Thank you. On the other tests it seems to perform well indeed. I was just exploring the various tests on this new tool. I had seen an example of a valve amplifier performing a near-perfect square wave reproduction on this test, so assumed it was possible, but if this topology is inherently going to always behave this way and that's realistically not going to affect my musical enjoyment then I'll concentrate my energies elsewhere 👍 🙂
You should be aware that any square wave irregularity can be observed for the sinus wave also ... in this case a peak at about 35KHz , preatty big by the look of the overshoot . You can accept any crap outside audio band , or not , your choice .
Last edited:
Resoldered a few spot that fixed 1/2 the problem. Still ringing at the left edge. What is the voltage rating for the cap at the NFB? Currently I am using 50v for testing.
After saw what dwjames did with grid stopper resistor, I realized I needed to try the vol. pot. up/down to find where it could stop ringing at the left edge. However, it didn't work out and the pot best is at 75% and below. As 6A3sUMMER's mention about NFB, so I kept adding 20pf at a time until 480pf. But at this value, the Frequency response is cut off a lot at 20Khz which I don't know it's good or bad. 1000pf looked perfect but frequency response cut off at 10khz so i guess it's not good.
The red channel is the one I am working on
Here are 3 that I think is good enough, Which one should I need?
1. Less frequency response cut off but still ringing at the bottom left.
2. More frequency response cut off at 20Khz but not sure if it's ok.
3. More frequency response cut off at 20Khz and no more ringing.
After saw what dwjames did with grid stopper resistor, I realized I needed to try the vol. pot. up/down to find where it could stop ringing at the left edge. However, it didn't work out and the pot best is at 75% and below. As 6A3sUMMER's mention about NFB, so I kept adding 20pf at a time until 480pf. But at this value, the Frequency response is cut off a lot at 20Khz which I don't know it's good or bad. 1000pf looked perfect but frequency response cut off at 10khz so i guess it's not good.
The red channel is the one I am working on
Here are 3 that I think is good enough, Which one should I need?
1. Less frequency response cut off but still ringing at the bottom left.
2. More frequency response cut off at 20Khz but not sure if it's ok.
3. More frequency response cut off at 20Khz and no more ringing.
Attachments
There's another way to approach that ringing. Restore C1 to it's original value. Now look at C4 and R7. These make up a low-pass filter, or "step-filter" as it is sometimes called. This is a standard feature in amps with global feedback. Try increasing C4 to 200pF and see if that reduces the ringing. The "step" filter actually limits the HF response. C1, or the "phase lead" cap in the feedback loop, is more intended to target a specific phase issue in the output transformer, than it is to limit the overall frequency response.
The next step is to check for stability. You can do that by dabbing a .1uF cap across the test load resistor and see if a 10kHz square wave holds up or collapses. If the amp goes into oscillation you will see it immediately.
Look at this Williamson schematic I developed (with some help!). You see a fairly steep low-pass filter in the front (around 40kHz) but a very light phase-lead cap across the feedback resistor. The amp is very stable and the bandwidth is flat well past 20kHz.
I bring up stability because a stable feedback amp will sound much better than one that flirts with oscillation.
The next step is to check for stability. You can do that by dabbing a .1uF cap across the test load resistor and see if a 10kHz square wave holds up or collapses. If the amp goes into oscillation you will see it immediately.
Look at this Williamson schematic I developed (with some help!). You see a fairly steep low-pass filter in the front (around 40kHz) but a very light phase-lead cap across the feedback resistor. The amp is very stable and the bandwidth is flat well past 20kHz.
I bring up stability because a stable feedback amp will sound much better than one that flirts with oscillation.
If post 1 is the schematic then you have inherent phase problems with a Se it doing phase splitting on sec. There will be phase issues.
Just coming back to this as I literally do find square waves in the music that I make and listen to. I'm sure for someone who only listens to classical and acoustic recordings this is true, but a lot of electronic music intentionally has all sorts of simple and complex waveforms going on.
I guess it's likely not an issue for my ears regardless, but it bothers me that this overshoot represents unwanted extra high frequency energy which my power stage and speakers are being asked to amplify and reproduce and that seems like a negative thing.
