Lepai LP-2020A+ mini-review and measurements

Thanks goes out to a friend for donating this amplifier for testing! (You know who you are)

You can hardly enter five forum threads on budget audio without seeing the Lepai LP-2020A+ mentioned. At about 20 U.S. dollars (30 with an AC adapter included), this compact and 20 WPC advertised amplifier looks like a solid deal.

However, despite its popularity, actual information about it seems scarce, aside from the occasional teardown or two. It’s built around a Tripath TA2020-020 class D amplifier chip, which, with its fine specifications and almost 90% efficiency, is what allows for this fantastic power to size ratio.

The specifications printed on the box (that’s apparently used for two models)

However, class D amplifiers require far more attention to detail in the circuit board design and component selection than traditional class AB chip-amps (like the classic LM3886). Did Lepai get it right? Does the LP-2020A+ live up to the specifications quoted on the box, or perhaps even the ones quoted in the TA2020-020 datasheet?

Let’s find out! (The less technically minded may want to skip to the verdict)

The test set-up

The test set-up

The LP-2020A+ was powered by a 5 A industrial switchmode power supply. THD+N measurement was done with an HP 339A, low-passed at 30 kHz. A 4/8 Ohm switchable resistive load was used.

Output impedance/damping factor was calculated by setting a loaded voltage of 5,51 V (8 Ohm) and 5,2 V (4 Ohm) and observing the voltage change when the load was disconnected. A VCC of 13,5 V was used for all tests, 13,4 V was the lowest voltage observed.

The results (updated 9.12.2012):


A-weighted spectrum captured at 5 W into 8 Ohm.

More spectrums (HF noise is included in the unweighted ones):

Sadly, but perhaps not surprisingly, the Lepai did not fulfil the claimed 20 W/ch at <0,05 % THD.

It starts distorting after about 9,6 W into 4 Ohm and 5,6 W into 8 Ohm. It doesn’t show on the THD+N measurements, but the clipping waveform during mild clipping is unorthodox and in my opinion rather intrusive. It causes a popping sound in the tweeters of my test speakers.

It does however far surpass its specification of 80 dB SNR, measuring in at 90,1 dB A-weighted and 87 dB unweighed from 20 Hz to 20 kHz.

Absolute noise measures in at roughly -76 dBV(a). This is 10 dB higher than higher-end amplifiers I’ve measured, and can be audible on many speakers – it is very obvious on mine. If noise above the audible spectrum is included, the numbers falls considerably; if measured unweighted from 10 Hz through 30 kHz on my HP 339A, the noise floor is at roughly -66 dBV. For comparison, that’s nearly two orders of magnitude above the -84 dBV of the Luxman L-120A that resides in my system.

The LP-2020A+ exhibits this behaviour during mild clipping. Picture taken at 0,3 % THD+N.

There is also a constant 200 mV P-P noise at about 160 kHz. Even with the low-pass filters in place, this somehow caused the 339A to go haywire when measuring at very low levels (<250 mW). This noise is probably not an issue in everyday use.

There is a constant noise present at about 160 kHz.

Scope view of a 20 kHz sine wave at 5 W into 8 Ohm. The HF noise can be seen as the abnormal thickness of the trace.

A more noteworthy detail about the LP-2020A+ is that it seems to have a considerable notch in the frequency response above 10 kHz, even with the tone controls disabled. After investigating the issue, I was able to conclude that the notch was not present at the input of the TA2020-020 amplifier chip.

Captured at the output, this curve shows that the frequency response peaks at about +1,6 dB at 30 kHz, and is up by over 1 dB at 20 kHz.

I’m not certain about what causes this anomaly, but I’d wager that it has to do with the way Lepai have designed the output filter. This could very well be an audible problem with this amplifier.

The high-frequency anomaly aside, the LP-2020A+ seems to be flat throughout the frequency spectrum down to about 40 Hz, rolling off by 1 dB at 17 Hz.

Build quality

The LP-2020A+ freed of its case

Build quality wise, you do get what you pay for – the design has a fair few cut corners that deviate from the design guidelines in the TA2020-020 datasheet, in order to reduce component count. The electrolytic capacitors are of varying Chinese brands not renowned for their quality, the potentiometers are pretty nasty and power supply filtering is … minimalistic. All in all, it’s still better than I expected. Soldering quality is decent and the aluminium case is solid.

However, something that struck me about my unit was this:

My LP-2020A+ had very poor heat sink mounting.

Thermal paste was scarce.

Yes, there was an almost millimetre-thick gap between the chip and the heat sink! The over-temperature protection would kick in after a few minutes of heavy 4-Ohm loading. Thanks to the efficiency of the TA2020-020, this would probably not have been noticeable in every-day use. However, one must wonder what it does to the chip lifespan.


