freeDSP-aurora - DSP with 8 I/Os, USB Audio, S/P-DIF, ADAT, Bluetooth and Wifi contro

[454Casull]

Originally Posted by sudermap

Unipolar supply is unfortunate though, since it limits the usable output to around 1-1.5vrms rather than the standard 4vrms expected with XLR. Still, this doesn't really affect anything to do with sound quality, just limits the max output by about 10dB. Not a big deal unless your power amplifier has low gain.

For what it's worth, the TPA325x chips need tons of gain with PFFB (Post-Filter Feedback) enabled, and that's how they get their best performance.

I think also the NC400 as well needs huge amounts of gain, probably more than could be obtained from a maxed-out balanced line output though.

[454Casull]

Originally Posted by sArj

That analog I/O is pure evil. A ton of capacitors in the audio path is really bad.
Why didn't you use the bipolar supplies?! A simple PWM inverter with a linear post regulator is so much cheaper than a huge bag of Nichicon caps and will gill give you so much better results in terms of frequency response, CMRR and crosstalk. It would even look so much cleaner and take much less board space.
This really destroys Aurora for me...

Is it that the capacitors are used to create virtual grounds or something? They're not actually used as part of low-pass filters, are they? KiCAD is not working for me right now unfortunately.

[sudermap]

Originally Posted by 454Casull

For what it's worth, the TPA325x chips need tons of gain with PFFB (Post-Filter Feedback) enabled, and that's how they get their best performance.

I think also the NC400 as well needs huge amounts of gain, probably more than could be obtained from a maxed-out balanced line output though.

Amplifiers using ICs that require high gains should have their own gain staging. If amplifiers can't accommodate consumer output levels then that's an amplifier problem, not a source problem.

Originally Posted by 454Casull

Is it that the capacitors are used to create virtual grounds or something? They're not actually used as part of low-pass filters, are they? KiCAD is not working for me right now unfortunately.

They're used as high-pass coupling caps. Since the design is single supply they're needed to avoid a DC offset. As implemented they don't affect performance (re: measurements look fine / meet DAC spec).

[dspverden]

Originally Posted by Kyrre

@dspverden and whoever is competent:

To my post above (I could not find a way to edit it): So I need 4x i2s output instead of 1x TDM8, thus: Is the linked project (expansion slot to 4x i2s converter) perhaps the solution as it is, provided the TDM8 signal (or an alternative type of signal protocol) can be routed to the expansion slot? Is that possible?

(I guess not, and that some control function is selecting which i2S to send to the expansion, and then this extra board is just a multiple of the same i2s signal, but I had to ask in my ignorance.)

And just to be precise: I do not mandatorily need the analog outputs, I have my own solution for that. So the question should be answered in this context. In other words, if the TDM8 connected DAC on Aurora stops working, that will be acceptable, though not wanted. I imagine that could happen if one replaced TDM8 on board with i2s. I only need 4xi2s output on the HW level.

PS: According to an expert on AD forum, this very chip AD1452 is the one to use if you need conversion from TDM8 to i2S, so perhaps there is something here?

Thanks for any response.

No that doesn't work. If you want to split a TDM8 streams into 4 I2S streams you will need some logic. The stream format is similar but still too different. A little FPGA or CPLD will do the job.

Raphael

P.S. For all waiting for a bugfix, I feel really sorry that you have to wait so long. I had very stressful time the past two month in my main job. But it is getting more relaxed now, therefore, I can spend more time on Aurora again.

[454Casull]

Originally Posted by sudermap

They're used as high-pass coupling caps. Since the design is single supply they're needed to avoid a DC offset. As implemented they don't affect performance (re: measurements look fine / meet DAC spec).

Ah. I did mean to say high-pass instead of low-pass.

Still sucks that they're electrolytics though. With so many of them, it's only a matter of time (10-15 years?) before some have to get replaced.

[Studley]

@dspverden: Hi, I sent an email to your website email address but I’ve not had a response.