Would it be such a bad thing to snub/low pass this transformer secondary at 30khz for example, which should be totally out of the way of any valid red book musical signal?
Also, one test I didn't do with the scope was to see how things looked if I only applied the input signal to one half of the 6SN7. It really looks to me like this flick could be a phase issue with combining 2 waveforms with slightly different timings.
Sorry, but you 'literally' do not find square waves in your music. Play a square wave from your CD player and see what you get. It can't do it.
Overshoot is not "unwanted extra high frequency energy" - it is the absence of higher order harmonics from an infinitely fast rising/falling edge. Your CD player will exhibit what you perceive as nasty overshoot - you believe red book is putting out extra high frequency energy??? Then why the overshoot from your CD player? Start to understand this concept and you will start to understand what your amp is doing.
Or you can band aid a perceived problem by intentionally limiting bandwidth. It would not surprise me in the least that your brain will tell you it sounds better when the square wave is pretty.
And please, this thread needs to isolate the issue to an amp without negative feedback.
Overshoot is not "unwanted extra high frequency energy" - it is the absence of higher order harmonics from an infinitely fast rising/falling edge. Your CD player will exhibit what you perceive as nasty overshoot - you believe red book is putting out extra high frequency energy??? Then why the overshoot from your CD player? Start to understand this concept and you will start to understand what your amp is doing.
Or you can band aid a perceived problem by intentionally limiting bandwidth. It would not surprise me in the least that your brain will tell you it sounds better when the square wave is pretty.
And please, this thread needs to isolate the issue to an amp without negative feedback.
I'm not looking to argue or frustrate. Just hear to learn and share 🙂 If that's the case, and this is isolated to a test environment, then that's great and I'll leave it alone.
I'll do a test later and make a track of a simple synth melody using a single square wave oscillator, save it as 44.1KHz 16bit and play the wav file via my DAC and check the scope again. Just so I can understand better what a real world scenario is like 🙂
I was just worried that this amplifier recreating signals like that from my music may trigger this overshooting and that the ringing would create additional signal, which whilst only audible for non-humans, would create additional peaks of energy (which I say because it looks like the overshoot signal reaches 15v peak to peak, whereas it looks like if reproduced the square wave as intended and did not ring, then the signal would be more like 11v peak to peak) for my gear to deal with, whether I can hear it or not.
I think there's some confusion here about the difference between Gibb's Phenomenon (shown in post #48) and square-wave testing with a wide-bandwidth signal. Yes, CD players produce square waves like the one in #48 and that is indicative of the limited frequency response. But we test amplifiers with a wide-bandwidth signal in order to determine the amplifier's actual behavior. In that case, leading edge overshoot does indicate an excess of HF information, and ringing indicates resonances and/or phase shifts in the highs that can lead to oscillation. You can easily see this by running a Bode plot analysis.
Here's a 10kHz square wave and Bode plot for my Williamson amplifiers with 20dB feedback. First, a "standard" Williamson step filter of 4.7K ohms in series with 200pF capacitor, or about a 100kHz low-pass, with a phase-lead cap of around 150pF. Stability is not too good:
Here's the same amp with a much steeper step filter at the input, about 40kHz. There's still some rise at the leading edge but no ringing. This version is much more stable into a partially capacitive load. The HF resonance points are still there (needs better tuning!) but are pushed much further down.
Here's a 10kHz square wave and Bode plot for my Williamson amplifiers with 20dB feedback. First, a "standard" Williamson step filter of 4.7K ohms in series with 200pF capacitor, or about a 100kHz low-pass, with a phase-lead cap of around 150pF. Stability is not too good:
Here's the same amp with a much steeper step filter at the input, about 40kHz. There's still some rise at the leading edge but no ringing. This version is much more stable into a partially capacitive load. The HF resonance points are still there (needs better tuning!) but are pushed much further down.
Last edited:
And in case you're wondering, is that bump at 100kHz audible? Not directly, but it makes for a "coarser" sound in the highs. The stabilized amplifier is much more pleasant to listen to.
dwjames,
thank you you open this threat, I wasn't pay attention until you posted it. Hope, you can sort it out. I will stop now so others don't get confuse with 2 different amps.
