So i figured i try to find help here before i throw in the towel..
The basic idea is a mix between:
1. Supertriode Chip Amp found on tubecad.com here:
More Super-Triode Amplifiers
2. Mixed-mode Feedback found on sound.westhost.com here:
Variable Amplifier Impedance
In other words, i am trying to combine this:
with this:
The goal is to have a chipamp that has a Triode signature AND the damping factor of a typical SE Triode (say Zo of 3-4 ohm).. best case would be a chipamp that passes a blind test as a SE Triode amp, with a bonus of adjustable damping factor.
Here is what i end up with in LTSpice:
The idea is as follows:
1. The triode on the inverting input attenuates the output signal by a factor of mu and feeds it back to the inverting input. The grid of the same triode receives feedback from the 0R22 resistor (R7) which will then reduce the damping factor.
2. The triode on the non-inverting input is simply a cathode follower that receives same plate supply as the triode on the inverting input. This kills two birds with one stone: buffering the signal input to the chipamp and improving PSRR by allowing the chipamp's CMRR to iron-out any PS ripple as we are feeding the same ripple to both inverting and non-inverting.
3. Input signal is fed to the cathode follower (triode on the non-inverting)
Here is the actual schematic, Left channel shown. Chip being used is TDA7265. 10mA CCS is formed by the LM317s and 120R resistors.
and here is the PCB design..
I should add that the High Voltage 150V supply is generated from a 555-based boost converter, not unlike the typical nixie power supplies (12V in --> 150V out, same 12V as being used to power the heaters). This is formed by U6 and the parts around it found on the PCB design above.
The problem:
It sounds very bad. Best way to describe it is like there are rice grains on top of the speaker cones. Imagine poorly tuning to an FM radio station and you'll get the same "grainy" noise. Cheap oscilloscope trace shows the output has spikes, even if it was being fed a sinewave.
What i have done so far:
1. Generate the 150V supply from a mains transformer, bypassing the 555 boost converter (i remove the 555 from the socket).
2. Generate the bipolar 24V supply for the TDA7265 also from a mains transformer. Previously i use an SMPS module that converts 12V to +/-24V.
Both actions above were based on the assumption that SMPS are causing the spikes on the output signal).
3. Bypass the 0R22 resistor (so it's only Supertriode mode, not mixed mode feedback).
What i suspect is causing the problem:
1. Too low closed loop gain causing oscillation. TDA7265 must have more than 25dB of closed loop gain to be stable. The tube i use have mu of around 33. This translates to around 30dB of closed loop gain (with 0R22 resistor bypassed).. so i don't think it's this.
2. Poor PCB layout causing oscillation?
3. The TDA7265 is simply a poor choice for this application for unknown reason.
One more hint: the heatsink is rather hot even with no input signal. Are grainy sound and hot heatsink signs of oscillation? What can i do?
The basic idea is a mix between:
1. Supertriode Chip Amp found on tubecad.com here:
More Super-Triode Amplifiers
2. Mixed-mode Feedback found on sound.westhost.com here:
Variable Amplifier Impedance
In other words, i am trying to combine this:
An externally hosted image should be here but it was not working when we last tested it.
with this:
An externally hosted image should be here but it was not working when we last tested it.
The goal is to have a chipamp that has a Triode signature AND the damping factor of a typical SE Triode (say Zo of 3-4 ohm).. best case would be a chipamp that passes a blind test as a SE Triode amp, with a bonus of adjustable damping factor.
Here is what i end up with in LTSpice:
An externally hosted image should be here but it was not working when we last tested it.
The idea is as follows:
1. The triode on the inverting input attenuates the output signal by a factor of mu and feeds it back to the inverting input. The grid of the same triode receives feedback from the 0R22 resistor (R7) which will then reduce the damping factor.
2. The triode on the non-inverting input is simply a cathode follower that receives same plate supply as the triode on the inverting input. This kills two birds with one stone: buffering the signal input to the chipamp and improving PSRR by allowing the chipamp's CMRR to iron-out any PS ripple as we are feeding the same ripple to both inverting and non-inverting.
