Discrete audio amplifer project PT20.1 distortion measurement

[@PeterKese]

@7:40 - here's how you are getting the 24.100 KHz spike: Your base frequency 20 KHz is somehow squeezed into 44.100 KHz (CD style) sampling rate. The nyquist frequency of that is 22.050. If you mirror 20.000 over 22.050 you get exactly 24.100 so what you are seeing is the 20 KHz aliased over nyquist (Monty Montgomery has a nice explanation at kzbin.info/www/bejne/fHicm5iji5aHbas). Anyways, either your source file or your audio device or anything in your audio path is 44.100 rather than 96/192 KHz. It could be windows drivers if you have old windows.

[@PeterKese]

In such situation, I'd probably start by rendering higher sampling rate test files in Audacity. Also I'd experiment with 96KHz vs 88.2 KHz (which has less noise or ghosts) to see what is matching your system's sampling rate better.

[@PeterKese]

Maybe this is what is going on: Imagine that the computer has only one single digital audio converter that it can drive either at a multiple of 44.1 KHz or 48 KHz at any single time. If your oscilloscope software decides to sample the audio at 192 KHz, then it will configure the device for a multiple of 48 KHz (usually ADC and DAC are in the same chip and running at the same frequency). If you are then playing a file sampled at 44.1 KHz on that same audio device at that same time (e.g. playing the sine wave generated by Audacity), it will need to resample that 44.1 KHz PCM signal to 48 KHz PCM and by doing that, it may introduce some aliasing.

[@PeterKese]

Another thing worth noting is that the FFT algorithm that is producing frequency spectrum for the graph is usually working on sample windows that are multiples of 2^n samples - for example 4096 samples or 8192 or 16384 samples (this is how Fast Fourier Transforms work most efficiently). You'd therefore wish to squeeze the sine wave in such a way that 4096 is divisible by the wavelentgh of your sine wave (wavelength in PCM samples). For example, if the Audacity file is generated for ~1 KHz signal at 48 KHz sampling rate, the actual FFT friendly frequency should be either 996.09375 Hz or 1007.8125 Hz (instead of 1000 Hz). Similarly for the 44.100 KHz sampling rate the ~1KHz FFT friendly frequency should be 1001.2939453125 Hz. I don't think this is really related to the aliasing issue, but if you wish to be really precise, this is worth keeping in mind.

[@johnyang799]

It's been a year since I started following the discrete amp project. A lot of things happened. I learnt a lot during the university, as well as on KZbin, on the forums. Today I just built an power amp in 2 hours for design proficiency topic. I feel really good about being myself.

[@andymouse]

I love watching this, I learn a lot, cant wait to see the boards made....cheers.

[@JohnAudioTech]

First! A lot of relevant blab in the first half of the video. Measurements begin around 13:30

[@Dazzwidd]

Well, I get the feeling that this will be a superb amplifier once it's built up on a well designed PCB. Avoidance of Ground loops and having common ground points for specific parts of the amplifier will be key, as well as low impedances in the high current ground returns. I suspect that oscillation you experienced to begin with may not have happened if it was built on a decently designed PCB. Actually... when you were doing these distortion tests, did it come to mind that you added those capacitors? I wonder how much they could be affecting your circuits behavior at high frequencies. It clearly doesn't matter for below 20khz, which is what's important anyway.

[@logotrikes]

I sometimes visit even though I might not be particularly interested in the topic at hand. Your take is what attracts me, especially if it's LM1875 or 3886. You do a thorough job John....

[@twotone3070]

Pleased it's come together for you. Well done on sticking with it.

[@1959Berre]

Snickers' distortion is never an issue.

[@イエンスヨハンセン]

Amazing work! Can’t wait to see the final schematic.

[@MassimoTava]

I’m planning on buying one of your boards or a kit If you make it.

[@Gengh13]

For the pcb I recommend KiCad, it's open source, so it will be around for the rest of our lives.

[@envisionelectronics]

I have an HP 8903 you can borrow. I’m in the Dayton area.

[@JohnAudioTech]

Hi Aaron, Thanks for the offer. I might be interested. Please contact me at [e-mail]

[@HammyTechnoid]

A frequency sweep would be telling... when exactly that non harmonic spike begins to grow. It's probably a gradual appearance.

[@cassvirgillo3395]

Hey John, 'Ol Snickers treads lightly. Great job on the amp, should sound really clean. As a long time loyal sub, Thank you for sharing your knowledge and experience, the time making videos. Learning a lot from your channel. All the best from Oregon, C.

[@isoguy.]

Awsome amp project, would really like to see a final boxed amp, power supply,good cables, bnc's and a good pair of speakers. You would think a supplier of test equipment would send you a one of their units to evaluate its effectiveness in doing this sort of work; would be a good bit of free advertising for them with an honest review. Here' hoping. 👍👍👍👍👍

[@JohnAudioTech]

I'd like to build a stereo amplifier out of this at some point. As far as test gear, the big companies seem interested only in the larger channels.

[@demoncloud6147]

Try using metal shielded casing, I recently made an amp (LiPo powered) and put everything inside a broken CPU power supply casing, very very low distortion due to Faraday Casing effect !

[@jontallbacka9406]

Get it shipped out for production John! We want it...

[@zodak9999b]

John, when you were using their generator and measuring 1kHz last time, wasn't there a pair of odd signals 1kHz on either side of 44100, like at 43100 and 45100? Now at 20kHz, there's an odd signal 20kHz away at 24100. And again with 19kHz there's a signal at 25100. I'm not sure how it's happening but there's definitely a bug in their generator, since it doesn't happen with Audacity.

