Member
Joined 2009
Paid Member
I'm sure pcb guys can make anything you can draw, I can only guess, from a position of ignorance I have to say, that for a low cost pcb it's usually better to stick to simple, and one drill size would be even better (but not possible in this case).
Crossfeed - yes we need this
Low bias for battery - don't know, if the supply is clean a battery only helps with portability - assuming the source is also portable. I don't think we've optimized this design for portability - there are many other designs out there intended to be very portable.
Line Out - this is a good idea, not sure if impacts the pcb or is simply up to the constructor to add connectors and source selection switch ?
that's not how I would do it.
I thought headphones were more like 90dB to 100dB per milliwatt not per Watt.
40mA bias allows upto 75mA of ClassA output current. Is it a push pull output stage?
Max power into 32ohm phones ~90mW
90mW gives ~110dB from 90dB/mW phones. That is more than loud and we are still in ClassA.
However, looking at peak current values gives somewhat different results.
12Vpp = 6Vpk. This voltage into a full range (no crossover) driver may draw double the current that a nominal resistive impedance would draw.
Peak maximum current draw could be 6Vpk / 32ohms * 2 <=375mApk.
I would design for that level of ClassA output. It will be enough for quality listening even when the average level is 30dB below the peak ClassA limit.
I would also design the output stage and the driver to achieve that 375mApk output when driven into ClassAB.
Is my logic/understanding correct?
Andrew,
Thanks for your comments.
I think you are dead right about db/milliwatt. My mistake.
This would make things better rather than worse, I'd say.
We cannot go too high on dissipation because things will get too hot. With a 24V single rail and 40mA bias we still have to dissipate just under half a watt per output device, and this is about all we can tolerate with a small driver (4793/1837) and enclosure.
However, with up to 25mW output in Class A, this should be more than enough.
Cheers,
Hugh
Thanks for your comments.
I think you are dead right about db/milliwatt. My mistake.
This would make things better rather than worse, I'd say.
We cannot go too high on dissipation because things will get too hot. With a 24V single rail and 40mA bias we still have to dissipate just under half a watt per output device, and this is about all we can tolerate with a small driver (4793/1837) and enclosure.
However, with up to 25mW output in Class A, this should be more than enough.
Cheers,
Hugh
Now, slots might be difficult, I say just select the correct pin spacing for a cap of that size and VW rating....
What's Switchable Crossfeed? It sounds very painful, can we do it?
Now, if 40mA is our chosen bias, what would low bias be? 10mA, 20mA??
If we use a single 4AH battery of 24V, the 40mA would exhaust 25% of the battery charge (1AH) after about 9 hours, taking account of LTP and VAS currents. That would be enough for two generous listening sessions. I say we don't need it.... but hey, this is a consensus design, so over to you. Might be that the best thing to do is build a ratsnest prototype, and try it out at different bias settings for subjective assessment.
Cheers,
Hugh
What's Switchable Crossfeed? It sounds very painful, can we do it?
Now, if 40mA is our chosen bias, what would low bias be? 10mA, 20mA??
If we use a single 4AH battery of 24V, the 40mA would exhaust 25% of the battery charge (1AH) after about 9 hours, taking account of LTP and VAS currents. That would be enough for two generous listening sessions. I say we don't need it.... but hey, this is a consensus design, so over to you. Might be that the best thing to do is build a ratsnest prototype, and try it out at different bias settings for subjective assessment.
Cheers,
Hugh
Or we could call it "class B" mode. Maybe we can get away with very low bias?
Or perhaps, we can put schottky's on the emitters for a kind of "non-switching" class B as was shown by PaulBysouth and Kenpeter in my Allison thread.
Now c'mon guys... All these exotic ideas and no one is willing to ratnest it and see what we can do? We're developing a product. We shouldn't shun a good idea because we can't predict its audible affect.
If no one else can, I might as well, if someone is willing to send me some spare parts...
- keantoken
Or perhaps, we can put schottky's on the emitters for a kind of "non-switching" class B as was shown by PaulBysouth and Kenpeter in my Allison thread.
