Simulation/modeling is constrained by the completeness and/or accuracy of component models; often times, variances from reality will be observed. The Atma-Sphere MA-1 Mk II.2 Mono Amplifier (with 14 6AS7G output and four 6SN7GT front-end/driver tubes per amplifier) has a measured output impedance (measured by an injection of a constant 1A of current at 50Hz) of 10.5 ohms.
Here's the URL for the SoundStage measurements on the MA1 Mk. II.2:
SoundStage! Measurements - Atma-Sphere MA-1 Mk II.2 Mono Amplifiers: Measurements (2/2002)
Measured power output with 1kHz test signal:
* 8-ohm load at 1% THD: 20W
* 8-ohm load at 10% THD: 120W
* 4-ohm load at 1% THD: 8W
* 4-ohm load at 10% THD: 80W
* 16-ohm load at 1% THD: 90W
* 16-ohm load at 10% THD: 130W
Of course, these results are using Ralph's original-spec 6AS7 output tubes... 😎
Here's the URL for the SoundStage measurements on the MA1 Mk. II.2:
SoundStage! Measurements - Atma-Sphere MA-1 Mk II.2 Mono Amplifiers: Measurements (2/2002)
Measured power output with 1kHz test signal:
* 8-ohm load at 1% THD: 20W
* 8-ohm load at 10% THD: 120W
* 4-ohm load at 1% THD: 8W
* 4-ohm load at 10% THD: 80W
* 16-ohm load at 1% THD: 90W
* 16-ohm load at 10% THD: 130W
Of course, these results are using Ralph's original-spec 6AS7 output tubes... 😎
The plate resistance of the 6AS7G is about 150 ohms. Fourteen 6AS7Gs in parallel calculate to 10.7 ohms. This is awfully close to the 10.5 ohms measured.
MOST IMPORTANTLY, the power figures quoted above for the Atma-Sphere amps are PEAK; they are not average or RMS.
I can get 70 volts peak to peak using two 150 volt supplies. That equates to 35 volts peak which is (35*35)/8 = 153 watts peak power with 12 output tubes. However, for a sinusoidal signal, that's 76 watts RMS. that's still a lot of power, but not as much as I hoped for using 14 tubes.
Notice that the power of the MA-1 goes up a lot when going from 4 to 8 ohm loads, but stays almost the same for 16 ohm loads as for 8 ohm loads. The amp is running out of voltage. It simply can't pump the same current through the 16 ohm load because there isn't enough voltage.
The most important thing I've learned is that ideally a circlotron OTL needs to be matched to the load for best performance, and having lots of output tubes will not give higher power into high impedance loads without adequate voltage drive. If you parallel lots of low impedance output tubes the voltage drive required from the supply is lessened, but this gets very large and costly to get the impedance of the tubes as low as (or lower than) the load. If you want to know the theoretical maximum power using an infinite number of tubes simply assume zero impedance for the tubes with the signal swinging to the supply rails across the load. The power can never exceed this value even with a bank of a million ideal zero resistance tubes.
If you want 250 watts RMS out of an 8 ohm speaker the FIRST thing you need are supply voltages that can support the power, and the second thing you need is a low enough parallel resistance on the output tubes so that the volt drop across the tubes doesn't dissipate most of the power. The "perfect" OTL tube would combine high voltage capability with low resistance. Unfortunately those two specs are difficult to optimize simultaneously in a tube design.
Let me rephrase that. It clips flat, as opposed to some very strange things I've seen with other designs I've tried. "Gracefully" was too kind of a word. "Controlled" would be more accurate.
Question
According to the datasheet the 6H13C has a maximum anode voltage of 250V, although the tube is typically run with 135-150 volt supplies.
Has anyone ever run this tube at a higher voltage? The spec even says 500 volts is acceptable when the tube is non-conducting. Raising the supply voltage would help increase the power substantially, especially into high impedance loads. I'd like to try 220 volts on each supply but I don't know if the tubes can realistically handle it. Obviously the bias current would be reduced to keep the plate dissipation within acceptable limits on each tube. It boils down to the voltage withstand of the tubes. If I can run the supplies at 220 volts I can get a lot more power out.
According to the datasheet the 6H13C has a maximum anode voltage of 250V, although the tube is typically run with 135-150 volt supplies.
Has anyone ever run this tube at a higher voltage? The spec even says 500 volts is acceptable when the tube is non-conducting. Raising the supply voltage would help increase the power substantially, especially into high impedance loads. I'd like to try 220 volts on each supply but I don't know if the tubes can realistically handle it. Obviously the bias current would be reduced to keep the plate dissipation within acceptable limits on each tube. It boils down to the voltage withstand of the tubes. If I can run the supplies at 220 volts I can get a lot more power out.
