.

D-class amplifier 2x25Wrms. There is also an adapter and carrying case. The battery lasts 6-7 hours of continuous playback. -protective/carry case. see photo Manual-Playback-Diatonic-manual. see photo Brochure-Playback-Diatonic-technical.

12/05/2026

In one of the projects, I needed a "clean" voltage to measure current and voltage. For the test, I used cheap Chinese DC-DC modules from 12V to 5V before making my own. I thought I had a measurement error or software error, because the ADC was very unstable and the values ​​were jumping for no reason. It quickly turned out that this was not the case, because the voltage was very "polluted". Then I started checking the DC-DC converters. With the LM2596, it turned out that this was not the chip at all. The elements were for the LM2596, and the inscription on the chip is LM2596, but the switching frequency, which should have been 150kHz for this chip, was only 50kHz and the AC ripple was very large, 300mVpp, due to inappropriate elements for this operating mode. According to the datasheet, I found out that this is the frequency for the LM2576. When I replaced the 33uH inductor with 100uH (due to lower switching frequency) and the capacitors from 100uH to 330uF it works OK. If you use tantalum capacitors, they should be rated at 2x the supply voltage and 100nF, parallel capacitors for peak smoothing. The AC ripple drops a lot, to about 10mVpp, which is acceptable. Now I have my own circuit, I was able to assemble it with the right LM2596 and it works great.
We measure the voltage noise by first measuring the oscilloscope noise itself and then subtracting it from the voltage measurement, the oscilloscope must be set to AC coupling, and the BW is limited to 20Mhz. This time the probe is set to 1x, otherwise we use 10x. Before measuring, we calibrate the probe to 1Khz (a nice rectangular signal, no spikes) the internal oscillator on the oscilloscope, so that it is adjusted for the input.
I also had a surprise with the MP1584EN chip. The datasheet mentions that the current without load consumption is about 100uA, but why does my module consume 26mA in idle mode? When I just looked at the circuit and compared it with the one in the datasheet, I realized that this is not this chip at all. According to the pins, it is aMP1593, or an ACL 4060 or 4070 chip, the difference is the max current 2A for 4060 and 3A for 4070. The chip itself is labeled MP1584EN. This chip is probably very popular and the Chinese just copied old chips with this label. Because it still works at 3A, otherwise it gets very hot even at 2A, because the elements here are not the right values. I also had a circuit with the real MP1584EN, and that works fine. The no-load current is only 0.22mA, this is energy saving for battery power, it has a soft-start, in 1.5 ms the voltage rises to the full set value, the max current is 2.5A, it works constantly at 1.8A, it is only a little warm, and at 3A it turns off because it starts to heat up a lot, just like it is mentioned in the datasheet. AC ripple is only 20-30mV. This chip works great, if only it is the right one :). After replacing the elements, the module with ACL 4060/4070 also works ok, it is necessary to change some values, just like it says in the dasheet and it is stable. In the photo, the circuit with MP1584EN.

28/04/2026

Testing older devices, if they are still within the tolerance range. In the middle we see the old one among the multimeters from the 90s, still working perfectly.

10/04/2026

OWON on duty, warming up the OCXO 10Mhz (1 ppb stability) module to operating temperature for calibration.