Shop OBEX P1 Docs P2 Docs Learn Events
Headphone Direct Drive — Parallax Forums

Headphone Direct Drive

cgraceycgracey Posts: 14,133
edited 2021-02-20 18:33 in Propeller 2

@Yanomani said:
Sorry by insisting in a some way off-topic route, but...

@rogloh said:
I have just tried and the levels are now far more satisfying on the headphones directly attached to the P2 pins on the RCA (via the 22uF coupling caps) and there is less distortion. Bass sounds good. This is still 8 bit audio remember.

@cgracey said:
75-ohm DAC pins can be ganged together to lower the impedance, you know.

Perhaps we could get rid of that headphone amp chip.

For two paralleled-DAC outputs, ~37.5 Ohm drive, 22 uF is only ~5% of the required capacitance.

Simple calculations shows ~420 uF minimum, for a -3DB cutoff rolloff at 10 Hz, so as not to "strangle" too much the bass response.

https://v-cap.com/coupling-capacitor-calculator.php

It seems like we should make a board which gangs four pins for each channel. Each pin goes through a separate 220uF series cap and each group of four (of the other sides of the caps) sums together to form left and right channels. This would result in about 19 ohms source impedance, which should be about perfect for driving headphones.

I need to get some 220uF caps and a 3.5mm stereo jack, in order to try this.

A lot of careful design went into the DACs to ensure they would be good for audio. Maybe they should just drive the headphones directly.

«1

Comments

  • RaymanRayman Posts: 13,800
    edited 2021-02-20 18:47

    I wonder if an audio transformer would work...
    Like this: http://www.audiomisc.co.uk/HFN/HeadphoneDAC/HeadDAC.html

    Nevermind. I don't think that works with this kind of DAC. Still, if there were a small, cheap transformer...

  • cgraceycgracey Posts: 14,133
    edited 2021-02-20 19:06

    Yes, the transformer idea is appealing, but, like you said, we need lower-impedance transformers than he was using.

  • pik33pik33 Posts: 2,347
    edited 2021-02-20 21:11

    These DACs are audiophile quality. ADCs are much noisier but it seems this can be also worked around.

    One DAC can output 3.3V @ 120 Ohm, and this gives ~1:5 divider when use 32 Ohm headhones. This gives ~1.8 mW, 90-110 dB, depending on headphones, enough for normal use. Maybe 2v/75 Ohm can be even better.

    I connected DT770 Pro headphones directly to pin0 and 1 via RCA. These are low sesitivity but good quality headpgones: it works, even via these caps (bass cut out) or dorectly with a bass and... DC. Now I need capacitors :)

  • cgraceycgracey Posts: 14,133

    If headphones didn't have a common ground, we could operate each channel in push-pull using two DAC pins. We could certainly drive a transformer primary like that.

  • cgraceycgracey Posts: 14,133

    I think ganging pins through 220uF caps is the simplest approach, but maybe those caps aren't cheap. Probably cheaper than a decent transformer, though.

  • @cgracey said:
    If headphones didn't have a common ground, we could operate each channel in push-pull using two DAC pins. We could certainly drive a transformer primary like that.

    Really posh headphones do have a 4 pin connector without common ground.

  • cgraceycgracey Posts: 14,133
    edited 2021-02-20 21:34

    @Wuerfel_21 said:

    @cgracey said:
    If headphones didn't have a common ground, we could operate each channel in push-pull using two DAC pins. We could certainly drive a transformer primary like that.

    Really posh headphones do have a 4 pin connector without common ground.

    Do you have any link handy? It makes perfect sense, but I've never seen such headphones, myself.

  • @cgracey said:
    If headphones didn't have a common ground, we could operate each channel in push-pull using two DAC pins.

    Maybe you still can if this is possible with three pins: Mix R and !L on pin A, L and !R on pin B, and !L and !R on pin Ground.

    -Phil

  • cgraceycgracey Posts: 14,133

    @"Phil Pilgrim (PhiPi)" said:

    @cgracey said:
    If headphones didn't have a common ground, we could operate each channel in push-pull using two DAC pins.

    Maybe you still can if this is possible with three pins: Mix R and !L on pin A, L and !R on pin B, and !L and !R on pin Ground.

    -Phil

    That's a mighty succulent idea, Phil!

    I need to think about that. I guess it would be fine for headphones, but not for a line-out where a common ground is expected.

  • @cgracey said:

    @Wuerfel_21 said:

    @cgracey said:
    If headphones didn't have a common ground, we could operate each channel in push-pull using two DAC pins. We could certainly drive a transformer primary like that.

    Really posh headphones do have a 4 pin connector without common ground.

    Do you have any link handy? It makes perfect sense, but I've never seen such headphones, myself.

    The keyword is "balanced". Usually using a 4.4mm "pentaconn" jack, but a 2.5mm TRRS is also common.