I would like to match the amp to my speakers. My speakers are 3 way JBL studio monitors. I have the impedance/frequency curve and IMO it's acceptable as long as the output impedance of the amp is 2 ohms or less. This means I'll need to use NFB, more than is shown in the schematic.
The efficiency is 90dB 1W/1M, and the power rating is 125 watts continuous RMS. The speakers are capable of deep powerful bass when driven with the right amp. I'm getting about 150 watts out of the simulation (75 watts RMS), or about half the rating of the speaker. That's close enough for me. Doubling the power will only buy me 3dB, so I'm good with 75 watts RMS. I'll end up leaving the supply voltages wherever they end up. From the trannys I have I'm guessing I'll have about 160 volts on each floating supply. That might push me over 80 watts RMS.
We have an amplifier called the MA-3. I have measured over 500 watts RMS at the output of that amp. The B+ runs about 150V.
The Soundstage! amplifier sample was tested with insufficient line voltage. Bascombe King used a variac when testing the amp on the bench, and did not pay attention to the fact that as the amp reached the higher output levels, the line voltage was sagging! Once that was corrected the amp made the correct 140 watt power level.
You can run higher B+; I have a 4-tube guitar amp that runs 190V.
So we are not talking about peak power, we are talking RMS. If the tubes lack transconductance, if the AC line voltage is low, if the power supply sags, all these and more can rob the amp of power.
The Soundstage! amplifier sample was tested with insufficient line voltage. Bascombe King used a variac when testing the amp on the bench, and did not pay attention to the fact that as the amp reached the higher output levels, the line voltage was sagging! Once that was corrected the amp made the correct 140 watt power level.
You can run higher B+; I have a 4-tube guitar amp that runs 190V.
So we are not talking about peak power, we are talking RMS. If the tubes lack transconductance, if the AC line voltage is low, if the power supply sags, all these and more can rob the amp of power.
I'm with SemperFi; stop worrying about the damned simulations --- just build the amps and enjoy! 😎
It became apparent that something is wrong with the simulation. Then I did some testing of the model and it's not even close to the specs of the tube. Arrgghh!
Lesson learned for trusting models found online. It was relatively simple to set up a test circuit to "trace" the simulated curves and after a few points it was obvious the model is wrong and behaves like a tube with a much higher plate resistance.
Actually I'm quite happy - now I can build my circlotron knowing I'll get a lot more power than I anticipated!
I didn't mean to mis-state the power of the Atma-Sphere amps. I was coming up with the same numbers for power coincidentally but mine were peak and theirs were RMS. They should state that in the specs IMHO. Sorry Atma-Sphere.
This is the first time I've gotten a bad model. From now on I'll trace a curve to be sure the model is at least close before I use it.
As soon as I can find my drill the grunt work begins on the chassis, although I do intend to search for an accurate model. If I find one it can be dropped into the circuit in no time. If not, even a model of a similar tube would be sufficient for analytical purposes.
Last edited:
Need help with hand calcs, not simulation
After calculating and recalculating again by hand the maximum power output I'm getting the same answer as the simulation I'm now running.
I found a model of the 6H13C that is accurate. When I calculate the maximum power out I get the exact same answer as the simulation I'm running. I've concluded the power claims are either inaccurate or I'm making a mistake. Hopefully the latter, but I can't see the error of my ways yet. If someone can point out specifically where I'm making an error I would appreciate it.
The resistance of the 6AS7G in its linear region is about 150 ohms. This is determined by examining the plate current vs. voltage curves. Ten parallel 6AS7Gs would reduce to 15 ohms.
It seems to me the peak voltage and power that the load will ever see happens when one of the parallel bank of triodes is cutoff and the other is fully conducting. If both sets of triodes conduct it can only reduce the voltage across the load.
The figures below labeled A, B and C show the circuit used to test the model, the circlotron output showing the parallel banks of tubes as an equivalant resistance, and the simplified circuit showing the maximum instantaneous power available using 150 volt supplies. Please ignore the typo in Fig C.
The tubes and the load form a voltage (and power) divider across the supply. The 150 volts is reduced by a factor of 8 ohms / (8 ohms + 15 ohms). This ends up allowing only 52 volts to reach the load. At 8 ohms, power is (52*52)/8 = 340 watts.
