[Xakepa]

Last mod of my AV-710 sound card turned bad - I fried the 5V LDO regulator. Input is +12V from the PCI bus; it's bypassed on the input with generic 10uF/16V elecrtolytic, and on the output with two of those 10uF/16V in parallel. Regulator is AMS1117-5.0 LDO.

What I did was to replace the input cap with Pani FM 470uF/25V, and the output cap with Illinois Capacitor AFX (os-con) 33uF/16V. Those 5V feed the analog section of the DAC; DAC is modded with two 47uF os-con powercaps (AVDD and AGND) on the analog side. The DAC is driving 10K pot in PIMETA.

What I suspect is that the ESR of the output cap was too low and the regulator died on oscillation (it went very hot). IlCap lists the ESR of 33uF/16 as 3.5ohms at 120Hz (70miliohms in 100-300kHz), and Tangent is mentioning ESR of 1-2ohms as sufficient for LDO.

Now I'm starting from scratch - what types of caps and values would you recommend. I'd rather be safe than sorry this time around.

Thanks.

 

[Garbz]

How about a 10uf 16V on the input and a 22uf cap on the output.

The datasheet says a lot about stability. Capacitance is required. They suggest a 22uf TAG cap, which are intrinsically low-esr, so I doubt the ESR of the OS-Cons caused the problem. That said I'd keep output capacitance low. IMHO increasing capacitance just increases problems in many circuits.

Input capacitance could be increased. I don't see anything that could have caused the cap to fry capactiance wise. Are you sure you got the polarity and the soldering right?

 

[tangent]

Quote:

Originally Posted by Xakepa Tangent is mentioning ESR of 1-2ohms as sufficient for LDO.

I don't say "sufficient" on my site, because that would imply that lower is okay. My power supply docs say the ESR must be between these two values. Lower and it oscillates. Higher and it oscillates.

In any case, I don't talk about the 1117 on my site, so extrapolating from what I do cover without checking it is just begging for trouble.

Quote:

Originally Posted by Garbz I doubt the ESR of the OS-Cons caused the problem.

I'd bet against you on that. In the LM1117 datasheet, it says the ESR must be between 0.3 and 22 ohms. The AMS1117 datasheet doesn't give an ESR value, but I'd bet it's the same way, so the cap's ESR is too low. Some ESR is absolutely necessary with almost all LDOs. How much depends on the regulator design. If this regulator would work into arbitrary ESR loads, they'd have given it a different part number.

Rather than change the cap, why not change the regulator? There's no good reason to use an LDO here. If you want to feel that you're doing something high-spec, use a high output current regulator to get lower output impedance, like the LM338, instead of the LM317.

 

[Garbz]

hey you are right. Stupid vague datasheet I was reading. Why can't the entire world just use the same one.

 

[mono]

I'd be checking to see if that reg has shutdown protection. How long did it work before failing? I'm just trying to rule out the possibility of some other problem.

Anyway, you have another option. Those two caps in parallel on the output, aren't actually in parallel. "FB13" (ferrite bead) is after CE12, before CE13. You could either replace FB3 with your choice of resistor (value), or break the plane in half and bridge it.

To bridge it I mean, look on the card where the text "FB13" is. If you cut a horizontal line barely above the text, starting about 3mm before the "F" and extending horizontally to the right about 12mm, it will divide the copper there such that FB13 is connected to the regulator output but now disconnected from CE13. By scraping off the coating to expose copper, you can solder a resistor bridging this cut to allow keeping FB13 and still having added resistance before your very low ESR cap.

However, there's another reason I wonder about this failure. Both before and after that FB13, in parallel with CE12 there is a SMT ceramic cap, CB20. After the FB13 in parallel with CE13, there is another SMT ceramic, CB21. Their ESR is certainly lower than your Illinois AFX.

 

[tangent]

Quote:

Originally Posted by Garbz hey you are right. Stupid vague datasheet I was reading. Why can't the entire world just use the same one.

That's one reason National charges more for their parts than the knockoff vendors. They spend that money on things like better datasheets.

Quote:

Originally Posted by mono After the FB13 in parallel with CE13, there is another SMT ceramic, CB21. Their ESR is certainly lower than your Illinois AFX.

Yes, and the ferrite breaks that theory. As frequency goes up, its impedance goes up, so the regulator doesn't "see" the falling ESR of the ceramic.

I haven't seen the schematic, but I can tell you that the ceramic is there for some other purpose. They put the ferrite there to let them have their cake and eat it, too. Why they didn't just use an NPN regulator, though, baffles me. They created a problem that they then had to solve. Weird.

 

[Xakepa]

Thanks for the input, guys

Mono, AMS1117 have a thermal protection and it definitely kicked in after about 1h. The chip was very hot, above 100C. I thought bootstrapping is overloading it - it was feeding 3V LDO->entire digital part; so I disconnected it from 3V reg. Maybe 2h later it switched off in a similar way, but that time comp refused to start untill I plugged the card out. I waited 10 min, checked everything (was OK) and replugged it for no avail - LDO didn't heat up at all. There's no short b/w terminals even now, but again comp won't start. It could be a slopy soldering job (I was dumb enough to try to solder on the ground plane), but ... I don't know

CB20 and 21 are puzzling me. My Fluke gets OL for capacitance, but reads 361ohms resistance on both, which doesn't make sense...From what I understand, if those are ceramic caps, they have far lower ESR than the electrolytics, so (even with FB13) because CB20 is right on the output the ESR is already too low for the regulator to be stable...

Could anyone make sense why the output is so involved? Pair of caps, FB, another pair of caps...seems like the same thing for 3V reg - pair, FB, single ceramic cap...what it's all about?..

