Alright, before I was complaining about stereo problems, but it's not so bad now. I suspect the gate capacitance was giving hell in the treble...
Before it was somewhat painful to listen to treble, I equalized it out. But now it seems apparent that the treble was being distorted heavily. This is probably why it sounded so different...
Alright, it's not so sharp anymore. It's definitely softened up and the measured characteristics are better. Not so imposing anymore, and not alien sounding to everything else I've heard. However resolution is still good and I'll even risk to describe the sound as "warm". Steely Dan's sounding good...
Noise is also less, and more tolerable. Noise was sharp before, and as I mentioned, hum has decreased. Apparently poor treble leads to poor noise, indirectly. I can hardly believe that the scope shows the signal buried in noise, but I don't hear any of it. Just noise. No obvious oscillation (save for the weirdness caused by sampling trash); it all comes from my soundcard.
Man, I must really have a ways to go before I can tell what sounds "right".
But I digress; this sounds pretty good.
- keantoken
Before it was somewhat painful to listen to treble, I equalized it out. But now it seems apparent that the treble was being distorted heavily. This is probably why it sounded so different...
Alright, it's not so sharp anymore. It's definitely softened up and the measured characteristics are better. Not so imposing anymore, and not alien sounding to everything else I've heard. However resolution is still good and I'll even risk to describe the sound as "warm". Steely Dan's sounding good...
Noise is also less, and more tolerable. Noise was sharp before, and as I mentioned, hum has decreased. Apparently poor treble leads to poor noise, indirectly. I can hardly believe that the scope shows the signal buried in noise, but I don't hear any of it. Just noise. No obvious oscillation (save for the weirdness caused by sampling trash); it all comes from my soundcard.
Man, I must really have a ways to go before I can tell what sounds "right".
But I digress; this sounds pretty good.
- keantoken
This is the part I don't understand. Maybe you can show what it looks like with a simple 1kHz sine wave at the input.
High-BW scope reads interesting pulses at 55mV pk-pk. Touching critical points doesn't alter it's frequency or it's amplitude.
Turning off monitor removes it completely!
After that, there is a very low-level signal that looks much the same, I'm thinking it might be a TV in another room. Switching on the scope caused jump in output voltage which seems to confirm my theory about an appliance (probably thermostat of fridge or AC).
- keantoken
Turning off monitor removes it completely!
After that, there is a very low-level signal that looks much the same, I'm thinking it might be a TV in another room. Switching on the scope caused jump in output voltage which seems to confirm my theory about an appliance (probably thermostat of fridge or AC).
- keantoken
Last edited:
Wow, 55mV p-p isn't tiny. Could be coming up your ground? Have you seen the common mode choke in the low noise preamp thread that I tried? Helped a lot with nasty noise similar to what you're seeing. I got a number of these common mode chokes out of old CRT monitors that people seem to throw out these days.
Edit: can you tell (roughly) the frequency of the pulses?
Edit: can you tell (roughly) the frequency of the pulses?
They are around 50KHz. They move down to 28mV after I connect the the scope ground... Remember, my bench is 2 feet away from my monitor! And this is a good monitor, with a nice strong flyback... Out of all the monitors I have owned, this one has given me the most interference.
Here is a shot of 1KHz at normal listening volume. As I discovered with the filtering knobs, most of it is ultrasonic, above 20KHz. All of this comes from my soundcard. After turning volume all the way down and turning off monitor, my low-BW scope shows .5mV of noise (excluding ripple), the other one shows 5mV...
The noise is just noise... Nothing I can get a trigger on, no obvious distinct frequencies...
- keantoken
Here is a shot of 1KHz at normal listening volume. As I discovered with the filtering knobs, most of it is ultrasonic, above 20KHz. All of this comes from my soundcard. After turning volume all the way down and turning off monitor, my low-BW scope shows .5mV of noise (excluding ripple), the other one shows 5mV...
The noise is just noise... Nothing I can get a trigger on, no obvious distinct frequencies...
