E1DA Cosmos ADC

[IVX]

more reasonable for a hobbyist than an AP box

and a lot more precise 2.5mm interface has my Load-board + suitable cables.

 

[johny_2000]

This LPF prototype has two stages that can operate individually or in cascade, depending on the choice of micro switches.

 

[IVX]

This LPF prototype has two stages that can operate individually or in cascade, depending on the choice of micro switches.

Hmm, I see no way to cascade two LPFs with existing switches yet, do you? Or do you mean that would be useful?

 

[JensH]

With special cables it should be possible. But if it is useful, I don't know.

 

[IVX]

but why that be needed? H3 is already reduced for about 17db, H2 for 8db.

 

[JensH]

The only reason I can think of is if the DAC used has a relatively high level of distortion. Not your DAC of course.

 

[IVX]

The only reason I can think of is if the DAC used has a relatively high level of distortion. Not your DAC of course.

it is a good idea, I gonna try the $13 Meizu HiFi DAC with LPF

 

[IVX]

Meizu HiFi DAC with LPF. Actually, I can try to prepare a special audio file with 0/180dg L/R to use 2 unbalanced outputs of Meizu as one balanced. H2 should be reduced.

 

[IVX]

But balanced DACs are the best of course, this is 9038SG3 with LPF:

 

[BKDad]

There might be some advantage to adding a .33uF cap in series with the input of the LPF blocks. Cut down the lower frequency noise coming out of the DAC, maybe including some AC mains junk.

 

[BKDad]

Since I didn't feel like cleaning the lab bench this morning and I was curious, I simulated a BPF instead of an LPF for this application. Similar topology.

Here is the response at the AES17-2015 edges:

Obviously, the same kind of filter could be built for 10 KHz, and one for a bandpass of 18-21 KHz for making IMD tests.

Not suggesting anything - I just wanted to follow up on a comment I'd made a few days ago. I'm not sure the BPF is better than the LPF with a .33 uF in series with the input in a practical sense. I guess you save one cap per input with the BPF, if you count the 0.33 uF's.

 

[BKDad]

Hmm, I see no way to cascade two LPFs with existing switches yet, do you? Or do you mean that would be useful?

What I would think would be useful would be a selection of three BPF's. One centered at 1 KHz, one at 10 KHz, and one at 19.5 KHz. The first two would be to work with the APU for extreme measurements of 1 KHz and 10 KHz.

The one at 19.5 KHz would be for using two inputs from the two DAC channels, one at 19 KHz and one at 20 KHz for IMD measurements.

It turns out that if you design the BPF's for a Q of between two and three and cascade two stages, not only do you really get rid of junk, but the bandwidth is such that everything outside the AES17-2015 window is attenuated by more than about 11 dB at the edges and more further away from the edges, so the noise contribution of the DAC would be pretty low. The measurement would then be dominated by the filters' noise and the ADC residual.

Single stage BPFs are no more complex than LPFs. Of course, two stages would be double. But, it sure looks like there's room on your PCB for that.

Of course, a balanced to unbalanced output converter as you have now would be great as well.

 

[IVX]

BPF requires an individual adjustment, LPF doesn't. When I tested the very first-time harmonics filter(actually for my AP 4 years ago, AP's gen is too noisy), I made LPF and BPF both are MFB. LPF was better, as I remember, it was better for THD and for THD+N as well. Perhaps, it was too weak attempt and I had to retry that test but I see excellent LPF's results with no any trimpots.

 

[BKDad]

I just tried a Monte Carlo analysis of the BPF's. Assuming I got it right (!), it doesn't look too bad with 1% caps and .5% resistors.

Green is 500 runs for a single stage. Blue is 500 runs for two stages in series.

I'd attach the LTspice circuit file, but I'm not sure how to do that... I can't comment on distortion, because this is just a simulation and I'm not sure I trust the opamp models.

 

[AnalogSteph]

I'd attach the LTspice circuit file, but I'm not sure how to do that...

Just zip it or attach a .txt ending to the filename.

 

[BKDad]

Thanks to AnalogSteph , the LTspice file is attached. If you choose to run the simulation, you'll almost certainly need to place your own opamp component.

I just tried using an Analog Devices opamp in both the LPF circuit and the BPF circuit. Indeed, the LPF shows lower distortion than the BPF filter. As you said. So, probably a fully active BPF is a dumb idea. I need to look at that more.

But, adding a .33uF in series with the input, or 33nF in the case of a 10 KHz LPF, adds no distortion. It just isn't as effective at low frequency rejection as the fully active BPF. No tuning, either.

I also tried a slightly different BPF topology. It's not quite as aggressive in rolloff, but it produces pretty much the same distortion as the LPF design. At least in simulation.

 

[IVX]

commonly NPO=C0G caps have 5%.

 

[BKDad]

I see that Mouser has 1%, but they are probably not low cost enough or easy to find.

Anyway, I shouldn't be trying to tell you your business, which you are quite successful at.

 

[IVX]

huh, it is not really business, I just like to win that's it

 

[BKDad]

Well, I should only watch from the sidelines, not yell at the player.