Axiom Home Page Forums General Discussion The Water Cooler PCB Fabrication

Has anyone dealt with any of the various online PCB (printed circuit board) fabrication companies? I was looking at PCB123, which seems ideally suited, as I don't want to spend a lot of money on software (they give you the software for free, but the design files can only be sent to them, not exported to be used with other manufacturers--although once you place the initial order, $50 extra does get you those files).

Also with all the surface mount stuff, I'll also be looking for a place that does assembly. PCB123 does recommend a few.

I wonder who holds the record for the most 0 reply threads. I think I'm in the running. Not even a, "what you building Chris?"

Seriously, what ARE you building?

I want to try to build a simple "digital preamp", something which can take an HDMI input (PCM only), split it into 8 separate channels, and send each to a Sabre32 Reference DAC running in mono mode, then have that output to balanced, XLR jacks.

I'd be happy to get it working with only a volume control. But I found my DSP code for a Linkwitz-Riley crossover, and I think I can get it to be phase coherent (that'll stop me from defending Pioneer's global crossover--got to run some simulations first). As long as I'm including a DSP chip, if it has enough buffer memory I can do time-alignment without any big deal. Plus at that point individual channel trims are almost free.

I've been reading design guides for these new, heavily integrated chips, and they actually seem easier to design with than the old ICs I used to play with in my teens (because everything is handled for you). Other than the sheer pin-count, and the fact that everything is surface-mount these days (both of which increase the difficulty of PCB design), I don't think it'll really be that hard.

In fact, if you didn't want volume control, and single-ended (unbalanced) outputs were good enough, you could actually connect a Silicon Image HDMI receiver directly to 4, stereo Sabre32 DACs, and just power the thing up without too many extra pieces.

Nice project. How much will that run you?

The chips are actually pretty cheap (e.g. $50 each for the DACs). Looks like the PCB fab will run about $160/board for a 4-layer design.

I guess it all adds up, but less than $1000 for the full DSP version, and probably about $750 for the volume control only.

Chris, i can ask a few people, i don't know if i will be able to come up with anything for you. But i can try.

I was going to ask what you were building, but was afraid I wouldn't understand what it was....

I found someone who is custom building DA boards using the DAC I want: http://www.twistedpearaudio.com/digital/buffalo.aspx (The I2S input is the format HDMI receiver chips for output of their audio streams.)

Their I/V (current to voltage) line stage also looks really nice:
http://www.twistedpearaudio.com/linestages/ivy.aspx

So who would be interested in a receiver with only HDMI inputs, no analog, no S/PDIF; PCM only, no Dolby or DTS, but up to 24-bit/192kHz; 16 channels of balanced, XLR outputs, no RCA?

It'll use a home-brew algorithms to turn up to a 7.1 channels into, rear, heights, wides, etc., along with multiple independent sub outs. In any combination you want. Say Charles wants to run 16 EP800s and nothing else, fine. Want Left, Right, Center, Left Surround, Right Surround, Left Rear, Right Rear, Left Wide, Right Wide, Left Height, Right Height, 4 Subs, and a ButtKicker, it'll do that too.

I've just about finished the DAC board schematic, it looks good. Picked top quality parts. You know the stereo output on the Oppo Special Edition? I'm using one of those DACs in mono, differential mode for each channel. Along with a pair of opamps designed to handle ultrasound imaging, or even radar. Giving it a dynamic range of 135 dB, with -120 dB of THD, virtually noise-free (especially into a balanced input). The problem is the DAC board itself already has a $500 bill of materials. That's not really bad for 16 channels, but that doesn't include the DSP board (I'm not sure how much horsepower I'll need yet for all these channels @192kHz), the power supply(ies), or the HDMI interface.

I'm trying to avoid having to license code from the big boys, because for a small run of electronics that'll be a huge percentage of the cost. I am afraid I'll run afoul of patents, even if I figure out the channel processing on my own.

Anyway, that's an update. I'll probably have all the DAC parts narrowed down soon (need to find an I2C repeater which will allow me to address multiple slaves which would normally all answer on the same address). I probably won't do much of the PCB layout myself. I found a company who will look over designs, give implementation ideas, design the PCB, and send it to manufacture. I have to pay them for the work, but that's the same as hiring someone at a company to do this job, but I only have to pay them once.

