Trying to make things measure better under resistive loads vs reactive loads are two different things. Just change the load to a simplistic drive load model and you will find that out quickly if you pick a frequency in the highly reactive range.
Why not? Of course, beyond a certain point, it's pointless to waste time and/or money pursuing better numbers.
There is no fixed number. Which number is dominating in a system determines what part might need improvement. For example, if the Modulus 86 and the parallel version do not sound the same in the same system, then you start to figure out what is causing the difference, and what needs further improvement.
Member
Joined 2006
One of my collaborators can reliably detect THD down to about 0.3 % (-50 dB). By 0.1 % (-60 dB) it's inaudible. That's one data point anyway.
At some point the hunt for zeros becomes academic. In case of the Modulus-86 and Parallel-86, the THD is a nice selling point. However, the real benefit of the composite amplifier is that it has very high loop gain. I find that 40-50 dB of loop gain at 20 kHz is needed for an amplifier to sound clean. Very few amps can do this. Bruno Putzeys' amps are the only ones I can think of at the moment. The high loop gain means that power supply noise and inter-channel coupling is dramatically reduced, thus, cleaning up the sound.
It all depends on your design objective. Do you want a straight wire with gain or do you want an effects box? With the MOD/PAR-86, I went after the straight wire with gain.
Tom
At some point the hunt for zeros becomes academic. In case of the Modulus-86 and Parallel-86, the THD is a nice selling point. However, the real benefit of the composite amplifier is that it has very high loop gain. I find that 40-50 dB of loop gain at 20 kHz is needed for an amplifier to sound clean. Very few amps can do this. Bruno Putzeys' amps are the only ones I can think of at the moment. The high loop gain means that power supply noise and inter-channel coupling is dramatically reduced, thus, cleaning up the sound.
It all depends on your design objective. Do you want a straight wire with gain or do you want an effects box? With the MOD/PAR-86, I went after the straight wire with gain.
Tom
May I be permitted to guess and say that the test was for low harmonics of the test signal/s.
If crossover distortion were to be "tested", then one finds that the detectable levels appear to be much lower compared to low harmonic THD.
Hi Tom.
Now that you have such 'blameless' amplifiers, are you using them in anger yet ?...ie are you running a decent system now and forming long term experience of your modules ?.
Dan.
Listening tests are really complicated when we get into audibility of distortion, the speaker residual release of energy can mask amplifier distortion as well as room reflections and modes.
Higher amplifier gain makes the sound more clean because interaction from the speaker has reduced interaction with the amplifier, I was just reading about it again a few days ago.
Different wires influence sound in different ways, pure thick straight wire really introduces some skin effect. I actually had to find the appropriate wire for speakers to keep negative effects to minimum. Selection was based on listening, testing, and the selection of configurations were based on engineering judgement prior to the tests because each cable factory had there own approach, and it was not easy to make a sound selection unless many factors are considered. Hafler even has a tuning procedure to achieve what they think will be closer to a straight wire when you apply speaker load. This just tells me that speaker load needs to be brought into the straight wire performance measurements.
That's likely true. Crossover distortion is pretty easy to minimize, though. Just add loop gain...
In my case, the test signal was actually music of my collaborator's choice. The tests were performed while I was working on the DG300B. I went through several iterations on the driver circuit, each having different THD. The DG300B is single ended class A with no global negative feedback...
We've performed the same tests on my solid state designs. Years ago, I built an LME49811+STD03 amp, which, due to sub-optimal layout has higher THD above 1 kHz. This has been compared against the MOD86.
It's relatively easy to get very low THD in a headphone setup, so one can compare the clean headphone sound with the amp+speaker sound as well.
Tom
Last edited:
Could you explain this please? As far as i know, high amp gain means low loop gain means higher Z_out means more "interaction from the speaker" with the amp.
😕
Ben
That's interesting, but may not be realistic of most distortions, which generally create more than one harmonic at a time, even if one is dominant (such as 2nd in a Class A circuit, especially tube or FET).
I listened to both the 24bit (in Chrome) and the 16bit (in Firefox) files, and the 5% second sounds just like the pure sinewave to me. As I CAN hear the difference with the 1% second harmonic, I presume the 5% is (mistakenly) the same as the the pure sine.
The 2nd and 3rd have different "character," but by the 4th I can hear the harmonic as a separate tone as much as, or even instead of a different "character" of the fundamental, and higher ones are also heard as separate tones, thus making this rather useless as examples of distortion (other than the 2nd and 3rd).
I agree with AndrewT that crossover distortion, which generates an array (dare I say a "full spectrum") of high-order harmonics, is a better test of audibility for the spectrum of low levels of THD that a bipolar transistor amp might generate. For a sine wave one hears a tone with a "character" rather than a fundamental and harmonic separately. With a multitone or musical signal, there's lots of 'grit' generated from both harmonics and IMD by crossover distortion.
I'm using my amps with pleasure in a stereo setup. I've used my amps on my Dali 3A speakers as well as a pair of Alpair 6P-based full-range speakers. I like them. My opinion of the amps haven't changed since I first plugged them in. They're still incredibly good amps.
Considering that you haven't defined what a "decent system" is to you, that's the closest I can come to answering your question.
Tom
I am referring to "amplifier gain" is the last stage closed loop gain which output is connected to the speakers. Sorry for the confusion.
That doesn't really change my confusion, as this still means lower loop gain which leads to higher output impedance.
The "amplifier gain" as you define it will be a little less than 0dB in most designs, as the last stage often is a voltage follower.
Raising SNR:
The linear regulators on the MOD86 need to drop quite some voltage as they are fed directly from the HV rail (How much depends on the transformer secondary voltage of course). Is there some headroom in respect to power handling / thermal design if i wanted to use the MOD86's +-15V rails for additional circuitry?
Ben
Worst case current draw through the LM317/337s is 15 mA. At ±28 V rails, this results in just shy of 200 mW of dissipation, so you have some room in the budget for additional circuitry.
Tom
Be aware that those rails are potentially noisy (in the context of the normal output noise of low noise regulators), carrying noise which is strongly signal-correlated.