Nice explanations all.
One thing I was thinking is my tweeters have a 50kHz frequency response - they're amazing. Even though I can't hear such frequencies I think it would make sense that I can hear my tweeters trying to reproduce such frequencies if they are present in the signal. Surely there is some kind of subsonic artefact of a tweeter trying to replicate a 50kHz noise?
I also think hiss from speakers is at least partially caused by noise, so less hiss = less noise.
Then there's some interesting things to think about in terms of the affect these supersonic frequencies have on the amp (as Mike was suggesting). I understand the mechanism, I just did not expect the affect to be so dramatic (which I will confirm tonight hopefully).
One thing I was thinking is my tweeters have a 50kHz frequency response - they're amazing. Even though I can't hear such frequencies I think it would make sense that I can hear my tweeters trying to reproduce such frequencies if they are present in the signal. Surely there is some kind of subsonic artefact of a tweeter trying to replicate a 50kHz noise?
I also think hiss from speakers is at least partially caused by noise, so less hiss = less noise.
Then there's some interesting things to think about in terms of the affect these supersonic frequencies have on the amp (as Mike was suggesting). I understand the mechanism, I just did not expect the affect to be so dramatic (which I will confirm tonight hopefully).
Errm..
On a logarithmic frequency scale, the difference between say, 25 kHz and 50 kHz is scarcely noticeable. What is probably relevant is that the figures are quoted for often 6 or even 10dB down from the median response and this means you don't actually get full response at 20kHz anyway. Quoting the response to 50kHz may still mean it sags at 20kHz, according to many published driver specs. anyway.
IMHO, unless you have some rare, honest figures quoting flat or -3dB response to 50 kHz there, it's more about specsmanship than reality, in audible terms.
On a logarithmic frequency scale, the difference between say, 25 kHz and 50 kHz is scarcely noticeable. What is probably relevant is that the figures are quoted for often 6 or even 10dB down from the median response and this means you don't actually get full response at 20kHz anyway. Quoting the response to 50kHz may still mean it sags at 20kHz, according to many published driver specs. anyway.
IMHO, unless you have some rare, honest figures quoting flat or -3dB response to 50 kHz there, it's more about specsmanship than reality, in audible terms.
My signal generator is the Systron-Donner Datapulse Versatester 1. Old, from the 70's! Thanks for the offer.
First of all, where is the ultrasonic noise coming from? It is present at all? Have you measured it? Is it possible this is just a matter of the input stage being protected from rails modulated by the output stage? After all, each half of the output stage alone will show it's own high level of distortion, and far greater when AB kicks in. It makes great sense to filter this out of the frontend rails. Go in simulation and plot the FFT of the MOSFET source currents! Now plot the FFT of the rail voltage. Of course none of this will matter if you haven't taken the time to properly enter the rail capacitor ESR and ESL (say 22mR and 33nH).
I'm all but dead sure that the harmonics of the rail modulation will be larger than any ultrasonic noise imposed on the rails.
- keantoken
First of all, where is the ultrasonic noise coming from? It is present at all? Have you measured it? Is it possible this is just a matter of the input stage being protected from rails modulated by the output stage? After all, each half of the output stage alone will show it's own high level of distortion, and far greater when AB kicks in. It makes great sense to filter this out of the frontend rails. Go in simulation and plot the FFT of the MOSFET source currents! Now plot the FFT of the rail voltage. Of course none of this will matter if you haven't taken the time to properly enter the rail capacitor ESR and ESL (say 22mR and 33nH).
I'm all but dead sure that the harmonics of the rail modulation will be larger than any ultrasonic noise imposed on the rails.
- keantoken
Let me backpedal a bit. I'm not opposed to the idea that ultrasonic noise is causing problems, I'm sure it does cause something, but I'm not sure it is the cause of the effect we are observing. Let me propose a test. Why not filter the rails, and then add some ultrasonic noise intentionally?