Using a square wave input an average power of very close to 340 watts is attainable, but for an undistorted sinusoid the maximum power is one half the peak or 170 watts.
Where am I going wrong?
After calculating and recalculating again by hand the maximum power output I'm getting the same answer as the simulation I'm now running.
I found a model of the 6H13C that is accurate. When I calculate the maximum power out I get the exact same answer as the simulation I'm running. I've concluded the power claims are either inaccurate or I'm making a mistake. Hopefully the latter, but I can't see the error of my ways yet. If someone can point out specifically where I'm making an error I would appreciate it.
The resistance of the 6AS7G in its linear region is about 150 ohms. This is determined by examining the plate current vs. voltage curves. Ten parallel 6AS7Gs would reduce to 15 ohms.
It seems to me the peak voltage and power that the load will ever see happens when one of the parallel bank of triodes is cutoff and the other is fully conducting. If both sets of triodes conduct it can only reduce the voltage across the load.
The figures below labeled A, B and C show the circuit used to test the model, the circlotron output showing the parallel banks of tubes as an equivalant resistance, and the simplified circuit showing the maximum instantaneous power available using 150 volt supplies. Please ignore the typo in Fig C.
The tubes and the load form a voltage (and power) divider across the supply. The 150 volts is reduced by a factor of 8 ohms / (8 ohms + 15 ohms). This ends up allowing only 52 volts to reach the load. At 8 ohms, power is (52*52)/8 = 340 watts.
Using a square wave input an average power of very close to 340 watts is attainable, but for an undistorted sinusoid the maximum power is one half the peak or 170 watts.
Where am I going wrong?
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Most CAD models for vacuum tubes fail to account for Class-A2 mode of operation. Class-A2 operation usually results in a reduction in the effective impedance of the output tube; I don't see anything in your results indicating that you're modeling this response-step in the simulation.
Class-A2 means that the grid is driven MORE POSITIVE than the cathode for part or all of the waveform. This means the grid will draw current from the cathode and heat up (the 6AS7 readily handles the power-dissipation imposed on its grid). Class-A2 also requires a special driver circuit, one that can supply power to the output-tube grid, which is provisioned in the M-60 with the 6SN7-based follower stage immediately preceding the 6AS7 output tubes. 😎
Class-A2 means that the grid is driven MORE POSITIVE than the cathode for part or all of the waveform. This means the grid will draw current from the cathode and heat up (the 6AS7 readily handles the power-dissipation imposed on its grid). Class-A2 also requires a special driver circuit, one that can supply power to the output-tube grid, which is provisioned in the M-60 with the 6SN7-based follower stage immediately preceding the 6AS7 output tubes. 😎
Better yet just buy better speakers. With better efficiency and 8 ohms or better minimum dip in impedance. There is no need to build a huge amp. Most speaker designers build lousy speakers. They just stuff drivers in a fancy cabinet. They are beasts with punishing loads and sound dead and lifeless. Since, they need hundreds of watts to get them to come to life. At only ear crunching levels. There are plenty of those speakers around. Along with big bad sounding powerful amps that you are forced to mate them too. Off course the mark up is much better on those amps. They are after all heavy and filled with parts so you can charge more. Likewise the speaker cabinet looks really nice and cost 10 times more than drivers in them. The money is not well spent on better performing drivers. But, you can charge more if it looks pretty enough. There are far more bad speaker designers than bad amp designers though.
^
|
|
Don't forget that sometimes they make the speeks so's to cover up for the deficiencies of bad amps as well.
|
|
Don't forget that sometimes they make the speeks so's to cover up for the deficiencies of bad amps as well.
This is quite true. My speakers are JBL 4410 and 4412 studio monitors and they sound excellent which is one reason I have no problem sinking some $$$, real estate, and weight into an amp that can drive them to full power (150 WRMS). The sensitivity is 90dB so they aren’t that bad, and with a low impedance amp they are extremely flat. With a higher impedance output amp I have the ability to EQ the signal with a Finite Impulse Response digital EQ and compensate for frequency response with great accuracy.
This amp is enough for me. Building an amp any more powerful would simply be a waste. Anything over 100 watts RMS of clean undistorted power will do just fine.
It's my understanding that driving the grid positive will draw grid current and also result in a sharp non-linearity and the tube will go into a mode of conduction that will draw considerably more current. But at that point is there not significant distortion introduced and wouldn’t the plate dissipation be far above the rating?
I’ve experimented with the simulation model, and the anode current continues to go up as the grid goes above 0 volts. However, with a 150 volt supply and an 8 ohm load, the plate dissipation with a zero grid voltage is 45 watts. If this is kept up for any length of time the tubes will burn up, will they not?
