I spent some time doing some simulation of the JLH headphone amp. A Word file is attached.
I tried the amp with split supplies of +/- 12 V and a single supply scheme using a single + 24 volt supply and a +12 volt supply.
All seem to function fairly well. Distortion is low and I tried comparing them producing about 100dB spl on phones with 32 ohms/ 90 dB/mW and 300 ohms/94dB/mW. The better phones have higher sensitivity than what I assumed.
I have not made one but intend trying out one. Output impedance is very low at less than 0.2 ohms except the single supply unit which has a capacitor in series with the output and causes the Zout to rise with falling frequency.
The output stage has about 100mA through it and PSRR is just about 50dB. So a low noise power supply is required. A wall wart might not be suitable I think, unless you add a capacitance multiplier after it ?
I'm going to try out the single supply version with an onboard supply to which a wall wart can be connected if not a transformer.
The single supply circuit performs pretty well on paper. I can compare it with my opamp / discrete output driver headphone amp. Basically an opamp ( OPA2134 ) driving a pair of TO126 devices through some TO92 devices. It's part of my DAC and can be used by itself.
Another advantage of the single supply circuit is that the output cap cuts off all dc to the phones. So it's pretty safe if anything breaks down. However the performance of the output cap itself is to be checked. No output cap is of course the best way ...but....
I simulated a 6922 tube circuit ( at 40 volts) with MOSFET follower and it performs poorer on paper. Wondering if I should rig it up to see how it sounds.........
I finally got a pcb layout done. I've been trying to make a pdf file from the .brd file but couldn't figure out how. Didn't have much time to Google and check that. However I just managed to make a .tif file. Let me see how I can attach that.
I've made some slighty modifications to the original JLH circuit. It looks fine on simulation and stable also. I'll post that soon.
Does anyone know the inductance of a typical 32 ohm and 300/600 ohm headphone phone ? Couldn't find that data on several web sites.
Would like to know if anyone made a stock JLH single supply headphone amp earlier ? If so any comments ?
I've attached it as a word file.
Note that the + ve going peak will always be about 2.6 volts less than the supply voltage. The negative swing goes almost down to 0 Volts. So with 12 volts you can get a maximum of 9.4 volts peak to peak. This is only possible if the output dc voltage is not at 1/2 Vsupply but slightly less than + 5 volts.
As shown you can get peak levels of about 110 dB which can be damaging to the ear . With more sensitive headphones it will be even higher. Note that a 45 mA quiescent current should be fine for most listening purposes. The transistors will need a small heat sink to keep them reasonably cool. With a 110 deg/C per watt rating in free air it means a 30 degC temp rise . Not Ok here with ambient as high as 35 deg C in summer !
I found an error in the pcb layout. Will have to include two resistors and capacitors that I missed out. That's going to delay things some more. Meantime I found my old Tube/MOSFET hybrid phones amp. It certainly will have higher distortion, but it will be interesting to see how it compares.
JLH headphone amp with a single supply-2 - Feb2010.doc
That looks interesting. Just goes to show that 'most' circuit configurations have already been tried long ago.
I've just corrected more errors on my board design. Added four smd parts as I could not fit them otherwise. Forgot about them when I started and I don't plan to move everything around and spend a lot of time on it. The amp will still work without them.
I like big pads and fairly wide tracks because they don't peel off when experimenting and soldering and desoldering several times.
Hope to get something ready by the weekend. It's a very busy month and so I cannot say when it will really get done.
JLH steering output transistor base current, ignoring output voltage and current.
(except in the GNF loop, after the paralleling). Seems a fairly compatible topology
for parallel abuse. Certainly with the 74LS04, as you got 120 ohms extra insurance
built into each output totem.
Could piggyback a dozen or more if you could figure a way to get the heat out...
PNP Collector pulldown must be proportionally smaller the more gates you drive.
Sim a dumbed down inverter logic gate in LTSpice is not the same, and it does
not work in this application. You need the full internal schematic to understand
why the input has to be pulled down.
I got the board done only to find that I'd forgotten to reverse the image !
