After quite a bit of hard work, we have a new board for the audio ADC converter, made of 4 layers: analog signals at the top, digital signals at the bottom, ground and power in the middle. No shared vias, no duplication of power inputs, no nasty capacitors on the audio input path. And thanks to this redesign, we managed to make the board a lot smaller: 86 × 38 mm.
Next step: audio DAC board.
After careful consideration, I have decided to go for a single bias source, instead of two (left and right). The reference design for the AK5397 uses a single one, and I am worried that having two different voltage levels for left and right could create some unforeseen issues. This will make the PCB a bit less symmetrical than it would be with two levels, but better be safe than sorry.
Also, I have decided to get DVDDL and DVDDR (3.3 V) from AVDDL and AVDDR (5 V) respectively, instead of getting them from the +6V input, which should reduce the amount of power loss quite a bit. Finally, I have removed all capacitors from the audio input signals, as mentioned on this post.
I am now working on a 4 layer version of the PCB.
With a bit of luck, this should be the last major revision.
Following the advice of another mentor, I have decided to redesign my audio ADC module as a 4-layer board. This will give me the right impedance between power and ground, by putting them on two separate planes. The board will now be laid in the following fashion:
I just received my FPGA module from Knowledge Resources. In fact, rather than receiving it in the mail, I flew to Basel to pick it up from Mike Stengle in person. He has put together an elite team of FPGA experts who can help me tackle some of the challenges that I am facing, not just for the ISHIZENO i8, but for the real project that is motivating all this research: building a machine for STOIC.
On my previous design, I made a total newbie mistake: putting tantalum capacitors directly on signal paths. This was quickly caught by one of the experts who is advising me on the project, and I’ve decided to replace them with film capacitors. Unfortunately, the ones we need must have a 220 μF capacity, which is a lot for a film capacitor. As a result, their individual size is 28.0 × 42.5 × 41.5 mm, which is too large for our small ADC modules. Therefore, we’ve decided to mount them on the backplane instead. We will most likely use EPCOS film capacitors. Updated design coming soon…
Following the advice of a couple of mentors, I have decided to redesign the power section of my ADC module. The LM78L15 and LM79L15 used to convert ±18 V into ±15 V are too noisy, and using a linear regulator to go from 15 V down to 5 V would waste two thirds of the input power, while generating significant amounts of heat, which would induce additional noise.
A better solution would be to feed something like ±18 V and 6 V to the board, then use better linear regulators to get ±15 V, and keep the ADM7150 for going from 6 V to 5 V. Also, it would be best to get the 3.3 V that we need for the digital part of the ADC converter from the converted 5 V level, instead of getting it from the 6 V input.
To get the ±18 V and 6 V inputs, one of my mentors suggested that we use a regulator with high switching frequency (2 MHz) to filter out any remaining switching noise. And multiple such regulators could be used with frequency lock with a phase shift between them, delivering very low EMI emissions, and a virtually ripple free power source. This part will be done on the backplane or PSU board.
Awesome news: we just received permission from Soren Soekris, designer of the amazing Soekris dam1021 R-2R sign magnitude DAM, to order one of his DAC modules in order to get some inspiration from it. This should save us a ton of trials and errors.
We also got warned that:
Thank you Soren, very kind of you!
We’re done with the initial design of our audio ADC module. Latest changes include the addition of a U.FL connector for the master clock, a very clean ground plane, and a properly sized cutout. I’m especially pleased with the ground plane, which includes much less traces than with our previous iteration, and does not include any signals, be they digital or analog. Instead, it is used exclusively for grounding and power. Also, some extra space has been added between vias and traces, ensuring better isolation.
Next step: the R-2R sign magnitude DAM module.
After a very long week-end of work, I finally have a fully symmetric version of our audio ADC. Not only that, but I managed to cram low dropout voltage regulators to go from 18 V to 15 V, which will provide a stable power source for the four operational amplifiers. Now, the left and right bias lanes have their own regulators, and so do the left and right digital sections. This brings the total number of embedded regulators to eight.
The PCB shown above does not have any ground plane yet, but it will be added later this week. Nevertheless, I made sure that the two regulators used for the digital section could be powered through the digital section of the ground plane. I also left some room for two pairs of connectors that could be mounted on the PCB in order to feed the analog audio inputs, thereby bypassing the card edge connector for the most important signals. Doing so, the card edge connector would only be used for digital signals, clock signals, and power. In fact, I’m even thinking about doing something similar for the master clock signal. Now, I need to find the right SMD connectors for a pair of balanced inputs.
I also made sure to use larger traces whenever possible. This should help reduce impedance as much as possible. This is especially obvious when looking at the two pairs of left and right horizontal lanes visible on the upper part of the top layer (red) and used to carry the analog audio inputs: as straight and fat as can be…
Last but not least, I made sure that the bottom plane would be used only for grounding and short and (mostly) straight power lanes, at the exception of one low frequency analog signal, HPFE, which is used to enable the high pass filter. Routing this signal through the bottom plane for a few millimeters allowed me to make the design totally symmetrical, thereby saving up to 10 mm of length on the board. Everything else takes place on the top plane.
Coming up with such a tight design was a lot harder than I thought. But it’s not totally surprising when you realize than the board will be 100 × 40 mm, which is tiny considering the number of components that it must house, while using two conductive layers only.
After several weeks of trying, I finally managed to get in touch with someone at Penny & Giles and get pricing for the professional-grade PGFM9000 motorized faders. At $63 a piece for 100 units, they’re not exactly cheap, but they’ll be in line with the other components of the system, like the Avago and Bourns rotary encoders.
Happy…
ishizeno 