It appears that perhaps the amp can output large amounts of power for fast transients but can’t keep it up long term? I don’t know, I’m asking.
It’s going to take quite a bit of work to build this and before I do I’d just like to know and understand a little more about the power output. I appreciate the feedback I’ve received. I’ve built two amps so far, a 300B SE and a KT88 PP, both of which sound great. Now I’m ready for an OTL.
It appears to me that the circlotron is the best circuit. The Atma-Sphere input and driver topology work better than any other I've modeled at driving the output stage. I'm jazzed about building this amp, but the work involved in the computer simulation is easy and it hepls me to understand how the amp behaves, so I'll model and simulate it for a few days before I start drilling holes for 15 6H13C tubes per channel. That's gonna be fun.
Right!
This years hi-fi mission is to build these power amps (6 monoblocks in 2 chassis) :-D
However I have a small problem - what about the phonostage / preamp?
The completely balanced nature of these amplifiers got me thinking along these lines:
"A cartridge is a loudspeaker in reverse so if it is a good idea to be balanced for a loudspeaker why not do the same for the cartridge?"
So can someone help me and point me in the direction of decent fully balanced phono stage / preamp circuits?
I can build from a circuit diagram but I cannot design :-(
I have had a trawl through these froums but haven't had much luck in finding what I am looking for.
I am posting here as I figured that I am more likely to get a sympathetic response...
This years hi-fi mission is to build these power amps (6 monoblocks in 2 chassis) :-D
However I have a small problem - what about the phonostage / preamp?
The completely balanced nature of these amplifiers got me thinking along these lines:
"A cartridge is a loudspeaker in reverse so if it is a good idea to be balanced for a loudspeaker why not do the same for the cartridge?"
So can someone help me and point me in the direction of decent fully balanced phono stage / preamp circuits?
I can build from a circuit diagram but I cannot design :-(
I have had a trawl through these froums but haven't had much luck in finding what I am looking for.
I am posting here as I figured that I am more likely to get a sympathetic response...
If you have a low impedance driver (e.g., a sturdy cathode follower or a source follower) directly coupled to the grids and the output tubes are among those that are happy with AB2 operation (I've had good luck with 6L6GC/7027 types), then no, the distortion will not be high and if you follow the ratings in the tube spec sheets, the tubes won't be overdriven.
I need to examine the output stage biasing and look at the load lines for the operation of the tubes and put SPICE away until this makes sense. It should be a simple matter to draw the load lines and hand calculate the plate dissipation of the tubes.
The simulation appeared to take AB2 into account but I'm not certain.
Looking into it.
Thanks.
Allen Wright of Vacuum State Audio has what you need. He also publishes a Tube Preamp Cookbook you'll find helpful.
Stuart
400 WRMS!!!
I just discovered the output stage wasn't the limiting factor in my simulation, it was the driver. The driver was clipping before the output, but the grids on the output tubes went positive with respect to the cathodes over 10 volts before the driver clipped and there was no apparent distortion as the grids went positive (until the driver clipped).
It looks like the model is working in AB2 but for some reason the driver was clipping before full power was reached.
I inceased the voltage on the power supply to the driver section and now I'm getting 403 watts RMS with no clipping of the output stage using 150 volt supplies. The grids are going 15 volts positive with respect to the cathodes at the peak of the waveforms.
Awesome! This amp is scary! 😱
The driver is still a little "flat" on the bottom but its not clipping. The output shows a slight asymetry which, since its balanced, must be due to the input its getting, but otherwise it looks pretty good.
I just discovered the output stage wasn't the limiting factor in my simulation, it was the driver. The driver was clipping before the output, but the grids on the output tubes went positive with respect to the cathodes over 10 volts before the driver clipped and there was no apparent distortion as the grids went positive (until the driver clipped).
It looks like the model is working in AB2 but for some reason the driver was clipping before full power was reached.
I inceased the voltage on the power supply to the driver section and now I'm getting 403 watts RMS with no clipping of the output stage using 150 volt supplies. The grids are going 15 volts positive with respect to the cathodes at the peak of the waveforms.
Awesome! This amp is scary! 😱
The driver is still a little "flat" on the bottom but its not clipping. The output shows a slight asymetry which, since its balanced, must be due to the input its getting, but otherwise it looks pretty good.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
To paraphrase that great vacuum tube designer, Rene Magritte, "Ceci n'est-ce pas une amp."