Once again, no test setup or equipment list. Pointless to post this without the rest of the information. These are film, no shock Polyester has poor behavior. We all know that. Not only that, they appear to be pulse tests. Let's try a sine instead of a turn-on thump.
Okay, test circuit? Equipment?
Relationship to sine wave distortion?
All I can see is that you don't like distorted turn-on thumps.
You need to post data relevant to the discussion with everything needed for others to assess it without doing research and digging on their own. I'm sure I read this earlier in time and discounted it. That information is important in pulse circuits, pulse shapers and other circuits that manipulate pulses. The pulse has frequencies far higher than audio as well.
We all know Polyester capacitors aren't the best for audio depending on circuit position and circuit impedance. That's a given. What you need to do is read reports yourself and assess how similar the test conditions are to a circuit you're using is. If they aren't, time to design a test yourself using appropriate equipment and test conditions. Make it repeatable by someone else and suggest it may be relevant and wait for agreement from others replicating your tests.
If a test does not apply to your circuit conditions, you can either ignore that test and design your own, or ignore it. "Read across" has downed aircraft and killed people. That is what you are doing right now.
Relationship to sine wave distortion?
All I can see is that you don't like distorted turn-on thumps.
You need to post data relevant to the discussion with everything needed for others to assess it without doing research and digging on their own. I'm sure I read this earlier in time and discounted it. That information is important in pulse circuits, pulse shapers and other circuits that manipulate pulses. The pulse has frequencies far higher than audio as well.
We all know Polyester capacitors aren't the best for audio depending on circuit position and circuit impedance. That's a given. What you need to do is read reports yourself and assess how similar the test conditions are to a circuit you're using is. If they aren't, time to design a test yourself using appropriate equipment and test conditions. Make it repeatable by someone else and suggest it may be relevant and wait for agreement from others replicating your tests.
If a test does not apply to your circuit conditions, you can either ignore that test and design your own, or ignore it. "Read across" has downed aircraft and killed people. That is what you are doing right now.
All of these phenomena are displaced down in frequency well below the stimulus. Look at the top left photo in your post where the 1 to 0 square wave stimulus is shown. Depending upon the fall time, the fundamental and harmonics are orders of magnitude higher than the DA-attributed photo below it which is what I was referring to earlier in this thread. The bottom plot below shows a stimulus with a 10ms on time and a 1 sec off time with a 10us fall time for a 100uF capacitor. V(vda) is the DA part of the capacitor and I(R1) is the current flowing in the series resistance part of the DA model. The kind of stimulus used to expose the DA component is highly unlikely with a real-world music signal, but it does show that DA components are shifted many orders of magnitude down in frequency and I'd add shows a remarkable correlation with Richard Marsh's photo. For normal bandwidth-limited audio signals with rise times measured in 10's of us (and especially now with digital audio where the real music signal is bandwidth limited to a shade over 20 kHz), DA is not an issue. So, the proposition that DA charge is released post the stimulus event in a manner that can introduce non-stimulus related harmonics that would smear the audio signal but not be measurable is unfounded in my view - and that's exactly what Richard Marsh's photos show, but his interpretation on these matters differs of course. There was quite some debate on this on the Blowtorch thread if you recall, and those making the smearing claim could not point to measurements or DBT in support of their hypothesis. If you run the analysis with 5 kHz square wave, the down-shifting of the DA component is even more clearly visible.
The photo below is with a 10uF electrolytic, and the DA component reduced by a factor of 10 as well
Here is a DA model
I've certainly never had issues with the Nichicon UES-series when sized for, say, 0.2-2 Hz cutoff in AC coupling applications. In my HP-2 I use several UES-series in the signal path and also a generic (but good) electrolytic capacitor in the output. They all have ~14 V DC across them as the HP-2 runs from a single power supply rather than a bipolar one. I wanted to see if I could make a high-end, single-rail design.