3. Input signal is fed to the cathode follower (triode on the non-inverting)
Here is the actual schematic, Left channel shown. Chip being used is TDA7265. 10mA CCS is formed by the LM317s and 120R resistors.
An externally hosted image should be here but it was not working when we last tested it.
and here is the PCB design..
An externally hosted image should be here but it was not working when we last tested it.
I should add that the High Voltage 150V supply is generated from a 555-based boost converter, not unlike the typical nixie power supplies (12V in --> 150V out, same 12V as being used to power the heaters). This is formed by U6 and the parts around it found on the PCB design above.
The problem:
It sounds very bad. Best way to describe it is like there are rice grains on top of the speaker cones. Imagine poorly tuning to an FM radio station and you'll get the same "grainy" noise. Cheap oscilloscope trace shows the output has spikes, even if it was being fed a sinewave.
What i have done so far:
1. Generate the 150V supply from a mains transformer, bypassing the 555 boost converter (i remove the 555 from the socket).
2. Generate the bipolar 24V supply for the TDA7265 also from a mains transformer. Previously i use an SMPS module that converts 12V to +/-24V.
Both actions above were based on the assumption that SMPS are causing the spikes on the output signal).
3. Bypass the 0R22 resistor (so it's only Supertriode mode, not mixed mode feedback).
What i suspect is causing the problem:
1. Too low closed loop gain causing oscillation. TDA7265 must have more than 25dB of closed loop gain to be stable. The tube i use have mu of around 33. This translates to around 30dB of closed loop gain (with 0R22 resistor bypassed).. so i don't think it's this.
2. Poor PCB layout causing oscillation?
3. The TDA7265 is simply a poor choice for this application for unknown reason.
One more hint: the heatsink is rather hot even with no input signal. Are grainy sound and hot heatsink signs of oscillation? What can i do?
The best you can do is to look at the output with an oscilloscope. That'll probably tell you a lot about what's going on.
"Grainy sound" and "rice on the speaker cone" could be anything, including a bad solder joint.
If you want your chip amp to sound like a tube amp, I suggest digging up some chip amp based guitar amps, such as those by Peavy. They use various tube emulation circuits and get surprisingly close to tube sound.
Tom
"Grainy sound" and "rice on the speaker cone" could be anything, including a bad solder joint.
If you want your chip amp to sound like a tube amp, I suggest digging up some chip amp based guitar amps, such as those by Peavy. They use various tube emulation circuits and get surprisingly close to tube sound.
Tom
Hi,
What you are trying to do with the amplifier is way unstable.
It needs to be unity gain stable. Valve gain is irrelevant, or
more to the point it is the gain, not the amplifier, which
is effectively a buffer. With 30dB valve gain the amplifier
will only be stable with over 55dB of total gain, ridiculous.
rgds, sreten.
I really don't understand the point of adding another triode.
Looks simple enough to me noting the amplifier must
be unity gain stable (which of course most are not).
I can't see any problem adding variable impedance.
Though the cap from B+ to output labelled 0V is wrong.
What you are trying to do with the amplifier is way unstable.
It needs to be unity gain stable. Valve gain is irrelevant, or
more to the point it is the gain, not the amplifier, which
is effectively a buffer. With 30dB valve gain the amplifier
will only be stable with over 55dB of total gain, ridiculous.
rgds, sreten.
I really don't understand the point of adding another triode.
An externally hosted image should be here but it was not working when we last tested it.
Looks simple enough to me noting the amplifier must
be unity gain stable (which of course most are not).
I can't see any problem adding variable impedance.
Though the cap from B+ to output labelled 0V is wrong.
Last edited:
I am going to stay with this design until i make it work, even if i have to re-do everything with different chip. I suspect i need a unity gain stable chip to do this.
I did look at the output trace using a scope, albeit a cheap pocket one. As mentioned, feeding a sinewave results in the output being a sinewave with spikes. Indication of oscillation?