Now c'mon guys... All these exotic ideas and no one is willing to ratnest it and see what we can do? We're developing a product. We shouldn't shun a good idea because we can't predict its audible affect.
If no one else can, I might as well, if someone is willing to send me some spare parts...
- keantoken
A good point. The "sim" will assure that we have no fire or magic smoke , but won't tell us sqaut about what the final sound will be. The japanese Op's and my MJE's sim within 5% of each other (slightly different Vbe's)but I KNOW they sound different , either as a driver stage in a larger amp or as the main OP's on this one. I guess one could even try some TIP30/31's, dual supply , class/a or B , even ksa 992's as the input differential. This circuit will plot in a very similar manner with all the device I have mentioned.
I have gone "full circle" .. built an amp , changed input devices or voltage stage components .. THEN gone back (after hearing something I liked) and saw what was happening with my loop gain/UG/square wave to correlate "sound to sim". The simulation just gives one more "opinion" as opposed to just changing components and hoping for the best.
OS
Let's talk about crossfeed.
It involves time delays and cross connecting between channels. This must be done from a low Z source into a high Z source, uses two caps and three resistors. However, this approach means buffering, which implies either an opamp or an emitter follower output stage with CCS in the tail for low distortion. Because of phase shifts, the crossfeed circuitry cannot be included into the nfb loop, so the buffer and the input stage must be separated and electrically independent.
All this is possible, but it will double the component count and increase pcb design time. Crossfeed is generally done in the top end headphone amps from the better headphone manufacturers, so my thoughts are that if you want it for this project, it will double complexity, increase development time and otherwise delay the final product. Otherwise, keep it simple..... if you want crossfeed, use Jan Meier's design on Headwise, here is the link: Crossfeed HP Amp
Cheers,
Hugh
It involves time delays and cross connecting between channels. This must be done from a low Z source into a high Z source, uses two caps and three resistors. However, this approach means buffering, which implies either an opamp or an emitter follower output stage with CCS in the tail for low distortion. Because of phase shifts, the crossfeed circuitry cannot be included into the nfb loop, so the buffer and the input stage must be separated and electrically independent.
All this is possible, but it will double the component count and increase pcb design time. Crossfeed is generally done in the top end headphone amps from the better headphone manufacturers, so my thoughts are that if you want it for this project, it will double complexity, increase development time and otherwise delay the final product. Otherwise, keep it simple..... if you want crossfeed, use Jan Meier's design on Headwise, here is the link: Crossfeed HP Amp
Cheers,
Hugh
Last edited:
Hello
For those who want to use the crossfeed.
Here is a diamond buffer simulation circuit with its distortion spectrum. I's a simple circuit, so there is place to improve it.
I have choosen BD139 and BD140 for sim purpose but you can use better transistors.
The bias are 40 ma.
Bye
Gaetan
For those who want to use the crossfeed.
Here is a diamond buffer simulation circuit with its distortion spectrum. I's a simple circuit, so there is place to improve it.
I have choosen BD139 and BD140 for sim purpose but you can use better transistors.
The bias are 40 ma.
Bye
Gaetan
Attachments
Virtual Ground
Hi Hugh,
I have followed this thread with interest and would like to make a few comments if I may.
I am a keen HP user and have designed several HP amps some much the same as your current project.
What I would like to suggest is to design around a virtual ground system. It has the benefits of a true split supply (fed by single rail voltage), no possibility of earth loop and zero hum.
Hi Hugh,
I have followed this thread with interest and would like to make a few comments if I may.
I am a keen HP user and have designed several HP amps some much the same as your current project.
What I would like to suggest is to design around a virtual ground system. It has the benefits of a true split supply (fed by single rail voltage), no possibility of earth loop and zero hum.
Hi Nico,
This is a great idea, many thanks!! I had not even considered the earthing at this stage, but I know it can cause difficulties, much appreciated.
Gaetan,
The diamond buffer is a very good idea for the final drive stage, and will enable use of simple single ended input stages to power the crossfeed circuit.
However, I don't want to do a crossfeed here, I feel as it's covered so well in Headwize it's better to keep this project simple and easy to make.