On another note, I know that ultrasonic signals are used to "bias" magnetic tape for recording, though I don't know what it is for. I suspect it decreases noise and distortion. I wonder if ultrasonic noise could "bias" a ferrite speaker magnet in some way? Only certain types of ferrite are useful for RF work. On one of my really old threads someone said this was being done in some speakers, but they didn't provide any details.
- keantoken
On another note, I know that ultrasonic signals are used to "bias" magnetic tape for recording, though I don't know what it is for. I suspect it decreases noise and distortion. I wonder if ultrasonic noise could "bias" a ferrite speaker magnet in some way? Only certain types of ferrite are useful for RF work. On one of my really old threads someone said this was being done in some speakers, but they didn't provide any details.
- keantoken
Hello KT
One thing I can say is that any noise and especially rf noise will modulate the amp signals and it can even do intermodulations in the signal, so it can kill resolution and soudstage. Better to filter all noises anyway.
The bias signal in a magnetic tape reduce the nonlinearity and it's overcome the magnetic hysteresis (magnetic material do resist to being magnetized or releasing its magnetism) so it done much better recording with much lower thd.
But bias cannot be apply to ferrite speaker, those rf noise will mostly intermodulate with the music in a tweeter.
Bye
Gaetan
Last edited:
In most systems there's plenty of it there already - from switching diodes, mains etc - so it would be more productive and merciful on the ears to do an experiment to reduce it !
Absolutely agree !
But I have one question:
What is the lowest frequency that is usually called / designated a Radio Frequency ?
Hello
3 khz to 30 khz are VLF and so 3 khz are the lowest radio frequencies, 30 khz to 300 khz are LF in radio frequencies, and so on for the radio frequencies...
Bye
Gaetan
Last edited:
This seems to answer my question
Radio spectrum - Wikipedia, the free encyclopedia
but elsewhere I read:
Radio frequency (RF) is a rate of oscillation in the range of about 3 kHz to 300 GHz
Anyway, which ever is true < 3hz or 3khz, anything above 20khz can be designated as RF
That's my new factoid for today 😀
Radio spectrum - Wikipedia, the free encyclopedia
but elsewhere I read:
Radio frequency (RF) is a rate of oscillation in the range of about 3 kHz to 300 GHz
Anyway, which ever is true < 3hz or 3khz, anything above 20khz can be designated as RF
That's my new factoid for today 😀
Last edited:
Hi Greg,
Very nicely done PCB & I like the ground plane.
My single-sided boards does not hiss. My source is a modified Squeezebox 3 and it pick up some crackle from the wireless router (I can only hear that when the volume is at maximum whilst the Squeezebox is paused). I turned off the router & the crackle is gone.
I suggest you to ground the input of the Fetzilla with a short piece of wire & that will help to isolate the source of the hiss.
Cheers, Stanley
Very nicely done PCB & I like the ground plane.
My single-sided boards does not hiss. My source is a modified Squeezebox 3 and it pick up some crackle from the wireless router (I can only hear that when the volume is at maximum whilst the Squeezebox is paused). I turned off the router & the crackle is gone.
I suggest you to ground the input of the Fetzilla with a short piece of wire & that will help to isolate the source of the hiss.
Cheers, Stanley
Kean,
Wax vapor is highly flamable, it has a very low flash point. Not a good idea.
Ken
Thanks for the warning, now I know.
In listening tests and measured performance, for rail filtering the benefits seem to outweigh the costs. It seems our question though is which benefits correlate to the listening tests. Here my knowledge is almost worthless, so I will defer to others...
- keantoken
In listening tests and measured performance, for rail filtering the benefits seem to outweigh the costs. It seems our question though is which benefits correlate to the listening tests. Here my knowledge is almost worthless, so I will defer to others...
- keantoken
Swordfishy, I admire you efforts invested in this amp, but this kind of statement can be very misleading to many DIYers because people tend to generalize stuff.
Useful information would have to answer the questions:
which JFET, which MOSFET, which BJT was used, what was the Id (Ic), what was the Vds (Vce), what kind of freqv. compensation was used, etc...