I question the Ebay 6n23p tube, whether they are new. Maybe they are not new and mu has reduced to something that caused the closed loop gain to be less than 25dB and, subsequently, oscillation.
I did look at the output trace using a scope, albeit a cheap pocket one. As mentioned, feeding a sinewave results in the output being a sinewave with spikes. Indication of oscillation?
I question the Ebay 6n23p tube, whether they are new. Maybe they are not new and mu has reduced to something that caused the closed loop gain to be less than 25dB and, subsequently, oscillation.
Don't forget that adding that tube V2 in the negative feedback loop is going to cause some bandwidth limits and some more delay within it. You'll need to cut the loop feedback considerably, would be my guess. Try bumping up R5 as a test to see if that makes it stable.
I'm not familiar with that particular chipamp, but doesn't it need a resistor from the [+] input terminal to somewhere, to establish a voltage bias point?
BTW, there are easier, more stable, ways to reduce damping factor. Like just adding a series resistor in line with the speaker output.
I'm not familiar with that particular chipamp, but doesn't it need a resistor from the [+] input terminal to somewhere, to establish a voltage bias point?
BTW, there are easier, more stable, ways to reduce damping factor. Like just adding a series resistor in line with the speaker output.
Sreten,
I am on the same page suspecting the need for unity gain stable chip.. but before that:
How is the chip effectively a buffer? I am feeding the inverting input with signal output divided by mu (33). So it is like a feedback of 30dB. LTSpice sim also shows the output is approximately 33 times the input signal in amplitude.
I am on the same page suspecting the need for unity gain stable chip.. but before that:
How is the chip effectively a buffer? I am feeding the inverting input with signal output divided by mu (33). So it is like a feedback of 30dB. LTSpice sim also shows the output is approximately 33 times the input signal in amplitude.
Hi bwaslo,
I will try increasing R5 but i am worried it might increase the output offset as TDA7265 has 20k internal resistor across the non inverting input to ground (this answers your question as well).
Increasing DF is just a bonus from utilizing unused grid. Main idea is to get the triode signature first. Besides, adding 3-4 ogm power resistor on the output wastes too much heat and output power.
I will try increasing R5 but i am worried it might increase the output offset as TDA7265 has 20k internal resistor across the non inverting input to ground (this answers your question as well).
Increasing DF is just a bonus from utilizing unused grid. Main idea is to get the triode signature first. Besides, adding 3-4 ogm power resistor on the output wastes too much heat and output power.
Last edited:
In that case, try hanging a series RC to ground from the [-]input feedback point. Something like 5k in series with 50uF or so. Not as a fix, but to see if that's the problem.
I soldered 4K7 with 100uF between inverting input and ground.. unfortunately, i failed to recreate the grainy sound to begin with. There was simply sound coming from the speaker now. Note that there was no sound even before i soldered the 4K7+100uF.
It was my attempt to power up the amp again after yesterday. Yesterday it sounded grainy but now it doesn't even make any sound. Interestingly, if i touch my finger to the input jack, it still makes loud buzzing sound. But if i plug the input jack to a source (cellphone and laptop), no sound coming out.
The heatsink is still quite hot even with no sound from the speakers. I can hold my finger on it for 5 seconds. I traced the input signal with my cheap pocket scope and sure enough the chip got the signal to the non-inverting pin but there was no sound.
I have to add that it was hit and miss as well for the last few attempts. Sometime i have to reset the +/-24V power supply to get that grainy sound because it won't make any sound at all.
Perhaps now the chip is dead after suffering what presumably is crazy oscillation?
Anyway, can anyone explain to me whether unity gain chip is necessary for this design? Again, i think have to explain how the Supertriode thing works before you answer.. The output signal is fed to the triode plate via 10uF coupling cap. With the plate moving with the output signal, cathode will output the same signal but now it's divided by a factor of mu. This gets fed to the inverting input. So, i think the chip is not working as a buffer. It's simply working with negative feedback of mu, which is 30dB with 6N23P tube.