[@JohnAudioTech]

I was using my music player in the last video but the difference between the Audacity generated tone and the ARTA built in tone generator is interesting. I'm no expert in DSP, perhaps they generate the tones using a different set of math.

[@zodak9999b]

@@JohnAudioTech I have no idea, but (after watching the whole video) it looks like intermod distortion with a 44.1khz signal that's not even there. Weird.

[@johnyang799]

Probably old version or compatibilities with driver or something. I have used arta for years and was able to get 0.00006% THD no issue.

[@johnyang799]

Possibly 48000khz is not the sample rate of windows audio. Use native sample rate should solve this probably.

[@johnyang799]

It would be the best to use asio4all and use asio selection in arta.

[@HillsWorkbench]

Success!

[@JohnAudioTech]

Finally!

[@bryceholland8587]

John new here to your videos, i have watched alot of your old videos and wanted to ask you some questions. I have emailed but no reply. Whats the best way to reach you? Thanks

[@clee2423]

i was waiting this video since the last one, so, if i want to measure it at 4Ω, what the value of the resistor divider should i choose ?, and is the resistor divider configuration is depend on the output voltage of the amplifier ? should i paralel the 100rΩ with 2 zener diode back-to-back to clamp the voltage ?

[@JohnAudioTech]

The load on the amplifier has little effect on the attenuator. I ended up using a 4k7 and 100 ohm in the attenuator. After the attenuater, I used a 1k ohm resistor in front of the zener voltage clamp. I added this during the test when I was having problems. I was afraid of damaging the computers line in. It is possible the junction capacitance of the zeners will cause distortion to the signal at 20Khz. I'd have to research that.

[@clee2423]

@@JohnAudioTech umm, here's a drawing of the circuit that you said imgur.com/a/v1HccbY , is it correct or not, and do i really need pre-amplifier, and i just test my PC soundcard, i found out that my PC soundcard is not too good, it end up with 0.0090% THD with 0.030% THD+N, when i use my old iphone as signal generator, the THD drops to 0.0058% and the THD+N around 0.025%, do you think this numbers pretty good, or it can be better if i have separate audio interface and a good signal generator ?

[@1pcfred]

I could have used more of a soft start with this series. Where some other amplifiers were measured so I had a basis of comparison. The extra details given here are lost on me without some frame of reference to hang them all on. Sure I've seen distortion measurements but I'm not sure if they are directly comparable to what I'm seeing here. There a larger sample size would have been appreciated.

[@JohnAudioTech]

I will likely measure my next amp, the JAT EZ AMP, the same way. I wish for better distortion analyzer equipment in the mean time.

[@1pcfred]

@@JohnAudioTech you use your FFT analyzer on a variety of amps. That helps us viewers relate to what we're seeing. I feel the lack of that aid here severely. As I've nothing to compare and contrast with.

[@pda49184]

Nice work John.. Careful that Snicker doesn't upstage you..

[@TheTrueVoiceOfReason]

Snickers says that this amp is the Cat's Meow! Nicely done. Measures very well. Devil's advocate time, though. Will you be putting any input or output protection in the amp should the inevitable part failure happen? I'd hate to see anyone try to come back at you should they screw up their kit or not purchase components from a reputable supplier, or any other number of faux reasons to try to shift responsibility off themselves.

[@JohnAudioTech]

The amp will not have onboard protection circuitry but will have a provision to add an outboard DC & short circuit protection. I will clearly state that my board/kits are for use at you own risk. I'm not a big company with lawyers.

[@TheTrueVoiceOfReason]

@@JohnAudioTech That's good to know. In today's world, you can never be too careful.

[@Dazzwidd]

@@JohnAudioTech People need to take responsibility for themselves anyway. This capitalising upon litigation when something is largely their own doing is ridiculous. Also if you blow something up, fix it, learn and move on... be more careful next time. Definitely provide a disclaimer notice though

[@iblesbosuok]

Two 1800Ω in parallel then series with 100Ω will provide 10:1 ratio. Oh, hi Snicks... Cheers from Indonesia

[@Mr.Leeroy]

Someone buy this man a dScope or whatever he needs! ffs

[@lahattec]

Magnetic fields!

[@jimmyfriend532]

Looks as if you are using a windows XP computer and out of date software , There are a lot better software programs like visual Analyser , and a older computer would be better off running Linix OS, you wold be better off and don't cost anything, both the OS and Analyser software are free! I use them my self and am very impressed with the performance of both, on my old desktop.

[@joshua43214]

You are confusing ratios with fractions - a very common thing. The ratio is 10:1, not 11:1 A 10:1 ratio has 11 parts total, making the resulting fraction 1/11 or "one in eleven." This is an unfortunate byproduct of how this material is taught (been this was for decades), and causes huge amounts of confusion. It is useful at times to express a ratio as a fraction (like when doing stoichiometric calculations). People end up thinking that a ratio of 10:1 is the same as 1/10, or in this case 1/11 = 11:1. This is simple incorrect.

[@Dazzwidd]

I thought my feelings here were right! Purely out of thinking about it, not my education. As there is 1000:100 which is 10:1, simple. But the voltage is 1/11 haha it's trying to be tricky. Makes it pretty simple: if 10V is where it starts it's 10V/11= 0.90909V

[@saiprasad8078]

Hello

[@gilangramadhan8052]

Please play sound your project.... classic sound..... viewers not believed is you not player audio