Incidentally, I propose using a diecast aluminium case, and utilizing the longer sides to attach the four output devices. This will serve as a perfect heatsink, but will take some care with layout, of course. Amps one end of the pcb, power supply at the other. I have decided to use common mode chokes with an unregulated supply.
Cheers,
Hugh
This is a great idea, many thanks!! I had not even considered the earthing at this stage, but I know it can cause difficulties, much appreciated.
Gaetan,
The diamond buffer is a very good idea for the final drive stage, and will enable use of simple single ended input stages to power the crossfeed circuit.
However, I don't want to do a crossfeed here, I feel as it's covered so well in Headwize it's better to keep this project simple and easy to make.
Incidentally, I propose using a diecast aluminium case, and utilizing the longer sides to attach the four output devices. This will serve as a perfect heatsink, but will take some care with layout, of course. Amps one end of the pcb, power supply at the other. I have decided to use common mode chokes with an unregulated supply.
Cheers,
Hugh
Hello Hugh
All ok, I have done a diamond buffer for those who want an alternative to the opamp buffer from the Jan Meier's design on Headwise.
As input volume control, the one you suggest in your Aspen FAQ web page should work good.
Here it is for those who do not seen the page:
"A linear 100 kOhm pot with a simple 15 kOhm resistor from middle terminal (wiper, or centre; output) to ground. The 47K input impedance of the AKSA combines with this resistor to give around 11K4 from wiper to ground. Signal input is taken to the top terminal, bottom terminal is earthed, and the middle (wiper) terminal is then taken to the AKSA input."
I have others projects runing, but if I can I will try this Aspen headphone amp.
Bye
Gaetan
Last edited:
Member
Joined 2009
Paid Member
I like it, gives DIY'ers an affordable choice and keeps it simple.
Seems like the goals of the 'project' have shaped up along the following lines ???
1) we're looking for the 'Aspen' sound, something along the lines of AKSA so we will pick same active devices and as much of the proven topology as we're 'allowed to see'. This greatly reduces risk that significant prototyping is needed before a pcb is produced
2) keep it simple - no bulky heat-sinking, no fancy chasis, single rail choke filtered supply, no cross-feed, no exotic experiments, single pcb
3) Class A bias at 'normal' listening levels with 23 Ohm phones, flows into Class B for transients or higher listening levels = translates to 40mA idle
4) Avoid expensive parts or single-source specialized parts
Projects always go wrong for me, no matter how simple.
I thought so about the crossfeed too. It would be nice, but I think it should be included for a second option. I also don't like the idea of using opamps (for no good reason).
- keantoken
Hi Keantoken,
I listen to headphones a lot - at least 4 - 6 hours every evening while in my office. Yes I live with it.
Crossfeed for headphones is really essential. Most modern recordings are recorded directly off the instruments and this does cause listening fatigue as there is no natural mix of sound to both ears and is very anoying after a while.
Recordings are made for the masses who listen to speakers in a room. Crossfeed can be applied equally sucessfully at the output of the amp. There are elaborate networks for achieving this but creating just a little crosstalk already works wonders.
Of course if the performance is miced properly such as Shefield Labs and the like using two microphones on a dummy head this problem does not exist.
I am sorry Hugh, I have to support the crossfeed idea for anyone who is a serious headphne listener and for extended listening periods.
Kind regards
Nico
I listen to headphones a lot - at least 4 - 6 hours every evening while in my office. Yes I live with it.
Crossfeed for headphones is really essential. Most modern recordings are recorded directly off the instruments and this does cause listening fatigue as there is no natural mix of sound to both ears and is very anoying after a while.
Recordings are made for the masses who listen to speakers in a room. Crossfeed can be applied equally sucessfully at the output of the amp. There are elaborate networks for achieving this but creating just a little crosstalk already works wonders.
Of course if the performance is miced properly such as Shefield Labs and the like using two microphones on a dummy head this problem does not exist.
I am sorry Hugh, I have to support the crossfeed idea for anyone who is a serious headphne listener and for extended listening periods.
Kind regards
Nico