Not only have you added another pole inside of the feedback loop, but apparently you've done it in a haphazard fashion. You need to figure out the Bode plot with the tubes in the circuit.
You can improve stability margin a lot by connecting a resistor in series with a capacitor across the chip's inputs. I forget how you calculate this, but basically you are scrubbing off loop gain at higher frequencies where there's too much phase shift. This is how Tom's (tomchr) current pump works at unity gain, so hopefully he'll chime in.
You can improve stability margin a lot by connecting a resistor in series with a capacitor across the chip's inputs. I forget how you calculate this, but basically you are scrubbing off loop gain at higher frequencies where there's too much phase shift. This is how Tom's (tomchr) current pump works at unity gain, so hopefully he'll chime in.
Thanks Fast Eddie. Hopefully tom will chime-in.
Hi jerluwoo,
Yes i did use the non-inverting pin for signal input. I did not follow exactly the tubecad way of supertriode connection.
Essentially, my schematic is the same as yours. The only difference is that i put a cathode follower before the input signal goes to the non-inverting pin.
The grid on the triode at the inverting pin is not used for signal input. Instead, it's used as feedback from the output as well.
Sorry if my previous schematic is confusing to read.
Here it is again.
You can see U2 is the triode that goes to inverting pin which is part of the feedback. U2's grid is connected to 0R22 resistor on the output, essentially forming the mixed feedback mode.
U1 is the cathode follower that goes to non-inverting pin. U1 grid is for signal input.
So, would you say that essentially i should not have any oscillation problem with regards to the design? i.e I just need to ensure the closed loop gain is > 25dB which is achieved by using triode of mu=33 for U2?
Hi jerluwoo,
Yes i did use the non-inverting pin for signal input. I did not follow exactly the tubecad way of supertriode connection.
Essentially, my schematic is the same as yours. The only difference is that i put a cathode follower before the input signal goes to the non-inverting pin.
The grid on the triode at the inverting pin is not used for signal input. Instead, it's used as feedback from the output as well.
Sorry if my previous schematic is confusing to read.
Here it is again.
An externally hosted image should be here but it was not working when we last tested it.
You can see U2 is the triode that goes to inverting pin which is part of the feedback. U2's grid is connected to 0R22 resistor on the output, essentially forming the mixed feedback mode.
U1 is the cathode follower that goes to non-inverting pin. U1 grid is for signal input.
So, would you say that essentially i should not have any oscillation problem with regards to the design? i.e I just need to ensure the closed loop gain is > 25dB which is achieved by using triode of mu=33 for U2?
Last edited:
I hope he does. This is a composite amplifier and Tom knows his stuff about composite amplifiers.
The golden role with composite amps (just like discrete designs) is for the output stage to be faster than the input stage. Of course this is often not practical so you have to force your input stage to slow down with either Miller capacitors or local shunts. Otherwise, your amp will be underdamped and potentially unstable.
My only experience with composite amplifiers involves super high speed buffers that are designed for composite amplifier circuits. This avoids the whole issue you are facing.
Yes in general the amp will need to produce mu gain, but, your grid connection constitutes positive feedback to the triode, so the amp may not be producing mu gain or you have designed an elaborate oscillator based on the tubes internal capacitance. Try simply grounding the grid instead.
Last edited:
Thank you!
This weekend, I will try 1K in series with C2 (to ensure the TDA7265 has more gain than mu) and 47pF parallel with R2 for a cutoff freq of 33.9kHz.
Do tell me if these values are not reasonable.
I am just hoping the TDA7265 isn't dead yet
I will also try swapping the 6N23P, perhaps the mu of the current triodes has dropped.
This weekend, I will try 1K in series with C2 (to ensure the TDA7265 has more gain than mu) and 47pF parallel with R2 for a cutoff freq of 33.9kHz.
Do tell me if these values are not reasonable.
I am just hoping the TDA7265 isn't dead yet
I will also try swapping the 6N23P, perhaps the mu of the current triodes has dropped.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.