Luckily it can still be used as the TO92 leads can be made to fit the reversed holes. The board is now fully populated. Hope I get some time tomorrow to try it out. Not a very nice layout but functional.
Will accept 12-0-12 transformer or 0-12 transformer or +12 Vdc from a wall wart.
Time to crash out today.It's 12.30 am!
Will post details after the listening test.....to determine if it's worth posting the details at all.
This was just a hasty test. I don't want to make any sweeping statements that I might have to retract again.
Initial impression is that it sounds very good. Without a level matched comparison with another headphone amp using an opamp front end with TO92+TO126 output stage running in classA mode, I think it sounds better. But I should make a level matched test to be really sure..........maybe , but the sonic difference is quite large right now. Wonder if level matching will negate that. It sounds much cleaner I think.
Now to put it under the scanner to see what it is really doing. This might take some time.
The circuit diagram is attached.
I used a 12-0-12 Volt power transformer feeding two FR107 diodes to a 2200uF cap. This is connected via a 10 ohm resistor to another 2200uF cap. This is the supply to a capacitance multiplier made up of a BD139 with 470 ohm resistor from collector to base. 470uF cap from base to ground and 100uF cap from emitter to ground.
I don't hear any hum or hiss from the headphone. Must try this at night when the ambient noise is very low. The LED on board is a red 5mm diffused type and isn't very bright. The heat sink I've used is about 30x30x1.6 mm. Both output transistors are mounted on it using mica insulators.
The heat sink after one hour is about 39 deg C and the ambient temperature is 25 deg C.
When it's very quiet I can hear faint hum in the can's ! So a better supply regulator is required. I will of course rip up the grounding scheme of the board to deterimine if that is the cause. So more work ahead . A second listening session shows up the other headphone amp as sounding muffled in comparison. Is this one boosting HF ? Will have to run the tests now.
Note the cheapo yellow box 1uF input cap and standard Samwha elco's. Still sounds very good. No sloppiness in the bass. Nice and tight / deep bass and very good HF . On phones it has shimmer on the cymbals ( on some tracks)! Wonder how it would sound as a pre amp. That's for later!
Nothing much to say about frequency response. It's flat out to 20 Khz. About -0.04dB at 20 Khz . At 20 Hz it's down about -1.3 dB with a 33 ohm load and less than -0.2 dB with a load higher than 100 ohms.
Distortion spectrum is great with the 2nd harmonic prominent and rapidly decreasing higher harmonics. Tube like I'd say !
The max output is just about 2.8 V rms with a 12 V supply. Enough for most listening as it's quite loud.
Hum as expected is very high . Only about -65dB down and getting poorer as the load impedance decreases. Needs more work on the power supply. A simple capacitance multiplier is not good enough.
However the noise floor looks quite low , especially at HF and overall looks good. The low quiesent current isn't enough for a 32 ohm load for full output ( about 4 V peak ). That's expected. With a 0.135mA quiescent current the amp clips at just about +4V peak and the heat sink is running at about 45 degC when the ambient is 24 deg C . Distortion with a 33 ohm load also drops dramatically over the operating range with a higher quiescent current.
Will work on the power supply and see what I can do. I think the hum component ( 100Hz and higher harmonics ) should be brought down by 30dB or at least 20dB. Bigger caps yes, but then it will not bring it down fast enough. Maybe a TO220 regulator. I wanted to avoid using one ! Will try something tomorrow. I should also test it at higher supply voltages.
Will post some graphs later.
Interesting reading your progress
I thing using a 7812 is the way to go, and make sure the grounding is correct... proper star point taken as a spur from the negative end of the main smoothing cap. Even connecting grounds tightly together isn't good enough... it has to be on a spur to eliminate the AC ripple that flows in the PSU.
Yes the unit is star grounded. The signal paths and supply ground approach the ground node from opposite sides. But then again there could be other mistakes and I will be examining them today. The simulated capacitance multiplier has far lower ripple in it's output than the practical measured values. Must figure out what's wrong. It's wrong by about 60dB ! Maybe the simulation is wrong !