The distortion of the completed amp is so low that I had to resort to some tricks to even measure it. I did not have the APx555 at the time. "Only" the APx525 which bottoms out around -125 dBc on the analyzer (and -113 dBc on the source). So I designed a low-distortion notch filter and a 5th order elliptical filter to clean up the source. The resulting measurement is below with annotation added. The IMD products are from the measurement setup (SMPS powering the filters).
It's hard to argue that this isn't low distortion. Here's the multi-tone measurement:
The 60 Hz (and harmonics) are from my bench top DMM, which sits on top of the AP. If I forget to turn it off, there's enough mains hum that radiates into the input stage of the AP to show up around -130 dBV @ 60 Hz. That's one of my (very few) disappointments with the AP. It took me a very long time to find that. Ahhh... Life in precision audio. 🙂
In case you're wondering: If you're a frequent reader of ASR you'll notice that Amir places the peaks of the test tones at 0 dB, whereas in my tests they're at about -22 dB. That's because when you add up all the tones (and account for their relative phases) you get to the 0 dB point. I.e., 0 dB in my test is the RMS sum of all the tones, thus representative of the overall signal amplitude, where as Amir uses the RMS value of a single tone.
I believe the proper way to look at this is to measure the IMD 'grass' relative to the RMS value of the overall test signal, not the amplitude of one of the tones.
Probably, but not necessarily. It's the usual "buyer beware" situation. If there's hardly any signal across the capacitor, as would be the case for a feedback capacitor that sets the LF pole for example, there's hardly any variation in capacitance, thus hardly any distortion. Maybe "hardly any" is good enough in your application.
I would certainly not recommend using X7R as a coupling capacitor if you're aiming for the lowest distortion. In particular not if C0G/NP0 is available in that capacitance.
There's a slide deck by Bruce Hofer (AP) floating around in cyberspace. It shows that C0G/NP0 provides even better performance than PP film. I'll see if I can dig that up.
Tom
By coincidence I came across the original reference: Walt Jung & John Curl (1985) A Real-Time Signal Test for Capacitor Quality. Audio Amateur 4/85.
@nicoch58 would you please post the full reference (like I just did) or a link to the article instead of context-less blurbs? That way the rest of us can better relate to what you're trying to convey and you also credit the original authors.
Jung & Curl set up a bridge using a known good reference capacitor and the capacitor under test. They then applied a step input and measured the resulting output.
That's an interesting test and it would likely be very relevant for something like S&H amplifiers as we discussed earlier in the thread. Whether the conclusion also applies for sine wave testing remains to be seen. It would have been easy for Jung & Curl to push the sine wave button on the signal generator and measure, but sadly their article only shows the results of step response testing.
Tom
Hi Tom,
I think the Jung / Curl tests were done in a way to show something. There is probably not enough effect as a coupling cap to show much to worry about. Also, as you point out, no signal voltage across the cap means it shouldn't affect the signal very much. I think most people can agree with that. Moral: Choose your capacitor size intelligently as a coupling cap. Filter determining components become important as do any caps that do see a signal voltage across them.
Your levels look correct. Running all frequencies at 0 dB (dBm or dBFS?) may well overload the circuit or cause some issue. Some test sets work better with lower than 0 dBFS input level, depends on the A/D. A level meter might give you a good idea of actual total signal level. Math is boring.
I read a little of ASR and was not impressed.
Tom, what do you use a multitone measurement for? I find THD and IMD tests show pretty much everything. I have looked at multitone results with my gear.
I think the Jung / Curl tests were done in a way to show something. There is probably not enough effect as a coupling cap to show much to worry about. Also, as you point out, no signal voltage across the cap means it shouldn't affect the signal very much. I think most people can agree with that. Moral: Choose your capacitor size intelligently as a coupling cap. Filter determining components become important as do any caps that do see a signal voltage across them.
Your levels look correct. Running all frequencies at 0 dB (dBm or dBFS?) may well overload the circuit or cause some issue. Some test sets work better with lower than 0 dBFS input level, depends on the A/D. A level meter might give you a good idea of actual total signal level. Math is boring.
I read a little of ASR and was not impressed.