- keantoken
Attachments
Last edited:
Here is why I chose the alternate compensation method...
I think Cdom runs counter to amp stability. The amp's OLG depends on input transconductance - CURRENT gain. Cdom does not respond if the output voltage does not change, as is the case with a capacitive load - hence the indomitable instability into capacitive loads.
However, if we use current-type compensation that responds to current draw instead of voltage, we avoid this problem. We also boost BW incredibly. The reason it takes such a large Cdom to achieve the same stability is because very large output currents must flow before Cdom can respond adequately to a capacitive load. In effect, Cdom is a very bad current-type compensation since stability into capacitive loads depends on the voltage drop of the next stage.
You will notice that phase must take a large dip in order to achieve such a steep rolloff. To me this indicates compensation is being applied only where necessary.
Look at the OLG plots below for comparison. The BA15218N is the opamp found in one of my junk amps. The SPICE plot is from the headamp.
This is all still a theory anyways. More testing will reveal it's practicality. But obviously it works as an alternative, since my amp works.
In simulation, after removing the MOSFET, I found I couldn't get stability back where it was. I solved this with a 10n+10R cap to ground at the output. One of the troubles with this type of compensation is that it takes more caps, and larger ones. The reason it's not in use may simply be because large caps like this can't be fabricated practically onto a chip. However this doesn't bar discrete design, and if so it may be possible to make a discrete opamp that bests the best of chips.
With the alternate compensation, the headamp is capable of amplifying ultrasonic signals with high linearity.
- keantoken
I think Cdom runs counter to amp stability. The amp's OLG depends on input transconductance - CURRENT gain. Cdom does not respond if the output voltage does not change, as is the case with a capacitive load - hence the indomitable instability into capacitive loads.
However, if we use current-type compensation that responds to current draw instead of voltage, we avoid this problem. We also boost BW incredibly. The reason it takes such a large Cdom to achieve the same stability is because very large output currents must flow before Cdom can respond adequately to a capacitive load. In effect, Cdom is a very bad current-type compensation since stability into capacitive loads depends on the voltage drop of the next stage.
You will notice that phase must take a large dip in order to achieve such a steep rolloff. To me this indicates compensation is being applied only where necessary.
Look at the OLG plots below for comparison. The BA15218N is the opamp found in one of my junk amps. The SPICE plot is from the headamp.
This is all still a theory anyways. More testing will reveal it's practicality. But obviously it works as an alternative, since my amp works.
In simulation, after removing the MOSFET, I found I couldn't get stability back where it was. I solved this with a 10n+10R cap to ground at the output. One of the troubles with this type of compensation is that it takes more caps, and larger ones. The reason it's not in use may simply be because large caps like this can't be fabricated practically onto a chip. However this doesn't bar discrete design, and if so it may be possible to make a discrete opamp that bests the best of chips.
With the alternate compensation, the headamp is capable of amplifying ultrasonic signals with high linearity.
- keantoken
Attachments
Okay, things just got REALLY interesting.
My observations for the headphones parallel my observations with speakers.
However there is a reason you don't want to use low loads with this amp without a buffer. The plot shows the output transistor's Vbe nonlinearity being injected through C5. An interesting side-effect of the different compensation.
The distortion created, since it is derived from a diode curve, is purely monotonic. Not the worst type of distortion, but we want to avoid a lot of it. A loudspeaker usually will not have 4 ohm (the current load) impedance at 5KHz so we're protected from the worst of it (for example my Betsyk's have 14 ohms). At frequencies lower than 1k there's not much HF for the cap to pick up. With a higher load, this will not happen. Since the compensation is current-type, a higher load means less HF distortion, as opposed to Cdom compensation, where no matter the load, HF distorts equally.
- keantoken
My observations for the headphones parallel my observations with speakers.
However there is a reason you don't want to use low loads with this amp without a buffer. The plot shows the output transistor's Vbe nonlinearity being injected through C5. An interesting side-effect of the different compensation.