6A3sUMMER and rayma,
I got the idea and figured out the best with your help.
grovergardner
I will try your suggestion later and don't want to hijack dwjames's thread too much. So far, it's working great now.
thank you you open this threat, I wasn't pay attention until you posted it. Hope, you can sort it out. I will stop now so others don't get confuse with 2 different amps.
6A3sUMMER and rayma,
I got the idea and figured out the best with your help.
grovergardner
I will try your suggestion later and don't want to hijack dwjames's thread too much. So far, it's working great now.
Also, many current speakers have tweeters with large resonances above 20kHz. Matching such a speaker with such an amplifier
would be very unfortunate.
5 stars to anyone who can read all of this:
There are lots of definitions of "ring" and of "ringing"
Take a Perfect square wave, and disturb either the phase, or the amplitude of one, and only one, odd harmonic, you get "ring".
Hey, wake up, this is NOT an oscillation.
Just imagine if the phase and/or the amplitude of two or more odd harmonics are disturbed.
Non negative feedback Amplifiers do not have perfect frequency response; and they do not have perfect phase response.
Just exactly like the amplifier in Post # 1.
And matter how bad the "ring" looks on that and all non negative feedback amplifiers, it is not an oscillation.
A church bell "rings"
No square wave starts it.
A single impulse (Bang!) starts it.
There is no negative feedback here either.
Many CD players play back at a 44.1kHz sample rate (exactly what comes off of a Red Book CD).
Check out the Denon Audio Technical CD (C-39-7147-EX):
Zoom in on the 100Hz square wave of the Denon Audio Test CD; "ring" at 22.05kHz.
You get the following:
When played back on the old Analog Filter CD players, there was an exponentially decaying "ring" After the rise, and an exponentially decaying"ring" After the fall.
Even though the first screen shot in Post # 52 is from a square wave generator sent through a power amplifier, it looks a lot like those old analog filter CD players playing a square wave from the CD.
The CD player bandwidth is 22.05kHz (the analog filter cuts off frequencies that are higher than nyquist).
When played back on the original Digital Filter CD players there was an exponentially Increasing "ring" Before the rise, and and an exponentially Decaying "ring" After the rise. The same thing happens before and after the fall (call it "pre-ring" and "post-ring").
The largest "ring" amplitude is Always right before and right after the rise & fall transisions.
The first (top) screen shot in Post # 48 is an excellent example of that kind of "ring".
The bandwidth is still 22.05kHz.
In CD player's later designs, Marketing decided to oversample the 44.1kHz CD data stream.
Faster is better right? Sales will Increase, and Marketing says you need a brand new CD player that oversamples.
Red Book CD 44.1k datastream gets oversampled 2X or 4X; 88.2k samples/second, 176.4k samples/second respectively.
. . . Suppose some of those higher frequencies get past [get through] the new improved CD player (Nyquist at 44.2kHz or 88.4kHz) Then . . .
That will be a problem. Those higher frequencies can activate a power amplifiers "ring" frequency, and can inter-modulate with other high frequencies and that difference frequency may fall into the audio band.
Most things have tradeoffs.
Many old things require new things in order to "Fix" the old tradeoffs, at the expense [and expen$e] of potentially creating some new tradeoffs.
Do a lot more measurements, or just read the review test results of a 2x or 4x oversampled CD player, and find out where all the "noise" of 1X 44.1k sample rate goes . . .
It becomes new "noise" frequencies, higher frequencies than the audio band . . . which is nice.
But, those new "noise" higher frequencies are where power amplifiers global negative feedback starts to fall apart, and Intermodulation distortion increases.
No matter how small of an amplitude those artifact are, they can be measured . . .
Now that you know about them you might not sleep at night, worrying about them.
Hopefully, if you managed to read through all of the above, CD players work far better than many worry-warts think they do.
Some of you (not the original poster) who want to improve the non negative feedback amplifier in Post # 1,
If you have any single ended or any push pull amplifier that uses global negative feedback, then please do a 10kHz square wave test (using a non-inductive 8 Ohm load resistor with a 0.1uF capacitor across the 8 Ohm resistor . . .