Tom, what do you use a multitone measurement for? I find THD and IMD tests show pretty much everything. I have looked at multitone results with my gear.
Just thinking aloud, I see much worry and spilled blood about a non issue.
Phenomena discussed here are well below the threshold of hearing which in my book qualifies them as such.
"Non existing"?
Nobody says so, being measurable is proof of existence.
"Nice to know"?
Yes, why not?
Important? ...
Not terribly. 🤷🏻
Phenomena discussed here are well below the threshold of hearing which in my book qualifies them as such.
"Non existing"?
Nobody says so, being measurable is proof of existence.
"Nice to know"?
Yes, why not?
Important? ...
Not terribly. 🤷🏻
Found it!! Bruce Hofer's talk is from the 139th AES Convention in 2015. Here're some relevant excerpts regarding ceramic vs film caps.
From what I understand, the APx555 uses many capacitors (and resistors) in parallel to drive the distortion down. For the ~$20k cost adder of the 555 relative to the 525 there should be room in the parts budget for quite a few NP0 capacitors. 🙂
Tom
From what I understand, the APx555 uses many capacitors (and resistors) in parallel to drive the distortion down. For the ~$20k cost adder of the 555 relative to the 525 there should be room in the parts budget for quite a few NP0 capacitors. 🙂
Tom
It would be interesting to see if there was a difference between film and foil construction. They indicate a mix, which would be odd.
Anyway, I can believe that data.
Anyway, I can believe that data.
Oh, I agree. The article is from 1985, so the body of knowledge is larger now. You have to start somewhere. Also, their end application might not have been audio.
My testing procedure is as follows: I adjust the amplitude of the sine source to the desired output level (typically maximum output power or some established reference power level). I then note the RMS output voltage and enter that as the 0 dB reference for the analyzer. I then switch to the multi-tone signal with the same RMS voltage. Thanks to the way the AP software works, this results in an output of the DUT that has the same peak levels as it did with the sine wave. So I know my circuit is not clipping. I then run the FFT to get the multi-tone result.
So, if I use the max output power (i.e., the highest output before clipping) with a sine wave as the reference, the 0 dB reference in the FFT of the multi-tone signal, is the highest signal level that can be produced before clipping. I think that's the truest representation of the IMD (and, thus, dynamic range) of the system. It's also generally what pushes the system the hardest. It's generally easier to get good performance at 1 W than at 100 W in a 100 W rated amp, for example.
I measure three types of 2-tone IMD and then add the multi-tone for completeness. The multi-tone signal is much closer to a music signal than a sine wave, which I why I use it. I honestly see little value in the 2-tone tests if a multi-tone test is available. I provide the 2-tone numbers so my prospective customers can compare the performance of my products against others.
Ah. So I'm not the only one who thinks that. 🙂
I think it was relevant when it started. Now it looks like Amir has lost interest and/or is overwhelmed with review samples. There's been several examples where he's tested something and found that it didn't perform well only to later disclose deep in the review thread that, "oh, by the way, I had some mode enabled. Here's the correct result." That's pretty disappointing - and, frankly, not scientifically valid.
I commend Amir for his efforts into increasing the transparency in the audio field. I just wish he'd validate his setups before posting the review.
The culture there has drifted way too far in the direction of "how many dB SINAD per dollar is this kit?" in my view. It was better a few years ago.
Anyway. Back to capacitors.
Tom
"PP-film and foil" refers to polypropylene film and metal foil. Think Wima FKP. It's not a mix of different capacitor types.
Tom
Well, what the Jung-Marsh tests definitely show is don’t use an aluminium electrolytic or similar for a S+H!
Thanks Tom, I would call them film types though.
Performance at 1 watt is closer to how people really listen to their system (except me of course! lol!, my speakers are about 86 dB/watt and 4 ohm). The THD is normally worse at one watt unless it is a single ended or tube product. The issue with looking at distortion up at higher levels is that the speaker system quickly outstrips the amplifier in distortion. It is valid for determining how well the amplifier is designed, no argument there. The average person doesn't understand from a system standpoint. That and maximum power is often more a test of the mains supply voltage and impedance. I guess that is how some really iffy "well respected" amplifiers survive in the market.