The distortion created, since it is derived from a diode curve, is purely monotonic. Not the worst type of distortion, but we want to avoid a lot of it. A loudspeaker usually will not have 4 ohm (the current load) impedance at 5KHz so we're protected from the worst of it (for example my Betsyk's have 14 ohms). At frequencies lower than 1k there's not much HF for the cap to pick up. With a higher load, this will not happen. Since the compensation is current-type, a higher load means less HF distortion, as opposed to Cdom compensation, where no matter the load, HF distorts equally.
- keantoken
Attachments
I've plugged in a 16 ohm speaker and have to come to terms with the fact that the best treble I've heard comes from this 2.5" speaker ripped out of a cheap plastic CD player...
All sound through this speaker is soft, but never lacking in treble... And there's probably no way I can get another one...
Here is a noise plot, FWIW. I've got not clue what this means, or if the simulation was correct. All BJT models modeled Rb, but it was always 10 ohms.
- keantoken
All sound through this speaker is soft, but never lacking in treble... And there's probably no way I can get another one...
Here is a noise plot, FWIW. I've got not clue what this means, or if the simulation was correct. All BJT models modeled Rb, but it was always 10 ohms.
- keantoken
Attachments
Hey Iko, I've got a number of those small chokes but I couldn't find the place in your thread where it is discussed.
The cord stretches about 2.75 meters, right behind my monitor, and it's one of those spiral cords so... I guess I probably should begin worrying about ultrasonic noise now.
- keantoken
The cord stretches about 2.75 meters, right behind my monitor, and it's one of those spiral cords so... I guess I probably should begin worrying about ultrasonic noise now.
- keantoken
Just like in this schematic ground goes through one of the coils and Vcc goes through the other.
http://www.diyaudio.com/forums/soli...e-low-noise-amplifier-lna-35.html#post2215018
I'll try to take a picture of it for you so you know what to look for. It looks like a little transformer. I measured several from 20mH to 30mH, and used those of highest inductance. I got mine from the power supply section of old CRT monitors.
Yes, be very, very, afraid!
Not to worry Iko, my amp reproduces ultrasonic noise with utmost accuracy!
Sound that martians would appreciate...
The noise isn't coming through the power supply though, it's coming from the cord and soundcard to my amp. Do you think the same configuration works on the input? Not sure how this would affect sound quality.
- keantoken
Sound that martians would appreciate...
The noise isn't coming through the power supply though, it's coming from the cord and soundcard to my amp. Do you think the same configuration works on the input? Not sure how this would affect sound quality.
- keantoken
Might. There's something else that might help you, something I haven't implemented yet, but it's in the "to do in case I need it" folder:
http://electronicdesign.com/article...-This-Noise-Rejection-Stuff-Anyhow-21808.aspx
http://electronicdesign.com/article...-This-Noise-Rejection-Stuff-Anyhow-21808.aspx
Just plugged in a random CM choke, it helped a little but there are still some transients.
More proof that my soundcard really sucks: switching to 44.1KHz (from 96KHz) decreases noise frequency considerably. You guessed it, all of it is sampling noise. Then when I switch to 96KHz again, noise is even lower! But restart the computer, and it's noisy as hell again. Why can't my soundcard make up it's mind?I'll try installing 3rd party drivers, again...
That circuit is interesting. There is also the possibility of transformer coupling as in this article:
http://www.diyaudio.com/forums/diya...ponent-grounding-interconnection.html?garpg=4
- keantoken
More proof that my soundcard really sucks: switching to 44.1KHz (from 96KHz) decreases noise frequency considerably. You guessed it, all of it is sampling noise. Then when I switch to 96KHz again, noise is even lower! But restart the computer, and it's noisy as hell again. Why can't my soundcard make up it's mind?I'll try installing 3rd party drivers, again...
That circuit is interesting. There is also the possibility of transformer coupling as in this article:
http://www.diyaudio.com/forums/diya...ponent-grounding-interconnection.html?garpg=4
- keantoken
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.