DIY iPod digital out

[dsavitsk]

Originally Posted by error401

Check out how dsavitsk did his HPDAC supply, you could probably poach most of that design, just respec it for 3V input.

I believe that it is happy with input DC as low as 1.8V -- look at the datasheet though.

There are tons of good switching chips out there, but I'd probably use the TPS61040 dsavitsk chose.

I don't know that it is anything special, other than being cheap and easy to find. I does need that ferrite after it as otherwise it is pretty noisy. But, I found it by calling TI. If you get the right person on the phone, they are pretty helpful.

Originally Posted by joneeboi

Further, unlike the HPDAC, we could choose our own amplifiers.

Why couldn't you do that there? You were welcome to choose anything you wanted, it is just that most of those choices, other than the LM6172, were poor ones.

Originally Posted by 00940

The so8 reg101 was still the regulator we picked for our tentative "all-out" transportable USB dac with Ble0t.

What ever became of that?

[FS2]

How about volume control? The I2S data isn't affected by the volume setting from the scrollwheel. It'll either be a matter of using the DACs on-board volume control using a microcontroller, or doing it on the analog output. Unless there are DACs that allow easy hardware control of the volume settings.

Powerwise, I was looking at an inverting switcher to generate a negative voltage for the op-amp. Any advantage to simply using a rail splitter and a higher boosted positive rail?

And a general question for you guys who have better ears than I do: Is there any difference in the "stock" iPod's sound when charging it?

[error401]

Originally Posted by FS2

Powerwise, I was looking at an inverting switcher to generate a negative voltage for the op-amp. Any advantage to simply using a rail splitter and a higher boosted positive rail?

I don't think that's necessary if you do a balanced configuration. There will be DC offset on each output node, but the net offset will always be 0V (or very close to it). So since the DAC operates from 0-5V, your opamps can be single ended as well.

[FS2]

I thought it was... If, say, L+ and L- both have the same DC offset, and one is applied to the non-inverting op-amp terminal and the other to the inverting, the DC offset cancels out. Unless I've missed something...

[error401]

Originally Posted by FS2

I thought it was... If, say, L+ and L- both have the same DC offset, and one is applied to the non-inverting op-amp terminal and the other to the inverting, the DC offset cancels out. Unless I've missed something...

I'm talking about going balanced all the way to the headphones. In essence you have two amplified lines that are 180 deg out of phase for each side, instead of a single one and ground. The DAC itself provides these inverted outputs, and you simply need to use four opamps instead of two for your output stage.

To make things a bit more clear:

The DAC can only output a voltage between 0-AVdd. Since this single ended signal needs to represent an audio signal, it needs to be biased to exactly 0.5AVdd, which will be the output voltage of both the + and - outputs when outputting silence. If you follow both + and - with a non-inverting amplifier, they will both output Gain x 0.5AVdd - exactly the same voltage. Each output goes to the + or - input of a headphone driver. Since there's no ground reference fed to the headphones, the ground-referenced voltage is meaningless at the headphones - only the difference matters. They see 0V since both + and - are exactly equal.

Problem is that normal headphone connectors don't allow this (it requires 4 wires instead of the usual 3), and the convention is to use XLR connectors which are rather massive. You could rewire your headphones yourself and use a less insane connector like a mini-DIN or 4-pin 3.5mm.

Alternative options are to only use DC blocking caps with only one DAC output, or a proper balanced to SE converter (conventional designs usually need 3 opamps/channel). Or a DC servo I guess, but the converter usually makes more sense.

[FS2]

Right, I'd thought that you were outputing to a common 3 terminal headphone, not going balanced all the way.

Been going over the design I have, trying to flesh it out. It'll need a DAC, micro for volume control, 6 op-amps for the output, a +/- voltage for the op-amps and a lower +ve voltage for the DAC.

[joneeboi]

I just ran through the Opus DAC thread, and I like the jumpering idea. My great idea is to use tiny switches instead, but progress is super slow. I was looking at the "Recommended External Components" section in the WM8740 datasheet and I wonder if we could open a discussion on the output stage in between the chip and the amplifier. Trying to fit this bad boy into the 1455C8 really limits coupling caps, as film and tants I've seen are usually well over 100V. If we slip the DAC board in the bottom slot, we can fit those giant caps lengthwise above the board. That still limits cap choices, but it still accommodates most popular boutique caps. We could also leave the option of opamp balanced-to-SE DC-eliminating output onboard too. Was there another option? DC servo, IIRC. I actually don't understand what that means, so maybe you guys can enlighten me.

[error401]

Originally Posted by joneeboi

I just ran through the Opus DAC thread, and I like the jumpering idea. My great idea is to use tiny switches instead, but progress is super slow. I was looking at the "Recommended External Components" section in the WM8740 datasheet and I wonder if we could open a discussion on the output stage in between the chip and the amplifier. Trying to fit this bad boy into the 1455C8 really limits coupling caps, as film and tants I've seen are usually well over 100V. If we slip the DAC board in the bottom slot, we can fit those giant caps lengthwise above the board. That still limits cap choices, but it still accommodates most popular boutique caps. We could also leave the option of opamp balanced-to-SE DC-eliminating output onboard too. Was there another option? DC servo, IIRC. I actually don't understand what that means, so maybe you guys can enlighten me.