EDIT: Found the solution to the I2C problem: http://ics.nxp.com/products/i2chubs/
"Addressing selected devices if there are multiple devices in the system with the same I2C address"

I would definitely try one of those. After I have the money. Which would be after the Axiom 30th year anniversary celebration.

And now after making sure the basement finally gets done and I have Hesta Prynn DJ my Basement Grand Opening.

I look forward to hearing more about this project.

I realized I said "receiver" in my last post, but this is obviously a pre-amp/processor, with the XLR outs, and no internal power amps.

Speaking of power amps, that was a previous project which I never finished, because the PCM to PWM controlled amp chips I wanted to base it around materialized too late, and then with much lower specs than promised. But anyway, I was used to the power supply design for large devices. Where every Watt counts, and heat is can become a serious problem.

I started looking at power supply designs for the digital and analog stages in this device today. Even something as inefficient (but with very clean output) as a shunt regulated PS is feasible.

I'm narrowing down options now. But it looks like for the digital stage, I'll be using a transformer which can output 9 Volts from 115/230 AC, and then regulating that down to 5 Volts into the digital board. From there I'll be using very fine regulators to get the 3.3 and 1.2 Volt feeds where needed. The analog devices (basically just the DAC output, and opamps feeding the XLR jacks), will need a bipolar (+/-) PS ending up with just under 4 Volts. That has to be the absolutely cleanest power.

The system will be comprised of several boards. The DAC board I've talked about. Also the HDMI input board, which will connect to the motherboard, or DSP board mounted on the bottom of the chassis. The DSP board will connect over to the UI board which will provided the front panel controls. The DSP also outputs to the aforementioned DAC. The PS boards and transformers will be to the left and right of the DSP board (near the air vents in the sides of the case).

The DAC board will have 2 rows of 8 XLR jacks. I'm using vertical mount models, which would normally be facing up off the board. But instead the board will be vertical, parallel with the rear of the case. The XLR jacks themselves will be fastened to the back panel, so the solder joins will not be stressed on plugging or unplugging the cords.

The HDMI board will be above the DAC board. There'll be 5 HDMI ports, and 1 output. Similarly to how the XLR jacks were the mounting point of the DAC board, the HDMI jacks will be screwed to the back panel.

So I have basic ideas about the HDMI board, and only rough guess how the DSP board will go. The UI is still up in the air. Won't need too much, volume knob, menu/arrow/enter keys, input selector, and a display. I've not decided about an on screen display yet. That depends on how hard it is to encode into an HDMI compatible interface and format.

Yeah, I'm interested in this, but I'm also wondering how much I'd miss the other features in a pre-pro that aren't part of this design. But even if I don't use it as my actual main pre-pro, I'm still interested in getting one to compare and see how it works.

I'm still not sure on the assigning of the channel outputs, so maybe you could go into more detail on that. Also, how will calibration work? I'm assuming some basic adjustments will be included, right?

Also, I'm thinking for my personal uses I'd be more comfortable with 6 or 7 HDMI inputs, but 5 would work for now.

5 HDMI ins/1 out, is the biggest chip I can get right now. But it isn't even as flexible as the 4-in chip. I've not done much looking at the HDMI board yet. But I am considering the 4-in chip paired with a separate 1-out chip. But what I could do is use two 4-ins together, for 8 total.

I'm building this as a platform (actually, the way I've designed the DAC board, I could easily substitute in a cheaper un-balanced RCA, based design with 8 stereo DACs, or even two 8-channel DACs). I'm not ruling out features, just not going to bother licensing them at first. But the DSP chips I'll be using have enough power to decode Dolby/DTS lossless up to 96 kHz (at least), and run either Audyssey or Trinnov processing. I'm still looking at how to get a mic hooked up. But that's an analog in, and makes things more difficult. I could also possibly support HDMI Audio Return Channel, it's in the chips, just needs to be assigned to an input (but it could possibly have AC3 audio, which needs a license).

It'll definitely have distance delay compensation. There's 256 kB of ring buffer space per DSP. That's allows up 42.667 ms of delay for 8 channels (each DSP will be handling 8 channels). Multiply 42.667 ms by the speed of sound, and that allows for speakers to be up to maximum difference of 47.6 feet.