I entered this with a hunch that the Lepai LP-2020A+ would not deliver its advertised performance. The TA2020-020 chip that it’s built around is rated for 20 W into 4 Ohm at 5 % THD. Despite that, it is still a very decently performing amplifier for the price as long as you stay within its limits. 5 clean Watts per channel into 8 Ohm might not sound like a lot, but for casual near-field listening with small bookshelf speakers it should still be plenty; “normal” background listening level is generally only around 0,1 Watt!

The biggest issue with this unit is the build quality. The casing is good quality extruded aluminium, but the quality control and above all the potentiometers (volume control, bass and treble) are of atrocious quality. The unevenness and scratching of the volume control can be fairly bothersome, especially when close to zero. Since the volume control is connected to adjust the gain of an op-amp rather than simply attenuate the input signal, a bad connection in it can lead to very loud popping and oscillation.

Sound quality wise, there isn’t much to comment on. There is a slight notch in the treble that could colour the sound, but it is a minor issue in this market segment. Other than that LP-2020A+ has a background hiss that’s 10 dB (twice as loud) higher than the considerably more costly amplifier I usually have in my system. Depending on what speakers are used and how sensitive one is to background hiss, this could be an issue, and probably the biggest sound quality concern with the Lepai.

All in all, would I recommend it? The answer is a resounding¬†yes! While it’s hardly “audiophile quality”, you aren’t going to find a better 5-watt-per-channel¬†amplifier for the price – and with proper stereo amplifiers often costing several hundred Dollars, this little Lepai introduces a whole new standard for budget amplification.

Image gallery




Preamp and tone amp op-amps (JRC 4558)

Output low-pass filter. The blue capacitors are 47 nF.

The Tripath TA2020-020 amplifier chip

Solder side


Distortion? Let’s get analogue: HP 339A in the house!

HP 339A distortion measurement set

Through a stroke of luck and friends in the radio broadcasting business (Thanks, Mathias!), I’ve found myself with this classic bit of kit on my bench.

Basic specifications for the HP 339A

While the 339A can’t quite match the performance of a high-end computer sound card, it’s a very competent piece of measuring equipment for evaluating audio frequency power amplifiers and other things that you don’t really want to connect to your computer – at least for us poor DIYers who can’t afford proper audio scopes.

The bitter-sweet part about this particular unit is that it’s barely been used at all, and that it did come with a calibration report – from 1988. Going through the performance checking procedures in the manual and cross-referencing measurements with my computer (equipped with an ASUS Xonar DX sound card and Vitrins Multi-Instrument 3.2) indicates that it’s in decent shape. Thankfully there aren’t many “time-sensitive” components in the signal chain at all, so I still feel fairly secure in using it. However, I can’t fully trust it to be absolutely accurate without at least a check-up by a proper cal lab.

A 1.45 V, 1 kHz sine wave generated by the 339A’s internal oscillator. Click for a slightly larger picture.

Connecting the internal oscillator to my computer shows that the internal oscillator is within both the frequency (+/-2 %) and THD (<0.0018 %) specs. In fact, it performs so well that the reasonably competent ADC on my sound card can’t even resolve its noise; the measurement above roughly equals the specified noise figures for the ADC chip itself – even disregarding the output buffer.

The true RMS output voltage of a Xonar DX sound card. The monitor output doubles as a fantastic label holder.

Aside from its signal generator and distortion analyser functions, the 339A incorporates a true RMS voltmeter. This is a very useful feature, since it allows the operator to quickly measure both the signal generator’s output level as well as the input level, without the need for separate tools or messing about with the connections.

The HP 339A makes it easy to switch between measuring distortion, input and output levels

Output from monitor connector at clipping of the input

It furthermore offers a monitor output for connecting a spectrum analyser or an oscilloscope. The monitor output is essentially hooked up in parallel with the meter, i.e, the fundamental tone is heavily suppressed (by -100 dB) and the distortion products come through untouched at 1 V full-scale. A very useful feature for poor sods such as I, since it offers a convenient “computer adapter” for heavier-duty equipment.

For the uninitiated, a distortion analyser like this works by suppressing the fundamental frequency (say, 1 kHz) with a very sharp notch filter and then measuring the amplitude of the harmonics at 2 kHz, 3 kHz, 4 kHz and so forth as well as noise produced by the equipment being tested. The level of the harmonics and noise is then presented on the meter in dB-below-the-fundamental and %. The monitor output essentially provides a distortion-only output – and the distortion is what we’re after.

To finish this up, some pictures of the guts of this thing:

The top cover only reveals the bottom of the boards – very unsatisfying.

Flipping it over reveals a maze of aluminium. HP went all out with the shielding in this one.

Sadly, I couldn’t find a clear map over which “sections” do what in the manual.

The power supply section beautifully displays the gold-plated circuit boards.