I did not rule out PSSR, nor crosstalk.
I am referring to
I am referring to
and went on to summarise with
Both these issues affect amplifier performance and these two issues do not have anything to do with either PSSR or crosstalk.
"Handling of the grounds" is one of the little known but vital issues that Tom must have addressed to get his 4ppm THD.
Though I haven't seen his PCB layout, his wiring instructions or his circuit, I would be very surprised if he hasn't addressed EXACTLY your concerns in the stereo version.
I look forward to his measurements of same.
Please! 3.4 ppm THD+N. 1.3 ppm THD... 😀
Yeah, you don't get there without understanding the system. That's for sure.
~Tom
Hi Tom, may you comment on how you get the THD figures better than the test equipement specification itself ? seems the AP525 is given for 1.4uv noise floor and 107db thd which is 4ppm. is this about averaging or fft size or bandwith ? Also I m wondering if there is a trick that can be used to reduce the thd at low volumes, as the curves show that the thd at full power is better than the one at low power ? Cheers
Hi,
I have 300VA 2x18V and 300VA 2x25V transformers.
Which one is more preferable for Modulus-86?
Thanks.
I have 300VA 2x18V and 300VA 2x25V transformers.
Which one is more preferable for Modulus-86?
Thanks.
2 x18v will give you 20-25 watts max. Plenty of power for easy load passive speakers 88dB 1W (and above) in typical UK/European homes. If you're not planning a disco this will be fine. This will give you about 1 dB less than the recommended 22v secondaries by my reckoning (looking at Tom's own figures).
Unit can be smaller, lower power consumption and heatsinks can easily be accommodated in a small internal structure. Cooler running and very little difference in maximum volume even compared to 25v rails.
The 2x25v will give you more power than Tom's specification (but within spec for LM3886) and push the output devices harder, so spec for bigger heatsinks and good ventilation. The subjective volume difference will be barely audible.
I have just started on a build (for my Dad's passive speakers) and have a transformer with 18v secondaries. His speakers are 94dB/1w sensitivity.
So... There is THD and THD+N. THD is the ratio of the powers of the harmonics to the power of the fundamental. I measured the THD from the FFT plots and calculated the THD from the magnitude of the harmonics relative to the fundamental. No tricks employed there. I ran with a 512k FFT length and averaging of 32 to get sufficient frequency and amplitude resolution.
For the THD+N, if a harmonic falls below the noise floor, its amplitude is assumed to be that of the noise floor at the harmonic frequency. This leads to the -20 dB/dec slope in the THD+N vs output power plot.
The best case THD+N residual of the APx525 is about -112 dB THD+N. The worst case is a few dB worse. The absolute THD+N limit depends on which input voltage range is chosen and the amplitude of the incoming signal. It's basically an SNR game. Higher signal -> higher SNR -> better THD+N -- until the AP changes range. So the key here is to prevent the range changes.
The sweet spot of the instrument is at an RMS input amplitude of 8-9 V RMS. By using a voltage divider on the output of the amplifier under test, it can be ensured that the input signal to the AP stays within the sweet spot of the instrument. Of course, one has to choose a resistive divider that does not add appreciable THD or noise by itself. Thankfully, a pair of 1 kΩ metal film resistors do the job nicely.
The other knob to turn is the noise bandwidth. Lowering the analyzer noise makes it possible to measure very low levels of THD+N. For measurements at 1 kHz (FFT and THD+N vs output power), I use a 22 kHz measurement bandwidth. For the THD vs frequency, I use a 60 kHz bandwidth to ensure that the third harmonic of the 20 kHz fundamental is captured.
These are the workarounds that I employ to make sure I measure the performance of the Modulus-86 and not simply that of the APx525. The APx525 is the second-best audio analyzer in the world. Second only to the $40k APx555... The Modulus-86 pushes the instrument to its limits, and in some cases a bit beyond.
~Tom
2x25 VAC will give you about ±34 V. That's well suited for driving 8 Ω speakers, but with 4 Ω speakers, you will likely hit the current limit for the LM3886 near the rails.
2x18 VAC will provide about ±24 V. That will work well for both 8 Ω and 4 Ω speakers.
Personally, I'd go with the 2x18 V AC transformer. You'll get slightly less power than with ±28 V rails, but you'll also get less dissipated power --> smaller heat sinks.
±24 V ~ ±28 V is the sweet spot for the LM3886 in my opinion.
~Tom
Tom We appreciate if you can desing two-channel amp more than 80W 8 ohm per channel. 35w is really not enough even for shelf speakers (100-150w)
Really? How much power do you actually need? I did the math on my setup and concluded I needed 3.24 W per channel to reach an SPL of 90 dB at the listening position. I go through the math on my Taming the LM3886 website.
At 38 W, my 87 dB/W*m efficient bookshelf speakers will reach 101 dB SPL at the listening position.
As previously noted, I am working on a Parallel-86 board. This board will allow for operation at +/-35 V into 4 ohm loads. That's enough for well over 100 W of output power. This board will also be perfect for bridging, should you need even more power.
~Tom
Do this test and find out how much YOU need with YOUR speakers on YOUR music.
test-how-much-voltage-power-do-your-speakers-need
Modern music (??!) which has been compressed to have a dynamic range of 6dB or less doesn't need much power to produce ear bleeding levels. 😱
But I just played one of the most famous recordings of all time, the 1965 Barbirolli, du Pre Elgar Cello Concerto. For the first time this Millenium, I regret no longer having 2x200W 8R 🙁
I am wondering whether you considered looking at the impedance curves of your speaker drivers. Normall Fo is smack in the critical areas where the music energy is high, thus limiting the current going through it.
If that was an issue, you'd see extreme dips in the frequency response of speaker drivers, not to mention multi-way speakers. Note that multi-way speakers typically have impedance peaks near the crossover frequencies. For a 3-way or 4-way speakers, you can have impedance peaks all through the audio range, even though the speaker will produce a constant SPL vs frequency for a constant input voltage.
I have measured the SPL output of various speakers at a constant input voltage. The SPLs have been flat within the limits of the drivers.
What happens is that near the resonance of the driver less B*I is required to move the diaphragm a certain amount. So even though the driver draws lower current, the diaphragm still moves the same amount and the driver produces the same SPL as away from the resonance.
~Tom
Last edited:
The problem is that at resonance, that same issue causes the driver to not be so well controlled, so there are phase issues as well. Looking at SPL alone can be very misleading because some issues occur at more than 30db below which does not show up much on the SPL plot.