    Not a lot of headphones have it by default, but any headphones where each ear has its own detachable cable can in theory be upgraded to balanced input, so it's usually a separate cable to buy. Well, these ridiculously expensive sennheisers have one included, haha.

  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2021-02-20 22:38

    I use 2 DACs for bridged audio just to get some more volume out of a piezo speaker but have not tried direct drive to a low-impedance load. However, as soon as large caps are being considered, I find that an amp chip is very cheap and very tiny. There is this SD8002D from LCSC which is a class-AB with BTL with 2W into 4ohm load (3W max). I think this is cheaper and far smaller than a large cap and only requires a single DAC.

    Oh, btw, the chip costs 5 cents although you'd better hurry, they are down to their last 25,480 chips (and erco might be lurking around).

  • cgraceycgracey Posts: 14,133

    @"Peter Jakacki" said:
    I use 2 DACs for bridged audio just to get some more volume out of a piezo speaker but have not tried direct drive to a low-impedance load. However, as soon as large caps are being considered, I find that an amp chip is very cheap and very tiny. There is this SD8002D from LCSC which is a class-AB with BTL with 2W into 4ohm load (3W max). I think this is cheaper and far smaller than a large cap and only requires a single DAC.

    Oh, btw, the chip costs 5 cents.

    But, does it sound GOOD? What would Ahle2 and pik33 hear?

  • cgraceycgracey Posts: 14,133

    @"Phil Pilgrim (PhiPi)" said:

    @cgracey said:
    If headphones didn't have a common ground, we could operate each channel in push-pull using two DAC pins.

    Maybe you still can if this is possible with three pins: Mix R and !L on pin A, L and !R on pin B, and !L and !R on pin Ground.

    -Phil

    I think you would want to use maybe two DACs on the common pin? And what would channel separation be like?

  • And what would channel separation be like?

    Pin A to ground would be (R - L) - (-L - R) = 2R; pin B to ground would be (L - R) - (-L -R) = 2L.

    -Phil

  • The point is that if you can get a BTL amp chip for 5 cents, you won't have any roll-off because of the direct drive and it only takes one DAC and no room at all. That must sound better already and THD at headphone drive levels are much lower.

    But there are more expensive 30 cent chips from ON that feature low distortion (around 0.03% at 1KHz @250mW).

  • cgraceycgracey Posts: 14,133

    @"Phil Pilgrim (PhiPi)" said:

    And what would channel separation be like?

    Pin A to ground would be (R - L) - (-L - R) = 2R; pin B to ground would be (L - R) - (-L -R) = 2L.

    -Phil

    Wouldn't there wind up being some cross-talk between left and right due to dynamic impedance of the speakers?

  • Wouldn't there wind up being some cross-talk between left and right due to dynamic impedance of the speakers?

    Dunno. If there is, perhaps some additional math could cancel it out?

    -Phil

  • cgraceycgracey Posts: 14,133

    @"Phil Pilgrim (PhiPi)" said:

    Wouldn't there wind up being some cross-talk between left and right due to dynamic impedance of the speakers?

    Dunno. If there is, perhaps some additional math could cancel it out?

    -Phil

    I don't know.

    Do you think the virtual GND signal would need to be half the impedance of the channel signals, since it is shared?

  • Do you think the virtual GND signal would need to be half the impedance of the channel signals, since it is shared?

    -Phil

  • cgraceycgracey Posts: 14,133

    @"Phil Pilgrim (PhiPi)" said:

    Do you think the virtual GND signal would need to be half the impedance of the channel signals, since it is shared?

    -Phil

    That''s exactly what I thought.

    Phil, I'm having strange ideas about using a smart-pin mode that counts clock cycles until some target number of ones or zeros arrives, to linearize ADC readings for those strings of pushbuttons with resistors across them.

  • All good points on driving a set of good quality headphones. But it got me thinking, can some of us electrical engineer types come up with a agreed on, somewhat easy to use and somewhat easy to build, output amplifier, good up to the limits of the DACs (RF region?) that would drive a reasonable AC or DC load, be it a relay, audio, or RF, maybe ?

    Conversely - A buffered input circuit for either DC or AC signals, with higher than normal spurs or RF components, that would otherwise damage the P2 i/o pins or internal circuitry of the chip?

  • cgraceycgracey Posts: 14,133

    @PropGuy2 said:
    All good points on driving a set of good quality headphones. But it got me thinking, can some of us electrical engineer types come up with a agreed on, somewhat easy to use and somewhat easy to build, output amplifier, good up to the limits of the DACs (RF region?) that would drive a reasonable AC or DC load, be it a relay, audio, or RF, maybe ?

    Conversely - A buffered input circuit for either DC or AC signals, with higher than normal spurs or RF components, that would otherwise damage the P2 i/o pins or internal circuitry of the chip?

    I think, depending on what the amp is needed for, the topology would have to vary widely to be economical for each application.

