Rob Flinn
Well-known member
Thanks again
Good DIY headphone amp?
- Thread starter jrmintz
- Start date
Help Support GroupDIY Audio Forum:
NewYorkDave
Well-known member
By the way... I haven't verified this for real yet, but the simulator says that the line amp circuit I posted earlier in the thread should be able to deliver hundreds of milliwatts into medium-Z phones with low distortion without the use of an output transformer. I simulated it with a feedback resistor of 27K (for a net gain of 20dB), a 2.2uF feedback coupling cap, and a large (~ 100uF) output coupling cap. For use with low-Z phones, you'd still want an output transformer.
I'll try this out for real sometime if I find the free time...
I'll try this out for real sometime if I find the free time...
NewYorkDave
Well-known member
No, the sleeve of the output connector doesn't have to be isolated from the chassis. Go head and use those good ol' chassis-mount TRS Switchcraft jacks you have laying around. :wink:
> a hard time driving low-Z phones without bad clipping on the lower half of the waveform. That's because a single-ended cathode follower cannot sink current, it can only source.
As the Designer, Jakob could use any value of pull-down current he wanted; that is rarely the limiting factor.
The key problem, especially with vacuum tubes, is the maximum current the device will supply. For best unclipped power in Class A1 operation, when output voltage swing is small (as it is here), set the idle current to about half the tube(s) maximum current.
In Jakob's plan, total current is 25mA and we might guess that most of that flows through the output. Say 20mA. That flows through 3K so the voltage on the bottom of the 'AU7/'88 is 60V. The plate voltage is 160V. Turning to the 12AU7 curves, at 100Vp and 0Vg we can get about 12mA each, 36mA for three units. So ideal bias current is 18mA, which is essentially where Jakob put it.
To get more current, he would need a smaller resistor (easy) AND more 12AU7/ECC88s (hard/costly).
> For high output into lower impedance phones, it would be better to use a push-pull output such as a White cathode follower.
The WCF has the same problem: you are limited by the peak current the tube will pass. A WCF won't run Class-B. It is worse than a resistor-loaded CF, because it needs twice as many tubes and they split the supply voltage. A resistor can live on less voltage than a tube. The WCF may have a lower supply current because it can idle at half the peak current. (But the higher supply voltage often makes the efficiency about the same.) The WCF will have half the small-signal output impedance, which may be a factor in some cases. What it excels at: up to 90% of clipping-point, its distortion cancels very well. The CF may run 5% THD where a similar WCF may be 1% THD. (But as you say, "some second-harmonic distortion that singers may very well find pleasing".)
> Obviously Cathode followers are very low Z output
The small signal impedance is very low. But you can NOT load a CF with anything like its output Z and get large signals: it distorts. The best-power load for a CF is identical to the optimum for the same tube working plate-loaded: Rp to 2*Rp. 12AU7 is about 6K, we got three so 2K, we want to load it in 2K to 4K to get the most power. Anything less gives less power, and below 1K it falls very quickly.
However a single 12AU7 can make about 100mW in 15K, and we have three, so we can tolerate a lot of loss of power and still drive many headphones.
> If... you plugged some 8 ohm headphones in would you compromise frequency response...
Increase the value of the output cap. Although actually it does not have to be much bigger than 100uFd: the cathode impedance is 100Ω.
> ... or power output? How low impedance can the heaphones be for Jakobs circuit?
How good are your ears? It will make "some" power in 1Ω.
How much? Do the simple math. We know from inspection that the output stage runs about 20mA, and for impedances much lower than 3K "most" of the bias current can be diverted into peak current in the load. Assume 20mA peak, which is 14mA or 0.014A RMS. Multiply that by impedance to get Volts. Stick Volts and impedance in the formula for Watts. (Or you can just figure I^2*Rl.)