A DC servo is basically a circuit connected with a very slow low-pass filter. It ignores all the audio data and essentially only reacts to DC. It is configured to 'want' 0VDC into the output. Properly connected so it can control the output circuit, it will bias the output to 0V, thanks to the slow low pass, even if there is an intrinsic offset. If you wanted to use one, I believe you'd need two opamps: the servo and an output stage after the DAC. Since you can't inject feedback into the DAC, I don't believe you could connect a servo directly to the output (though I might be wrong...).

If you do bal->SE with a separate output stage, you need 6 opamps for the bal->SE and another 2 (or some other type of amp) for output. If you can fit it, that's what I would do. Use low-Iq opamps as much as possible and you can probably keep the Iq fairly low. It's going to hurt battery life though. You'll also need a negative-side DC-DC as well I believe, if you want to avoid coupling caps. You *might* get away without one by creating a buffered 1/2Vpos and using it as ground for the DAC IC. Not sure if it'd work though, I'd have to think about it more than I have.

Using one side of the DAC and amplifying it in the conventional manner only requires 2 opamps, so uses less space and power. If you have a battery life budget you might be able to use 'nicer' opamps as a result. But you need coupling caps, and boutiques aren't going to fit. If you're one of those that believe that boutique caps make a difference, that might be a showstopper.

[joneeboi]

Thanks for the help. It sounds like we might have to copy Twisted Pear more than I'd prefer. IIRC, they have the option for all three, right? I'm thinking of regulating USB and forgoing the whole battery mess.

With the boutique caps issue, I understand you can block DC with any appropriately spec'ed cap; the boutique bit would only be useful for preventing the use of poorer spec caps. Their sound difference is doubtless overemphasized and almost deified/idolized in some sense, and I can understand wanting to shy away from them. Nevertheless, I'd still argue they make a difference when placed directly in the signal path just like switching in lower-quality parts, audible or not. Is there something I'm missing? I'm sure the layout can be implemented such that their inclusion doesn't interfere with conventional non-boutique options.

I merely suggested the boutique option for those who want a simpler output stage as opposed to using six (expensive?) surface mount opamps. One lead to the positive leg, other lead to output jack. Six opamps in that casing along with switches/jumpers for hardware control on that tiny board is going to be tricky for this EAGLE-newbie.

[error401]

Originally Posted by joneeboi

Thanks for the help. It sounds like we might have to copy Twisted Pear more than I'd prefer. IIRC, they have the option for all three, right? I'm thinking of regulating USB and forgoing the whole battery mess.

That makes things easier. You should still watch your current consumption though, the iPod will suck some and really you shouldn't draw more than 500mA from USB. Since you'll need a boost converter, 50mA on a dual-opamp probably converts to 100mA or more on the USB supply.

If you're willing to make this non-portable (by depending on USB), why not use a higher voltage external supply. Then you don't need the onboard switcher and don't have to worry so much about current.

With the boutique caps issue, I understand you can block DC with any appropriately spec'ed cap; the boutique bit would only be useful for preventing the use of poorer spec caps. Their sound difference is doubtless overemphasized and almost deified/idolized in some sense, and I can understand wanting to shy away from them. Nevertheless, I'd still argue they make a difference when placed directly in the signal path just like switching in lower-quality parts, audible or not. Is there something I'm missing? I'm sure the layout can be implemented such that their inclusion doesn't interfere with conventional non-boutique options.

Well, no reason to say 'NO BOUTIQUE CAPS ALLOWED', but it's not going to be easy to do a layout to accommodate them, since they're so long. I think you're pretty much going to have to put pads for a small poly film cap and if you want to use boutiques, you'll need to airwire them. Twisted pear uses OSCONs for output; not exactly boutique .

I merely suggested the boutique option for those who want a simpler output stage as opposed to using six (expensive?) surface mount opamps. One lead to the positive leg, other lead to output jack. Six opamps in that casing along with switches/jumpers for hardware control on that tiny board is going to be tricky for this EAGLE-newbie.

Good opamps don't have to be expensive. OPA134 is fairly cheap, and sounds good. There are lots of others out there that aren't prohibitively expensive, but sound good. And you don't need 6 parts, you need 6 amps. You could use a quad amp in SO-14 and a dual amp in SO-8 for all the amps you need. Just two packages to route . If you go this route you'll probably still need a separate output stage though, so volume control can be done.

I can't really think of an easy way to switch between balanced and SE output stages. I guess it could be done by substituting some resistors around the positive side of the bal->SE and taking the output directly. How are you going to do volume control here? Maybe you can come up with something, but with the space considerations, it's probably better to just choose an output topology and go with it.