For example, 2SK170BL used with Id=2mA sounds grainy and murky, but with Id=80% of Idss it sounds perfect. The same device can sound totally different depending on conditions of use...
Juma..
just follow the thread and you will know....Bypassing it to the end is not at way to get the picture..Sort of like reading at book backwards....you miss the points...🙂)
just follow the thread and you will know....Bypassing it to the end is not at way to get the picture..Sort of like reading at book backwards....you miss the points...🙂)
It's probably a mix of actual audio band noise + some demodulated products from higher frequency RF noise.
. . . and it may be that your ears are a valid way of judging noise levels
It's just that my goal is getting the system to sound great so personally I assess noise levels more indirectly by listening for improved clarity, smoothness & sound staging etc and as I alluded to before I have noticed a clear correlation between reducing noise and these qualities being enhanced.
It may be that a good scope could also measure / assess noise but my digital scope is too noisy itself to see a meaningful difference
cheers
mike
Last edited:
Juma,
I was wondering when you would post - welcome to our thread!
You are right, however, ALL operating parameters are important, and your 80% Idss comment is right on the money. That is the reason we specified the GR version of the input jfet, so as to run it close to the 2.5-6mA Idss rating.
Nevertheless, the connection between formal measurements and sonics is so tenuous that a spectrum of (often strong) opinions is inevitable. I am not sure that SWF should be upbraided for saying what he said. Actually, I have found that just about any topology can sound good, if sufficient work is put into the component selection and operating points. The problem is, it seems at times like pinning the tail on the donkey. For example, I'm of the firm belief that too much global feedback does not improve the sound, and that a good figure is somewhere between 20-30dB, no more. I have no idea why; it's just a result of years of experimentation.
What causes noise? SWF, I have found that LTP input stages are very susceptible to noise, but singletons, with emitter/source current taken from the output rail like your Fetzilla, make very little noise. I'm sure there are good reasons for this, perhaps the fact that the first stage current is taken from the output, rather than from a rail.
Mike,
How are you getting on with your DC coupled approach? Are you pleased with the sound quality, and by any chance have you compared it with the AC coupled version here?
Cheers,
Hugh
I was wondering when you would post - welcome to our thread!
You are right, however, ALL operating parameters are important, and your 80% Idss comment is right on the money. That is the reason we specified the GR version of the input jfet, so as to run it close to the 2.5-6mA Idss rating.
Nevertheless, the connection between formal measurements and sonics is so tenuous that a spectrum of (often strong) opinions is inevitable. I am not sure that SWF should be upbraided for saying what he said. Actually, I have found that just about any topology can sound good, if sufficient work is put into the component selection and operating points. The problem is, it seems at times like pinning the tail on the donkey. For example, I'm of the firm belief that too much global feedback does not improve the sound, and that a good figure is somewhere between 20-30dB, no more. I have no idea why; it's just a result of years of experimentation.
What causes noise? SWF, I have found that LTP input stages are very susceptible to noise, but singletons, with emitter/source current taken from the output rail like your Fetzilla, make very little noise. I'm sure there are good reasons for this, perhaps the fact that the first stage current is taken from the output, rather than from a rail.
Mike,
How are you getting on with your DC coupled approach? Are you pleased with the sound quality, and by any chance have you compared it with the AC coupled version here?
Cheers,
Hugh
Last edited:
Hello
Most of the times, hiss noise are resulting from transistors or resistors thermal noises in the input stage but it could be the current consumption in the first stage transistors, check the input stage for any defect transistors or resistors or maby something causing wrong current consumption in the input stage.
Bye
Gaetan
Last edited:
What methods do you typically use to decrease noise?
I have a Tektronix 561B scope, with the 3A9 differential plugin. I can see signals down to 10uV. It has built-in LP and HP filters, and a BW of 1MHz. Unforuntately it is difficult to make good measurements because my signal generator is broken, and the main ripple is larger than any HF signals. I just remembered, I have some 24V switching supplies, I'll have to see how easy it is to wire them up.
- keantoken
- Status
- Not open for further replies.