1Ω load. 0.014A * 1Ω = 0.014V. 0.014V^2/1Ω = 0.000196 Watts. Or 0.2 milliWatts. Typical headphone sensitivity is, say, 90dB SPL at 1mW. Peak loudness will be 83dB. We usually want 15dB headroom between speech/music long-term average and peak levels. Listening level is 68dB SPL.
Re-calculating with 100Ω we get 20mW and 81dB SPL listening levels. Elsewhere I said about 11mW: this is the difference between "low"-Distortion (actually almost 5% 2nd) and gross clipping. It will be pretty "warm" above 10mW and go splatt at 20mW.
As the Designer, Jakob could use any value of pull-down current he wanted; that is rarely the limiting factor.
The key problem, especially with vacuum tubes, is the maximum current the device will supply. For best unclipped power in Class A1 operation, when output voltage swing is small (as it is here), set the idle current to about half the tube(s) maximum current.
In Jakob's plan, total current is 25mA and we might guess that most of that flows through the output. Say 20mA. That flows through 3K so the voltage on the bottom of the 'AU7/'88 is 60V. The plate voltage is 160V. Turning to the 12AU7 curves, at 100Vp and 0Vg we can get about 12mA each, 36mA for three units. So ideal bias current is 18mA, which is essentially where Jakob put it.
To get more current, he would need a smaller resistor (easy) AND more 12AU7/ECC88s (hard/costly).
> For high output into lower impedance phones, it would be better to use a push-pull output such as a White cathode follower.
The WCF has the same problem: you are limited by the peak current the tube will pass. A WCF won't run Class-B. It is worse than a resistor-loaded CF, because it needs twice as many tubes and they split the supply voltage. A resistor can live on less voltage than a tube. The WCF may have a lower supply current because it can idle at half the peak current. (But the higher supply voltage often makes the efficiency about the same.) The WCF will have half the small-signal output impedance, which may be a factor in some cases. What it excels at: up to 90% of clipping-point, its distortion cancels very well. The CF may run 5% THD where a similar WCF may be 1% THD. (But as you say, "some second-harmonic distortion that singers may very well find pleasing".)
> Obviously Cathode followers are very low Z output
The small signal impedance is very low. But you can NOT load a CF with anything like its output Z and get large signals: it distorts. The best-power load for a CF is identical to the optimum for the same tube working plate-loaded: Rp to 2*Rp. 12AU7 is about 6K, we got three so 2K, we want to load it in 2K to 4K to get the most power. Anything less gives less power, and below 1K it falls very quickly.
However a single 12AU7 can make about 100mW in 15K, and we have three, so we can tolerate a lot of loss of power and still drive many headphones.
> If... you plugged some 8 ohm headphones in would you compromise frequency response...
Increase the value of the output cap. Although actually it does not have to be much bigger than 100uFd: the cathode impedance is 100Ω.
> ... or power output? How low impedance can the heaphones be for Jakobs circuit?
How good are your ears? It will make "some" power in 1Ω.
How much? Do the simple math. We know from inspection that the output stage runs about 20mA, and for impedances much lower than 3K "most" of the bias current can be diverted into peak current in the load. Assume 20mA peak, which is 14mA or 0.014A RMS. Multiply that by impedance to get Volts. Stick Volts and impedance in the formula for Watts. (Or you can just figure I^2*Rl.)
1Ω load. 0.014A * 1Ω = 0.014V. 0.014V^2/1Ω = 0.000196 Watts. Or 0.2 milliWatts. Typical headphone sensitivity is, say, 90dB SPL at 1mW. Peak loudness will be 83dB. We usually want 15dB headroom between speech/music long-term average and peak levels. Listening level is 68dB SPL.
Re-calculating with 100Ω we get 20mW and 81dB SPL listening levels. Elsewhere I said about 11mW: this is the difference between "low"-Distortion (actually almost 5% 2nd) and gross clipping. It will be pretty "warm" above 10mW and go splatt at 20mW.
Similar threads