[DSP_Geek]

Originally Posted by 454Casull

Ah. I did mean to say high-pass instead of low-pass.

Still sucks that they're electrolytics though. With so many of them, it's only a matter of time (10-15 years?) before some have to get replaced.

Not the end of the world. You could bypass the output caps and apply the positive & negative outputs to the positive and negative inputs of a differential amplifier. It's the classic use case but instead of hum you're cancelling a DC offset. That'll get you a single-ended output, so for balanced output the negative diff amp simply swaps the input routing, pos out to neg in and neg out to pos in. The DC is still cancelled and you get a nice inverted phase signal.

IN THEORY since the inputs are differential pairs, by the same principle you should be able to simply bypass the input caps so long as both inputs are referenced to 0 volts. It will fail miserably if either input is left open. However, I have NOT tried this. Since most opamps aren't incredibly happy with inputs at either rail I'd want to breadboard a copy of the input circuit and test voltages at the opamp input pins to make sure they're at Virtual Ground (1/2 Vcc or 2.5 volts) as they should be, instead of Real Ground. If they're not, then the breadboarding just saved you considerable grief.

Or you could simply bypass the input electrolytics with some decent film caps, say around 2.2 uF. Unless your soldering skills are really on point I would advise against it, though, and any film cap with a high enough value to be useful would also have pin spacing wider than the electrolytics' pin spacing so the retrofit would be a bit of a hack job.

tl;dr: getting around the caps can be done -- so long as you know exactly what you're doing & you're okay with nuking the warranty

[454Casull]

Originally Posted by DSP_Geek

Not the end of the world. You could bypass the output caps and apply the positive & negative outputs to the positive and negative inputs of a differential amplifier. It's the classic use case but instead of hum you're cancelling a DC offset. That'll get you a single-ended output, so for balanced output the negative diff amp simply swaps the input routing, pos out to neg in and neg out to pos in. The DC is still cancelled and you get a nice inverted phase signal.

IN THEORY since the inputs are differential pairs, by the same principle you should be able to simply bypass the input caps so long as both inputs are referenced to 0 volts. It will fail miserably if either input is left open. However, I have NOT tried this. Since most opamps aren't incredibly happy with inputs at either rail I'd want to breadboard a copy of the input circuit and test voltages at the opamp input pins to make sure they're at Virtual Ground (1/2 Vcc or 2.5 volts) as they should be, instead of Real Ground. If they're not, then the breadboarding just saved you considerable grief.

Or you could simply bypass the input electrolytics with some decent film caps, say around 2.2 uF. Unless your soldering skills are really on point I would advise against it, though, and any film cap with a high enough value to be useful would also have pin spacing wider than the electrolytics' pin spacing so the retrofit would be a bit of a hack job.

tl;dr: getting around the caps can be done -- so long as you know exactly what you're doing & you're okay with nuking the warranty

Interesting, thanks for the perspective.

Bypassing the electrolytics is an viable approach, I think, even if the leads are way longer (i.e. stick out) because there shouldn't be very much EMI/RFI and the excess inductance probably won't hurt the function of the caps. 2.2 uF is a far cry from the 100 uF caps that are there right now, though. On that note, any idea why 100 uF was chosen as the cap value?

https://www.mouser.ca/ProductDetail/...FG4IpGWg%3D%3D

[DSP_Geek]

Why 100 uF? Haven't the faintest. 100 uF with 10 Kohms input resistance yields a -3 dB corner frequency of 0.16 Hz, whereas 2.2 uF at the same input R is closer to 7 Hz and 4.7 uF is about 3 Hz. The -1 dB points are an octave above the -3 dB rolloff so there might be a slight advantage to the 4.7 over the 2.2 if you want to hear blue whales

[tmuikku]

They form a highpass filter with the input impedance. the higher the capacitanse the lower in frequency, less phase shift for the audio frequenvies. 100uf is propably a convenient value, might be a size constrain, or same cap somewhere else as well for bulk discount and smaller inventory