  • cgraceycgracey Posts: 14,133
    edited 2021-02-21 06:05

    I spent some time tonight looking for transformers to solve the headphone problem and it seems hopeless. There is nothing that would be ideal and the cost would be at least $3.00 per channel.

    Instead of looking more, I just ordered a $10 assortment of electrolytic caps from Amazon that go up to 1000uf 16V. A cap'tator, as Beau used to say, is the simplest solution to this problem. I'll connect two DAC pins, each to a cap, tie the other cap ends together, and that will be a 37.5-ohm channel, which will be just about right for 32-ohm headphones. If you need 16-ohm output, gang four pins with caps. This will push the signal through to the headphones without any DC bias. I think it's the most practical way to hear what the DACs can do audio-wise.

  • Ahle2Ahle2 Posts: 1,178
    edited 2021-02-21 06:28

    @pik33 said:
    These DACs are audiophile quality. ADCs are much noisier but it seems this can be also worked around.

    One DAC can output 3.3V @ 120 Ohm, and this gives ~1:5 divider when use 32 Ohm headhones. This gives ~1.8 mW, 90-110 dB, depending on headphones, enough for normal use. Maybe 2v/75 Ohm can be even better.

    I connected DT770 Pro headphones directly to pin0 and 1 via RCA. These are low sesitivity but good quality headpgones: it works, even via these caps (bass cut out) or dorectly with a bass and... DC. Now I need capacitors :)

    Yes, being an audiphile for decades and been working with audio gears both professionally and as a hobby, I can tell you that the P2 audio DACs are of a very high quality. That is by ear and I will do some measurements in my home "studio" with semi professional equipment some time. @Chip a transformer WILL affect the sound, but many (almost always oldschool) audiophiles claimes the coloration adds to the quality. In my opinion anything that distorts the sound in ANY form makes it worse. But to be completely honest, using cheap consumer level equipment, you could hardly hear the difference between the P1 and P2.

  • pik33pik33 Posts: 2,347

    you could hardly hear the difference between the P1 and P2.

    Maybe on very entry level. As I have not enough money for true audiophile toys, I have an audio equipment, selected by "good enough quality at low enough price" The difference between 300 kHz noise shaped DAC on P1 and 300 MHz real DAC on P2 is huge

    I connected the P2 with DACs on to 120 W amplifier, then set the amplifier to 100% volume and what I hear was silence, if only the cable was good (or all kind of the noise induced in not as good quality cable - you can't tell about the cable until you connect it.)

    An interesting experiment: you can actually hear how the "noise DAC" works. Get a "headphone" spin2 demo, set the frequency to 0.01 Hz. Connect the P2 to a good amplifier, run the program and then set the volume at 100%.

    Capacitors are cheap. I hope I have several of them in my drawer, if not I can buy them tomorrow (we have near all shops closed in Sunday). I also need an optoisolator and a DIN 5pin socked: I need a MIDI IN accessory board.

  • Ahle2Ahle2 Posts: 1,178

    @Yes the difference between the P1 and P2 is HUGE; I was just mentioning that people that run the audio through the TV, hamburger speakers, laptop speakers or even mid range PC audio systems will not hear the difference because the "audio chain" affects the sound way more. The P1 was not even close to audiophile quality. I would argue, just by hearing, that the P2 is close to audiophile quality. Measurements will tell. It's good enough that my ears can't easily tell any obvious colorations when playing CD quality wavs and connected to my good quality main audio system. The P1 however.... it was good enough for retro bleeps.

  • cgraceycgracey Posts: 14,133
    edited 2021-02-21 09:25

    The P2 DAC topology is the simplest that exists, I think. Imagine 256 identical CMOS drivers with matched sink/source each driving an identical resistor, and all the resistors are tied together on the other side. That's it. The DAC is the power stage, too. No funny R-2R business and no active buffers. Just passive resistors being equally driven and tied together. Simplicity is best. I'm thinking these big, ugly 1000uF caps are going to take care of the transformer problem in a much cleaner and cheaper way, so that we can drive headphones nicely.

  • pik33pik33 Posts: 2,347
    edited 2021-02-21 11:27

    A homebred contraption may be much faster solution as there is no local Amazon shop here, and the global one says "Deliver: March 5-18"

  • jmgjmg Posts: 15,140

    @cgracey said:
    .... Simplicity is best. I'm thinking these big, ugly 1000uF caps are going to take care of the transformer problem in a much cleaner and cheaper way, so that we can drive headphones nicely.

    Transformers are certainly dinosaurs and clunky, but caps appear simple only on the schematic.
    Caps are physically large, and quite expensive, and worse, mean the current peaks come from the 3v3 rail.
    A simple, (much) lower cost and smaller amplifier, allows current from the 5V rail, and bridge drive is very simple.
    SO8 is common, but it also seems MSOP8 has a standard pinout in both class-AB and class-D amplifiers eg LM4890S, LM4990MM/NOPB, BL6306, IS31AP4991A etc

Sign In or Register to comment.