 

[mono]

Quote:

Originally Posted by Xakepa Thanks for the input, guys Mono, AMS1117 have a thermal protection and it definitely kicked in after about 1h. The chip was very hot, above 100C. I thought bootstrapping is overloading it - it was feeding 3V LDO->entire digital part; so I disconnected it from 3V reg.

I recently acquired another AV-710 and unlike my other one, this time I'd decided to mod the supply caps. I have 470uF 'lytics before both regulators, the original CE12 after Q1 (AMS1117), then FB13, then a 47uF tantalum in CE15, another 47uF tant down the supply track in pseudo-parallel (CE18, upper left corner of the Via DAC), then Os-Cons around the Wolfson DAC.

I had seen a thread where this bootstrapping was mentioned but ultimately avoided doing that to mine- but I did upgrade the caps quite a bit. I haven't plugged mine in though, I was considering putting RCA jacks on the bracket when I came across your post and haven't yet done anything with the output so it wasn't done, but if I'm going to fry AMS1117 maybe I'd be as well off to do that now rather than after more mods.

I don't know how much current this card consumes in total but that bootstrapping might be a bit much for AMS1117 or at least, as implemented without much SMT 'sinking. If I end up frying mine, I think I might have a 78L05 or other drop-in replacement around here somewhere, just a matter of finding 'em.

Quote:

Maybe 2h later it switched off in a similar way, but that time comp refused to start untill I plugged the card out. I waited 10 min, checked everything (was OK) and replugged it for no avail - LDO didn't heat up at all. There's no short b/w terminals even now, but again comp won't start. It could be a slopy soldering job (I was dumb enough to try to solder on the ground plane), but ... I don't know

That's one of the things I was wondering, maybe an intermittent short was the culprit.

Quote:

CB20 and 21 are puzzling me. My Fluke gets OL for capacitance, but reads 361ohms resistance on both, which doesn't make sense...From what I understand, if those are ceramic caps, they have far lower ESR than the electrolytics, so (even with FB13) because CB20 is right on the output the ESR is already too low for the regulator to be stable...

It makes sense because they're still mounted in-circuit. I don't know what their value is either but I'd guess 0.01uF, that along with FB these are HF filtering.

Quote:

Could anyone make sense why the output is so involved? Pair of caps, FB, another pair of caps...seems like the same thing for 3V reg - pair, FB, single ceramic cap...what it's all about?..

Remember that a PC not only uses a switching PSU but has several fast power consumers. There's significant high frequency noise to be rid of. One of the things I almost did, but just seemed too "over-the-top", was to completely sever the lines to the system 5V and 12V PCI contacts and put a power connector on it to run from off-board power source. One thing at a time though, I'm not convinced it's necessary either but I got this card for almost nothing, so,

 

[mono]

I plugged my AV-710 back into a system with it configured as;

12V system rail -> 470uF 'lytic -> AMS1117-5.0 -> stock 10uF/25V paralled w/SMT ceramic -> stock ferrite FB13 -> 47uF/10V tantalum paralleled with stock SMT ceramic.

I doubt it sounds different.

Nothing stands out.

I kept both LDOs separate and measured the top, casing temp of the AMS1117-5.0, it was 56C, or 32C above ambient. Measurement might be a bit lower than actual but I could touch it without skin loss...

I'm wondering if the bootstrapping on the 1117/3.3 was just too much current->heat for Xakepa's card to handle. 12V down to 5.0 is quite a drop for entire card's power considering it's a SOT-223 part without a lot of copper under it.

 

[Xakepa]

Glad to hear that! Enjoy!

I had the same expirience as you. Hardly anything changed, I just think the "grain" in the mid-highs of my BGs was a bit less.

IMO, output capacitors mod is by far the most rewarding

 

[mono]

Quote:

Originally Posted by mono I doubt it sounds different. Nothing stands out.

Previously I had my card in an old testbed system and didn't think it sounded any different. Now I have it in a more modern system with sufficient reserve processing power to use Foobar2000's PPHS Resampler, Ultra Mode. It may sound slightly better, I can't be sure if it does or it was the resampling or further listening or ??? So I don't discount the possiblity that it might have a minor improvement but it's slight enough that I can't be sure. I'm fairly sure it doesn't sound worse.

 

[Garbz]

I doubt that applying filters to the digital system to alter it and make one sampling rate fit inline with another that isn't a direct multiple of the first has any audible benefit really.

Oversampling yes, upsampling / resampling not much of a believer really.

 

[mono]

Quote:

Originally Posted by Garbz I doubt that applying filters to the digital system to alter it and make one sampling rate fit inline with another that isn't a direct multiple of the first has any audible benefit really. Oversampling yes, upsampling / resampling not much of a believer really.

It should be noted that I was using the analog output. I think the general theory is that upsampling at high enough precision can exploit the DAC's better resolution.

 

[Xakepa]

That's a bit tricky. My entire music collection is 16/44 redbook rips; if I try to upsample to 88kHz (2x44) kernel streaming, card drivers can't cope with it - it's too high for "2 channel mode", and not compatible with "2 Hi-Res" which defaults at 96.

I wasn't able to overcome this, so I settled for NOS KS - resampling to 96 sounds harsh to me.

If one can advice how could I run AV-710 at 88kHz KS I'll be very happy

 

[Garbz]

Quote:

Originally Posted by mono It should be noted that I was using the analog output. I think the general theory is that upsampling at high enough precision can exploit the DAC's better resolution.

The problem I have with it is that no extra data can really be added. DACs do perform better at higher sampling rates, but that is what oversampling is for where the output is in multiples of the input and thus no filters need to be applied to approximate regardless of precision. I wonder have you tried upsampling to 88.2khz instead of 96khz? There would be best of both worlds there.