Then please post a screen shot of that.
Those who do not do that test, perhaps should not complain about the wonderful amplifier in Post # 1.
The original poster was right to ask his questions in his original post.
And I bet he is a lot more knowledgeable than when he started; in spite of all the new items and possible fears, I bet he sleeps at night, right after listening to his amplifier.
Enjoy the Music!
There are lots of definitions of "ring" and of "ringing"
Take a Perfect square wave, and disturb either the phase, or the amplitude of one, and only one, odd harmonic, you get "ring".
Hey, wake up, this is NOT an oscillation.
Just imagine if the phase and/or the amplitude of two or more odd harmonics are disturbed.
Non negative feedback Amplifiers do not have perfect frequency response; and they do not have perfect phase response.
Just exactly like the amplifier in Post # 1.
And matter how bad the "ring" looks on that and all non negative feedback amplifiers, it is not an oscillation.
A church bell "rings"
No square wave starts it.
A single impulse (Bang!) starts it.
There is no negative feedback here either.
Many CD players play back at a 44.1kHz sample rate (exactly what comes off of a Red Book CD).
Check out the Denon Audio Technical CD (C-39-7147-EX):
Zoom in on the 100Hz square wave of the Denon Audio Test CD; "ring" at 22.05kHz.
You get the following:
When played back on the old Analog Filter CD players, there was an exponentially decaying "ring" After the rise, and an exponentially decaying"ring" After the fall.
Even though the first screen shot in Post # 52 is from a square wave generator sent through a power amplifier, it looks a lot like those old analog filter CD players playing a square wave from the CD.
The CD player bandwidth is 22.05kHz (the analog filter cuts off frequencies that are higher than nyquist).
When played back on the original Digital Filter CD players there was an exponentially Increasing "ring" Before the rise, and and an exponentially Decaying "ring" After the rise. The same thing happens before and after the fall (call it "pre-ring" and "post-ring").
The largest "ring" amplitude is Always right before and right after the rise & fall transisions.
The first (top) screen shot in Post # 48 is an excellent example of that kind of "ring".
The bandwidth is still 22.05kHz.
In CD player's later designs, Marketing decided to oversample the 44.1kHz CD data stream.
Faster is better right? Sales will Increase, and Marketing says you need a brand new CD player that oversamples.
Red Book CD 44.1k datastream gets oversampled 2X or 4X; 88.2k samples/second, 176.4k samples/second respectively.
. . . Suppose some of those higher frequencies get past [get through] the new improved CD player (Nyquist at 44.2kHz or 88.4kHz) Then . . .
That will be a problem. Those higher frequencies can activate a power amplifiers "ring" frequency, and can inter-modulate with other high frequencies and that difference frequency may fall into the audio band.
Most things have tradeoffs.
Many old things require new things in order to "Fix" the old tradeoffs, at the expense [and expen$e] of potentially creating some new tradeoffs.
Do a lot more measurements, or just read the review test results of a 2x or 4x oversampled CD player, and find out where all the "noise" of 1X 44.1k sample rate goes . . .
It becomes new "noise" frequencies, higher frequencies than the audio band . . . which is nice.
But, those new "noise" higher frequencies are where power amplifiers global negative feedback starts to fall apart, and Intermodulation distortion increases.
No matter how small of an amplitude those artifact are, they can be measured . . .
Now that you know about them you might not sleep at night, worrying about them.
Hopefully, if you managed to read through all of the above, CD players work far better than many worry-warts think they do.
Some of you (not the original poster) who want to improve the non negative feedback amplifier in Post # 1,
If you have any single ended or any push pull amplifier that uses global negative feedback, then please do a 10kHz square wave test (using a non-inductive 8 Ohm load resistor with a 0.1uF capacitor across the 8 Ohm resistor . . .
Then please post a screen shot of that.
Those who do not do that test, perhaps should not complain about the wonderful amplifier in Post # 1.
The original poster was right to ask his questions in his original post.
And I bet he is a lot more knowledgeable than when he started; in spite of all the new items and possible fears, I bet he sleeps at night, right after listening to his amplifier.
Enjoy the Music!
Last edited: