[f] AMB on Facebook

Channel and power supply configuration

beta22

Channel and power supply configuration

Postby johan851 » December 28th, 2017, 10:26 am

I currently have a γ2, M³, and a LM3875 chipamp for headphones and speakers. I'm interested in building a β22 to upgrade and replace both my M³ and chipamp. I probably listen to headphones about 80% of the time, but it would be nice to be able to use it for speakers as well. Just a pair of bookshelf speakers in a small room at moderate volumes.

I've been trying to do my homework as I think about this build. I stumbled on some background reading here and on head-fi - for reference, this post and this post.

I was planning to go with the recommended configuration for speakers - three β22 boards (single active ground channel) with a separate σ22 for each channel, offboard heatsinking for power supplies and β22 boards. I'm pretty sure that's going to be overkill for my application, actually, but it doesn't hurt to have headroom. After reading through posts like those linked above, I'm wondering if that's really the best option.

To briefly recap, the posts above discuss three configurations:
  1. Passive ground
  2. Active ground
  3. Active ground w/ supply rail current cancellation

If I'm thinking about this correctly, the jump from 1 to 2 (passive -> active ground) is a more significant improvement than 2 to 3. With a passive ground, the amp is dumping out of phase load current directly to the unregulated ground rail. Since the ground rail has real impedance and is used as a reference throughout the circuit, having its voltage bounce around is not ideal. With an active ground, that load current instead goes to a highly accurate buffer and current is sourced/sunk by well regulated power supply rails.

Going from 2-3 seems like a smaller improvement. Although the current on the supply rails is dynamic and will cause those references to move, this is mitigated by the very low output impedance of the σ22 and the high PSRR of the β22. A good argument for not using three supplies when you could use one (as in one thread I linked), but not a huge effect on performance. After all, this is kind of what we expect a power supply to do.

Am I thinking about this correctly?

In my planned configuration, with three supplies instead of one, I'll have situation 2. I guess it makes sense that I'll need to sacrifice something to coax the setup into speaker duty. I want to put the power supplies in a separate enclosure, and I'm planning on adding an LCDuino, δ1, etc., so the wiring is actually a little complex. I'm looking at $60+ just for powerCON jacks and connectors.

I'm wondering if I should go ahead and go to a four-channel configuration. If I did this, would it be reasonable to power speakers with only two σ22, one for each left/right pair? That would get back to balanced supply currents (configuration 3) on each power supply while using headphones. I don't have a balanced source, though I might build one in the future. For now, I could just have two active ground channels, and that would set me up well if I build a y3 down the road.

If one σ22 per side is sufficient, then...

Pros
  • supply rail current cancellation in headphone mode (class A)
  • less supply wiring from case to case
  • easy jump to a balanced input in the future
  • two active grounds instead of one working overtime to handle current from both channels
Cons
  • headphones need to be re-cabled
  • two δ1 attenuators instead of one
  • more complex input wiring

Of course, if one σ22 per side won't be sufficient, going to balanced or pseudo-balanced and four power supplies probably isn't worth the effort.
johan851
 
Posts: 72
Likes: 1 post
Liked in: 0 post
Joined: June 4th, 2010, 10:57 am
Location: Seattle, WA
Country: United States (us)

Re: Channel and power supply configuration

Postby amb » December 28th, 2017, 2:17 pm

Look at the "group build β22" at the β22 website gallery. That's a 4-board balanced configuration with two σ22 power supplies (one per stereo channel). I think this would be the best of all scenarios, even if you don't have balanced sources. You can convert your source signals from unbalanced to balanced (say, using a pair of α24s), or more simply, just connect the cold side input to signal ground. The latter of these two solutions makes the two cold side amps act as individual active grounds per stereo channel, giving you all the benefits of a 3-board configuration without twisting yourself into a knot with an odd number of amp boards and power supplies. It would also allow you to use one δ1 attenuator (even though that's not how the group build β22 was done, because balanced inputs are supported with a toggle switch).

This does require you to recable your headphones, but that's probably the best thing you could do to improve stereo crosstalk (i.e., there is no shared ground wire any more). To make your headphones compatible with other (non-balanced) amps, you can make a detachable adapter back to a TRS jack. It would also make driving speakers easier (with a 4-wire output), and increased output power (theoretically 4x compared to unbalanced, even though actual will be less). In your described application this should be more than sufficient.
User avatar
amb
Site Admin
 
Posts: 10359
Likes: 175 posts
Liked in: 479 posts
Joined: March 29th, 2010, 9:49 pm
Location: Sunnyvale, CA
Country: United States (us)

Re: Channel and power supply configuration

Postby johan851 » December 28th, 2017, 7:09 pm

Great, thanks! I've been wondering about this for a couple days, and I'm glad I was on the right track. I think this will be about as much work anyway, with better results.

Anything specific I should do on the σ22 setup? Should stick with 100VA transformers for each?

Also, if I include an unbalanced TRS connector, where should its ground be connected?

Edit - one more thing. If I anticipate using this in balanced mode in the future but connect the cold side to signal ground for now, I should configure those boards for the same gain as the hot side instead of unity gain, right?
johan851
 
Posts: 72
Likes: 1 post
Liked in: 0 post
Joined: June 4th, 2010, 10:57 am
Location: Seattle, WA
Country: United States (us)

Channel and power supply configuration

Postby amb » December 28th, 2017, 7:47 pm

100VA transformers should be fine. I think the group build β22 used 80VA units.

You can connect the TRS sleeve to one of the σ22’s ground. This is not ideal hence my recommendation to rewire your headphones to a 4-wire setup.

You can set the cold side boards’ voltage gain the same as the hot side.
User avatar
amb
Site Admin
 
Posts: 10359
Likes: 175 posts
Liked in: 479 posts
Joined: March 29th, 2010, 9:49 pm
Location: Sunnyvale, CA
Country: United States (us)

Re: Channel and power supply configuration

Postby johan851 » December 28th, 2017, 8:16 pm

OK, that's what I figured on the TRS grounding. I know the TRS configuration isn't ideal (unbalanced supply currents) but it'll be nice to have a TRS jack just in case. I'll re-terminate the headphones I have.

Yes, I saw 80VA transformers on the group build. And on-board heatsinking, too...
johan851
 
Posts: 72
Likes: 1 post
Liked in: 0 post
Joined: June 4th, 2010, 10:57 am
Location: Seattle, WA
Country: United States (us)

Re: Channel and power supply configuration

Postby johan851 » November 3rd, 2019, 10:31 pm

Resurfacing this thread, as I'm revisiting this build. I can't believe it's been two years, but this is what happens when you have kids. :shock: The balanced dual mono build was a slippery slope and resulted in a σ25, LCDuino-1, δ1, and more being added to the BOM, along with the associated complexity, size, and expense. I never got it started. I recently had to scale back my speakers, and I want to scale this project back as well in order to have any hope of getting it done.

In trying to understand what "light speaker duty" means for my situation, I plugged my desktop speakers into my M3 headphone jack yesterday and carefully cranked the volume while monitoring the internals with an infrared temperature gun. This M3 has a STEPS PSU (24V, LM338, 15VA transformer), 2" heatsinks for the LM338 and M3 output transistors, and 6x gain. I reached the necessary levels near the top of the volume dial. The LM338 heatsink stabilized around 60-65C, and the M3 ground channel heatsinks warmed to only 45C. If the little 15VA transformer and LM338 handled that, I imagine just about any flavor of β22 + σ22 is going to hold up fine.

As one of my goals is to drive current and future near field speakers well, I scaled down to a simple dual-mono passive ground build: two β22 boards, each with its own σ22, and a shared transformer in a separate enclosure. This seems like a nice configuration for speakers. The footprint can be fairly small, and the symmetry is attractive. But it seems like too much of a compromise for the headphones, even with the option to re-cable them to take advantage of separate grounds. I kept coming back to the elegance of the reference build.

I spent a couple more afternoons playing with options, and finally came across the Mini Dissipante 3U enclosure. It's a hair bigger than I'd like, but the large heatsinks can fit either a β22 or a σ22 board mounted vertically on each side plus two more boards on the bottom. The heatsinks have a C/W of ~0.4. I spent a few hours with paper and this very handy heatsink calculator (it even lets me put in the Thermasil III pad specs) and I'd like a sanity check on the following configuration, which I think gets pretty close to ideal for my purposes.

Right heatsink: σ22 mounted vertically. At 0.4 C/W (or ~1.6 C/W per FET) and Tj(max) at 100C, the thermal limit is 14W per FET. At ~8V drop across the transistor, that means 3.5A continuous current available per rail. One big advantage of the external heatsink is that the ambient temp is about room temp rather than chassis temp.

Chassis bottom: two β22 boards for L and R channels, using 2.5" onboard heatsinks.

Left heatsink: one β22 ground channel. With a heatsink this big, I think it can even provide the ground for speakers. Digging into the heatsink section of the β22 instructions, I calculated max average current for each L/R channel at 0.45A (+/-30V, Po=18W, i=1.5A, i*0.3=0.45A). The ground channel has to source that current from L and R with no voltage swing, which is rough, and most of the power is dissipated across the amplifier MOSFETs. Using the heatsink calculator again, it looks like the thermal limit would be around 15W per amplifier FET, which is 0.68A average (15W/(30V-8V)), or 0.34A average per channel. Reversing the average current calculation using 0.34A per channel for L/R, I get Po of about 10W max per channel. Not the full 18W, but plenty for these particular speakers.

If I reduce the supply voltage to +/- 27V the ground channel capacity more closely matches the L/R channel max and everything gets more reasonable.

Does this approach seem sane? I'm worried that I'm not factoring the quiescent current into the averages.

The asymmetry sort of bothers me, but otherwise this would have all the advantages of the reference build while even allowing the ground channel to be used for speakers. I figure I can leave connectors in the chassis to allow switching the speaker ground from active to passive if future speakers are too much for the active ground, or even install a switch on the back for easy testing of both configurations. I think an active ground is preferable to passive ground dual mono, especially for headphones.
johan851
 
Posts: 72
Likes: 1 post
Liked in: 0 post
Joined: June 4th, 2010, 10:57 am
Location: Seattle, WA
Country: United States (us)

Re: Channel and power supply configuration

Postby amb » November 4th, 2019, 12:06 pm

I didn't try the heatsink calculator you linked, assuming it's correct, and those are the numbers it gave you, then it looks sane. Just keep in mind all these calculations are for steady state situations, while music is highly dynamic. The calculations serve as good reference but actual dissipation will be differrent with music, depending on the program material, the speaker load, and actual average output current. One thing to point out: even though the active ground channel does not swing voltage, the output MOSFETs power dissipation is determined by the voltage across each of the MOSFETs multiplied by the current passing through the devices. If it's sourcing and sinking the return current from both channels, they are additive if the signals are monaural and in phase.
User avatar
amb
Site Admin
 
Posts: 10359
Likes: 175 posts
Liked in: 479 posts
Joined: March 29th, 2010, 9:49 pm
Location: Sunnyvale, CA
Country: United States (us)

Re: Channel and power supply configuration

Postby johan851 » November 4th, 2019, 12:47 pm

amb wrote:I didn't try the heatsink calculator you linked, assuming it's correct, and those are the numbers it gave you, then it looks sane. Just keep in mind all these calculations are for steady state situations, while music is highly dynamic. The calculations serve as good reference but actual dissipation will be different with music, depending on the program material, the speaker load, and actual average output current. One thing to point out: even though the active ground channel does not swing voltage, the output MOSFETs power dissipation is determined by the voltage across each of the MOSFETs multiplied by the current passing through the devices. If it's sourcing and sinking the return current from both channels, they are additive if the signals are monaural and in phase.

Right, that's the assumption I used for the ground channel amplifier MOSFETS - full additive current from each of the left and right channel across the full voltage (22V for +/-30V rails). Thank you!
johan851
 
Posts: 72
Likes: 1 post
Liked in: 0 post
Joined: June 4th, 2010, 10:57 am
Location: Seattle, WA
Country: United States (us)

Re: Channel and power supply configuration

Postby johan851 » November 5th, 2019, 1:07 pm

Another question - is it at all worthwhile to add a four pin jack to the active ground setup? I'm thinking that re-cabling headphones and running separate ground wires all the way back to the ground channel board output would reduce crosstalk due to the impedance of the TRS plug, jack, and return wire. I could probably do all of that for about $50 or so. Neutrik rates the contact resistance of their locking headphone jack to be < 6mΩ, the wire I'd run is probably another 5mΩ or so... seems a little over the top, but I imagine it might at least halve the crosstalk resulting from a shared ground, assuming 10mΩ output impedance on the β22.
johan851
 
Posts: 72
Likes: 1 post
Liked in: 0 post
Joined: June 4th, 2010, 10:57 am
Location: Seattle, WA
Country: United States (us)

Re: Channel and power supply configuration

Postby amb » November 5th, 2019, 11:23 pm

Yes, it's a good idea, if you could live with having to re-terminate your headphones to use a matching 4-pin connector too. The resistance of ring contact on a 3-pin TRS plug is the most significant cause of channel crosstalk.
User avatar
amb
Site Admin
 
Posts: 10359
Likes: 175 posts
Liked in: 479 posts
Joined: March 29th, 2010, 9:49 pm
Location: Sunnyvale, CA
Country: United States (us)


Return to β22 amplifier

Who is online

Users browsing this forum: No registered users and 1 guest

Quick Links

AMB audio DIY resources
AMB audio shop