The LDA low distortion power amplifier project
Welcome to this page, here you'll find a little information about the LDA low distortion power amplifier project and the development process.
After studying books on low distortion amplifier design and months of learning in the first half of 2018, I decided to have a shot at designing a linear, transistor power amplifier with output transistors biased in the low distortion region between class B and class AB. I also designed the associated circuitry: mains/line input and distribution, power supply, and protection circuits. And this became the LDA project, an amplifier that is suitable for the most demanding home audio requirements and is still cheap enough, so an average hi-fi enthusiast or hobbyist can probably afford it.
I haven't published the circuit schematics so far, but I may do so sometime in the future. There is no special invention in my amplifier circuit, I studied books mainly from Douglas Self and Robert Cordell, so the amplifier circuit is based on the concept that they set forth in their books and what Douglas Self called a blameless amplifier, except that this one is capable of higher output power.
The LDA162 (and a later slightly modified version, LDA172) power amplifier is capable of delivering 200 Watts into a 4 ohm load with very low distortion (the 1kHz THD of the first and second prototype measured less than 0.0007% @ 2.5 Watts into 3.5 ohms, and degraded to only 0.0013% @ 200 Watts into 3.5 ohms). Moreover the amplifier has very low THD at all power levels below 200 Watts and its distortion remains considerably low up to the highest audible frequencies. My intention was to design it to be robust and reliable.
If you want lower amplifier power, that is viable without component changes, just by lowering the supply voltage. At 110 Watts power and below, you may use only 1 output transistor pair instead of 2 pairs, which makes the component set somewhat cheaper.
The original and most expensive configuration (LDA 53DM) achieves protection via an ultra low distortion solid state relay, which is faster than a mechanical relay for overcurrent protection purposes, and in terms of distortion it is more stable (practically unchanging) unlike a mechanical relay. Later I designed two cheaper configurations with mechanical relays (LDA 43S, LDA MIN), although I don't like using mechanical relays in amplifiers, especially in low distortion amplifiers.
In November 2018, I built a mono,
then later a stereo prototype.
The distortion measurements revealed an extremely low THD at 1kHz.
Other, very challenging kind of distortions, like 19+20kHz Intermodulation (IMD) or Transient Intermodulation (DIM) are also very low.
This is the first prototype of the protection circuitry.
The PSSR350 power supply & relay board includes an ultra low distortion solid state relay, which has several advantages over a mechanical relay.
In January 2019, I designed the second version of the amplifier PCB, leaving out some unnecessary complexity, and built another stereo prototype with this new PCB layout. As you can see, this is a medium power version, with only one output transistor pair installed.
Measurement results are a bit worse at heavy loads, but I suspect it would be about the same if I had used 2 output transistor pairs. I also designed a new power supply board (PSMR101S) with mechanical relays for those who would like to go cheaper, because the ultra low distortion solid state relay is much more expensive than a mechanical relay.
In February 2019, I conducted an ABX style listening test with three young people, two of them were musicians. None of them could distinguish the LDA162 from a commercial amplifier (namely a Sony STR-DE598). The Sony has about 10 times higher harmonic and intermodulation distortion in the output voltage range at which we listened to music. The test music was (1) the first 30 second of Mike Oldfield's Five Miles Out, and (2) the first minute of 2CELLOS' I Will Wait. The official recordings were played through an EMU 1212M sound card. The loudspeaker was a pair of PM200's that I built back in 2003.
In May 2019, I built another stereo prototype in the lastly developed, cheapest configuration (called LDA MIN), which has only a simple protection circuit (loudspeaker protection and turn-on thump muting through a DPST mechanical relay). It measured just as its predecessors in terms of noise and distortion.
In June 2019, I made two little modification to the LDA162 circuit for somewhat better temperature stability and robustness. The temperature drift of the current source of the voltage amplifier stage became minimal, and the voltage amplifier transistor is now protected under output overcurrent or output short circuit conditions. I built a mono LDA172 prototype, which produced the same excellent measurement results as the LDA162.
In May of 2019, I started to sell a few LDA kits without any profit in Hungary. Since there was almost no interest at all, I later wrote the assembly instructions in English, and decided to sell some more non-profit kits on ebay.de. If you are in some other European country (not in Germany or Hungary), and you'd like to build this design and are willing to transfer the non-profit price of the kit plus the postage fee in advance to my Paypal account, then I will send the kit to you. More information about the non-profit kits here.
For those who are not persuaded by my ABX test, in the middle of August 2019, I conducted a much more revealing test, a so-called nulltest that again proved that the LDA amplifiers don't have an audible distortion. Please google "nulltest" if you don't know what a nulltest is. This test was done on one of the LDA162hc prototypes. The used sound file was a short section (approx. 1m35s - 1m45s) from the official recording of 2CELLOS' Thunderstuck played via an EMU 1212M sound card. I chose this recording, because there are artificially distorted cello sounds and beat sounds in it with a strong harmonic and transient content, and it can be thought (at least by mere human reasoning) that it is "harder" for a power amplifier to amplify distorted sounds without adding even more distortion to them. During the test, my 8-ohm PM200 loudspeakers were connected to the output of the amplifier, and the volume was set to what can be called "normal listening volume", about 75-80dB.
Here's the difference file that was generated with Diffmaker. In the beginning of the file there is a few tenths of seconds remnant (not subtracted) music, you can use this short section to set the volume on your amplifier to your preferred (normal) listening volume. Then sharpen your ears: can you hear anything in the following approx. 10 seconds (without pumping up the volume)? What you hear is the difference between the input (= above said music) fed to the amplifier and the output of the amplifier, while the said loudspeaker was also connected. I couldn't hear anything at 1.5m listening distance from my loudspeakers, although there was a little background noise from traffic. I also played around with Audacity to amplify this difference .wav file by 70dB (3162 times amplification) , so I may hear what is the difference. What I found in it is mostly the hum noise of the transformer and some weak music perhaps due to the imperfect gain matching of the input and output files when they were subtracted as the subtraction couldn't be done properly due to the little hum content in the amplifier output signal. The transformer in this LDA162hc prototype was a normal EI core, unshielded transformer (not a toroidal one), I think this caused the little hum. I don't upload the 70dB amplified difference file here, if you're interested, you can amplify it yourself by using the free Audacity application.
This test should be persuasive to everyone that the LDA amplifiers don't have an audible sound signature of their own. They are transparent.