Individual channels trims, adjustable in 0.5 dB steps. The master volume can also use 0.5 dB steps. The neat thing is, it's even smoother than that. When you go up or down a half dB, the volume ramps in 32, 1/64th of a dB steps. Even cooler, that's not even linear, it uses a logarithmic acceleration curve to reach the requested volume. It (un)mutes the same way.

Also multi-tap FIR filters for crossovers. That allows for phase aligned, highly customizable settings. I'll be doing the crossover work in the main CPU, and handing the resulting channels to the two DSP chips. Well, that's my initial plan. Still need to figure out what chips, and how many.

That brings me to your first question. HDMI carries at most 8 channels of PCM audio. Most of the time these are configured as 7.1, but the .1 isn't true, it's a full frequency range channel. That doesn't matter for this discussion. Just know the player is decoding what ever is on the disc to PCM and outputting how ever many channels result.

Those channels hit the first chip. This is where I'd be running Dolby's, DTS's, Audyssey, Trinnov or my custom processing to create additional channels from the existing. It can look at everything about those raw channels, and manipulate them in different ways. Say your disc has a 5.1 track, but you've got a 7.3 setup. So the processor looks at the surround channel, and creates rears from that. Then it moves onto the bass processing. You have two subs in the front of the room and one in the rear. You want all the subs to get the LFE channel from the disc, so it's mixed into all three, but you want stereo bass for the front of the room, so the crossover of the center is mixed into both front subs, but only the crossover of the left main goes into the left sub, and right to the right. The bass signal from all 4 surround channels is taken into the rear sub.

That gives you a total of 10 output channels. The first DSP handles channels 1 to 8, the second 9 to 16. These might be assigned as:
1. Left
2. Right
3. Center
4. Left Surround
5. Right Surround
6. Left Rear
7. Right Rear
8. Left-Front Sub
9. Right-Front Sub
10. Rear-Sub
But there's nothing fixed about those assignments. You can plug into any jack and send that signal to any speaker. (The Trinnov code actually has discovery routines to try to figure out where the speakers are in the room, and auto-route the channels.)

The order of processing in the first CPU is:
Channel creation (from existing signals, dematrixing, or sinks which will receive crossover information or mixes).
Crossover processing, remove the low-cut info, and prepare it for mixing.
Mixing, combine signals from other locations into the final assignments.
Send the channels to their proper DSPs.

The DSPs don't seem to do much right now. They're just handling the delay alignment. But they'll also run any room correction processing, EQing, standing wave attenuation, etc.

Great stuff, Chris. This could be your calling.

This is not far from what I already do at my job. It's selecting hardware and software, combining them together in such a way that it solves the problem at hand. The pieces are a little smaller, but they still have well defined ways of interacting. That is what surprised me. How well defined the functions and sample implementations of these really complex chips are these days.

As far as licensing goes, how hard and expensive would it be to implement something like DLNA? Will this platform allow for firmware updates?

Firmware upgrades, absolutely. They're essential to any modern piece of equipment. The main CPU has both USB and Ethernet interfaces.

As for DLNA, I don't know where I'd do the A/V decoding. The CPUs in the platform are pretty much accounted for, I'd have to add another chip just to handle that. Plus, it feels like something better accomplished by a device designed for the task.

Don't worry about the lack of HDMI inputs. This pre/pro will allow multiple preset states. So it won't assume there's only one possible device behind each port. I had initially planned to only have one input, and require the use of an external switch. But every HDMI input chip already has a hub built into it these days. Still the idea stands, that it'll be easy to switch all available input configuration settings with the push of one button.

Very cool.



I thought about it after I asked it and came to the same conclusion. I figured it wouldn't be something you'd care to concern yourself when anyone could just get a player of some kind that already has it. Still, some of your ambitions regarding this project have surprised me.

I was trying some PCB layouts yesterday and today. Was using the PCB123 software which the one manufacturer provides free of charge. The biggest problem with it is that has such a small library of parts, and doesn't import any normal formats. So I spend most of my time drawing footprints, rather than laying out the board. I didn't even get to any complicated components yet.

Going to try something else now. DipTrace gets good reviews.

Shame the software which the chip makers provide library files for is all in the $4000 range. Maybe when I'm selling hundreds of devices a week...

Ouch.