Give me a break. The listening test uses mp3 files!
<param name="initParams" value="soundAName=linearexample.mp3,soundBName=12dBexample.mp3" />
What do mp3 encoders do? They discard unneeded details. Like distortion, for example. And the bitrate of the mp3 files? See attachment below.
To seriously listen for low level distortion, try 24/96.
Attachments
Last edited:
If your 0.1% is second and/or third harmonic, then I agree you probably can't tell the difference between 0.01% and 0.1% THD of low harmonics.
But it is completely different when there is crossover distortion.
0.1% of crossover distortion with lots of odd harmonics and most of them above the 5th, it is very easy to dislike 0.1% of crossover distortion and many will even dislike 0.01% of crossover distortion compared to 0.01% of low harmonic distortion.
Last edited:
Yet some people can hear the low levels of distortion. The test asks the same question several ways so it can't be blind chance.
MP3 can use up to 32 frequency banks to define a sound stream, the content being compressed and lost is not always the distortion because the distortion is information, not blanks/zeros/repeated numbers etc. But if you don't believe it, don't. Klippel is no dum dum of a company.
Last edited:
I totally agree that the type of harmonic distortion is important, probably more so than just the absolute THD number.
Here is the FFT of an Aksa Lender driving 2.0Vpp into 50ohms (title incorrectly says 1vpp):
In the above graph, note that distortion is predominantly 2nd and third order with third lower than second - a pleasant sounding profile despite it being 3.5x higher THD than another amp (shown below) that is DC coupled and uses push pull and current feedback topology with only 0.002% THD, but the harmonic content is over higher orders and third is higher than second. IMO, doesn't sound as good even though lower in THD:
Last edited:
Suzyj,
Very nice compact design. Those are some very nice THD figures you have and I would love to see the harmonic distortion profile to see where the H2/H3/H4/H5 etc distribution is. Can I ask why you did not use SOT23 for the TO92's and perhaps SOT223 for the TO247's? It would have been a slick flat SMT package then. You could even make it on metal core PCB and bolt the whole thing to a heatsink for higher power operation.
Very nice compact design. Those are some very nice THD figures you have and I would love to see the harmonic distortion profile to see where the H2/H3/H4/H5 etc distribution is. Can I ask why you did not use SOT23 for the TO92's and perhaps SOT223 for the TO247's? It would have been a slick flat SMT package then. You could even make it on metal core PCB and bolt the whole thing to a heatsink for higher power operation.
The files may sound a little different, but I don't see how the amount of distortion can remain calibrated or known after conversion to mp3, which is a lossy type of compression. Especially, low bitrate 127kbps which requires discarding a significant amount of information to reduce file size. The best that can be done without information loss may be Monkey's Audio compression which is lossless and can compress audio up to about 50%.
Also, I don't think Klippel is a dummy, but his website designer or somebody over there might be.
Attached incredibly early not well calibrated dodgy (did I mention they're early) measurements of 1KHz THD into a 47 ohm load with no heatsink.
I'm badly noise limited, which is making the measurements harder than they could be. Also I'm finding that unless I run my oscillator at the point where I'm about to run out of loop gain, the oscillator harmonics dominate.
Not that I'm complaining, mind. Hard to measure is my goal
Attached piccie of my spreadsheet, plus plot of 2.5V RMS into my spec-an, both with and without 1KHz notch, which knocks my fundamental down ~50dB.
If I reduce loading on the amp (increase load resistor from 47 Ohms to 1K) the harmonics drop into the noise floor. I'm thinking (given that it's predominantly H2) that I'm suffering due to the output transistors not being in thermal contact with one another.
So it needs a heatsink, or at least some metalwork to ensure the output transistors are at the same temp.
Like I said, early, dodgy measurements. I'm waiting on boards for a new source that will hopefully make this process easier.
I've gone off SOT23s a bit of late, as I find they don't save as much PCB real estate as I'd like. TO-92's are quick and easy to deal with, plus they include a via for every pin.
My goals for this amp are something that's easy to build as well as performing well, as I'll probably be making a fair number for line arrays and such forth.
Attachments
Last edited:
I've been running a pair of SB12PFC-4 (12cm 4 ohm mid-woofer in my noiseUnit speakers) for a few days playing music in my study with these, with no heatsink, to see how they fare. Haven't blown them up yet, despite the occasional complaint about the volume from my husband.
I'm wondering if perhaps given that transistors are cheaper than heatsinks, the solution is simply to have many output drivers and no heatsink.
Last edited:
If net efficiency stays the same, then you should produce the same amount of heat with more transistors. Where would the heat go without a heat sink?
Output drive current capability for this board is essentially down to output transistor gain, number of output stages and quiescent current per output stage. If I double the number of output stages I can halve the quiescent current per stage for the same output current capability. The heat associated with that quiescent current is now dissipated across twice as many devices
overall heat dissipation under load is independent of number of transistors, so by increasing the number of transistors you're decreasing the heat each one needs to dissipate.
Essentially more transistors = more heatsinking area - the transistors themselves are little heatsinks.
Transistor plastic cases probably don't conduct heat as well as aluminum fins, or have as much thermal mass to possibly help handle transient heat loads. Also, they would need to have sufficient airflow similar to fins. With a heat sink it could be exposed on the back of a speaker cabinet, say, something people sometimes do.
Last edited:
Yes, very good point. I think the rationalle behind using a heatsink vs using more transistors depends on what overall power level we're wanting to do. At 5W it's unclear as to what's optimum.
Regarding THD with more drivers, adding more drivers increases capacitance, putting more demands on the stage driving it. Slew rate will decrease, all other things being equal. Whether that increases THD will depend on whether it's the slew rate that's dictating THD.
Regarding THD with more drivers, adding more drivers increases capacitance, putting more demands on the stage driving it. Slew rate will decrease, all other things being equal. Whether that increases THD will depend on whether it's the slew rate that's dictating THD.
Hi Suzi,
Could be. But then, what does the capacitance look like on the larger devices? Forced air cooling would help, but then you've created a small vacuum cleaner. Filter bags anyone?
Using a metal cored PCB gets away from DIY a bit. You could use a bunch of TO-126 cases (or similar) lined up in two rows that would fit a square section of aluminium tube down each side. Cheap heat sink while retaining the multiple output transistors. Higher power.
-Chris
Could be. But then, what does the capacitance look like on the larger devices? Forced air cooling would help, but then you've created a small vacuum cleaner. Filter bags anyone?
Using a metal cored PCB gets away from DIY a bit. You could use a bunch of TO-126 cases (or similar) lined up in two rows that would fit a square section of aluminium tube down each side. Cheap heat sink while retaining the multiple output transistors. Higher power.
-Chris