So, just curious now. What does a multitone test show you? IMD tests are very revealing (I use 19 KHz and 20 KHz equal level). You should see the sidebands on some amps. The THD tests are also very revealing and I am looking at noise floor, plus skirts. The two tone tests give me comparable numbers to the industry accepted ones. A multitone test is non-standard. I also fail to see how it is more difficult to an amplifier to reproduce that signal than the standard IMD test. The other two IMD tests are simply a different stimulus, I haven't seen where any provide more information. I am not disagreeing with you Tom, I just don't get any further information, therefore the other tests are a waste of time.
I have seen Amir make some technical errors. That's okay, generally that site isn't useful from what I have seen.
-Chris
Performance at 1 watt is closer to how people really listen to their system (except me of course! lol!, my speakers are about 86 dB/watt and 4 ohm). The THD is normally worse at one watt unless it is a single ended or tube product. The issue with looking at distortion up at higher levels is that the speaker system quickly outstrips the amplifier in distortion. It is valid for determining how well the amplifier is designed, no argument there. The average person doesn't understand from a system standpoint. That and maximum power is often more a test of the mains supply voltage and impedance. I guess that is how some really iffy "well respected" amplifiers survive in the market.
So, just curious now. What does a multitone test show you? IMD tests are very revealing (I use 19 KHz and 20 KHz equal level). You should see the sidebands on some amps. The THD tests are also very revealing and I am looking at noise floor, plus skirts. The two tone tests give me comparable numbers to the industry accepted ones. A multitone test is non-standard. I also fail to see how it is more difficult to an amplifier to reproduce that signal than the standard IMD test. The other two IMD tests are simply a different stimulus, I haven't seen where any provide more information. I am not disagreeing with you Tom, I just don't get any further information, therefore the other tests are a waste of time.
I have seen Amir make some technical errors. That's okay, generally that site isn't useful from what I have seen.
-Chris
Mulitone tests - and especially ones similar to those that Tom showed above - will generate lots of IMD components that will show up as base line hash and/or IMD components in amplifiers with suboptimal linearity. Really good amps will show 18 bits or more dynamic range (peak tone level to base line grass) and I think Amir has shown a few products (preamp IIRC) at 20 or 21 bits. These kinds of stimulus signals were difficult to generate prior to the advent of digitisation, but really offer new insights into amplifier performance not possible before. Importantly, they much more closely mirror an actual music signal with simultaneous multiple frequency stimuli.
Hi Bonsai,
Thank you. But looking at the grass using two tone is sufficient I think. You can see all you need to see. The electronics just sees a signal, the more tones, the more IM products, but that doesn't matter really. In addition, THD will make a real mess of multiple tones I find.
Anyway, I don't think a circuit has any more difficulty with two or ten tones at once. ONce frequencies beat and produce sum and difference tones, why would more of them matter?
I'm using an RTX 6001, so I can "see" way down there.
Thank you. But looking at the grass using two tone is sufficient I think. You can see all you need to see. The electronics just sees a signal, the more tones, the more IM products, but that doesn't matter really. In addition, THD will make a real mess of multiple tones I find.
Anyway, I don't think a circuit has any more difficulty with two or ten tones at once. ONce frequencies beat and produce sum and difference tones, why would more of them matter?
I'm using an RTX 6001, so I can "see" way down there.
Because in a lot of amplifiers
1. The amount of feedback changes with frequency
2. distortion profiles (harmonic content) vary dramatically
Where two tone IMD tests really help is when they are conducted at HF (eg 19 and 20 kHz) and at full power because they will expose SID which you won’t see with multi-tone as the amplitudes aren’t high enough.
1. The amount of feedback changes with frequency
2. distortion profiles (harmonic content) vary dramatically
Where two tone IMD tests really help is when they are conducted at HF (eg 19 and 20 kHz) and at full power because they will expose SID which you won’t see with multi-tone as the amplitudes aren’t high enough.