The STA326 was 24-bit DDX® 32 kHz to 192 kHz input sample rates, 96 kHz internal processing sample rate, 24 to 28-bit precision, ternary PWM output, [master gain/attenuation + dynamics control attack/release + auto volume-controlled loudness], digital gain/attenuation (+48dB to -80dB)... but most important (for me till now) was the digital I2S input (monochip true FDA, from digital inputs to PWM output).
My brother has a benchmark ahb2 + Gustard X16 DAC : My super tweaked STA326 is (way) more noisy, less detailed, but as balanced and nearly as pleasing to listen to (and more pleasing to listen to at low volume).
Nevertheless, i'll try the A8... (have to find a good inexpensive DAC with IR remote control).
Last edited:
STA326 have terribel noise.
https://www.audiosciencereview.com/...s/arylic-a50-review-wireless-amplifier.19944/
https://www.audiosciencereview.com/...s/arylic-a50-review-wireless-amplifier.19944/
The biggest problem with the STA326-era devices is that 24 bits of precision for the coefficients is inadequate at low frequencies. When you do the bilinear transform to generate the coefficients for the biquads, you end up with very small numbers for low pass filters (such as DC blocking or "rumble" filters). Most of those 24 bits are placeholder zeros after the decimal point, and there are just a few significant digits, resulting in quantization errors. As a result, the older Apogee/ST devices aren't going to measure as well, even though the chip can sound very good if used within its limitations. A lot depends on how the filters are configured during measurement.
ST addressed this issue starting with the STA311 chip, which adds another stage to the biquads to provide additional filtering. You can see that extra filter stage and ST's explanation on page 69 of the STA311 datasheet: https://www.st.com/resource/en/datasheet/sta311b.pdf. However, a better way to address this is to use more bits for the coefficients. For example, the ADAU1701 uses 28 bits and the ADAU146X series uses 32 bits. I can't see how many bits are used for coefficients or calculations in the MA2304DNS from the data posted on the Infineon web site.
ST addressed this issue starting with the STA311 chip, which adds another stage to the biquads to provide additional filtering. You can see that extra filter stage and ST's explanation on page 69 of the STA311 datasheet: https://www.st.com/resource/en/datasheet/sta311b.pdf. However, a better way to address this is to use more bits for the coefficients. For example, the ADAU1701 uses 28 bits and the ADAU146X series uses 32 bits. I can't see how many bits are used for coefficients or calculations in the MA2304DNS from the data posted on the Infineon web site.
Never liked that FX-AUDIO I had with STA326 even if it' sounded pretty good then.
It had som mute function that cutted in when I listened to solo piano at very low volume.
But no where close to sound like MA12070.
Time has passed since so better forget it.
ADAU1701 has an dynamic range of 104dB.
ADAU146x 139dB.
Easy choise.
It had som mute function that cutted in when I listened to solo piano at very low volume.
But no where close to sound like MA12070.
Time has passed since so better forget it.
ADAU1701 has an dynamic range of 104dB.
ADAU146x 139dB.
Easy choise.
A while back I bought an UD20 all in one solution baset on the STA335BW which is an I2S Amp. It was meant as a carry along to be fed by 12V SLA.
The integrated USB receiver uses a Savitech chip. To break it in I watched movies via my Desktop for some months. It did sound okay but nothing special, until I had the crazy idea to hook it up to my 280AH Lifepo4 battery bank paralleled by 60uf of MKP caps. That was a transformation that made me speechless.
I'm not saying it rivals my main rig, but I will certainly explore the available I2S Amps further, as well as trying out massive Lifepo4 batteries on all my amps.
The unit is still in stock form which gives me hope for further improvement (ext. Clock, separate USB supply, dedicated regs...)
I tried to feet I2S to it via my Dante system but was unsuccessful. Some logic programs the chip on the fly via I2C which I have no clue of.
Greets Klaus
The integrated USB receiver uses a Savitech chip. To break it in I watched movies via my Desktop for some months. It did sound okay but nothing special, until I had the crazy idea to hook it up to my 280AH Lifepo4 battery bank paralleled by 60uf of MKP caps. That was a transformation that made me speechless.
I'm not saying it rivals my main rig, but I will certainly explore the available I2S Amps further, as well as trying out massive Lifepo4 batteries on all my amps.
The unit is still in stock form which gives me hope for further improvement (ext. Clock, separate USB supply, dedicated regs...)
I tried to feet I2S to it via my Dante system but was unsuccessful. Some logic programs the chip on the fly via I2C which I have no clue of.
Greets Klaus
Windows Server2016 fully loaded into RAM
Diy Dante System with Monism modules giving out I2S
to AK 4490, 4493, 4495 as desired
transformer coupling skipping all active stages on dacs and amps going straight to the MA12070, TPA3255, TPA3118 input pins. Differential no caps in the signal path.
Many different Speakers Fullrange, Horns, Monitors....
Diy Dante System with Monism modules giving out I2S
to AK 4490, 4493, 4495 as desired
transformer coupling skipping all active stages on dacs and amps going straight to the MA12070, TPA3255, TPA3118 input pins. Differential no caps in the signal path.
Many different Speakers Fullrange, Horns, Monitors....
With relatively high efficiency loudspeakers one rarely plays at higher levels than 1W (full range that is). Open for debate of course but I never needed more powerful amplifiers than 10W ones. One better has some current instead of voltage 
The drawbacks of high power amplifiers are too many to justify their perceived necessity in many cases. Not all but many.
The drawbacks of high power amplifiers are too many to justify their perceived necessity in many cases. Not all but many.
Last edited:
I looked at 2.1 implementation for something like the Paul Carmody "Isetta". MA12070 with 8 ohm speakers in BTL and from 24v supply can deliver 2x12 or 2x13 watt before clipping. SE power is 1/4 of that. So for Isetta I would have 2x3watt and 1x30watt (BTL 4 ohm before clipping, the chip likes lower ohms/heavier loads so it seems).
