0.300 VDC (from soundcard) on LM3875 GC input !!!

[Leglandu]

Hi gentlemen,

I’m currently assembling a LM3875 gainclone from audiosector.com.

Problem : checking for DC presence at the output of my Xfi Extrememusic PCI soundcard, I unfortunately found out a stable 0.300 V on the stereo channel (the other channels are clean)

It seems not to be a “static” voltage, my controller is a very old one with a needle. Strangely, it's the same value on left and right channel…

I checked the card's electronic circuits without succes, too small and complicated. Nevertheless I like this card and don't plan to change it

Have anybody already got the same phenomena, not necessarily on a XFI card ?

Is that dangerous for my FE168EZ drivers, as the GC kit is passing the DC ?

If yes, what do you think is better at the input of the gainclone, a capacitor or a tranny ?

Thanks for your advices...

[Nordic]

The input caps on the amp, should take care of that...

[Leglandu]

Thanks Nordic. Simple and direct answer… you just killed my problems in 1 sec, so it let me a little bit frustrated

My questions are because the kit is originally designed for a direct entry, so I would like to get some advices or experiences about adding a capacitor or a tranny at the input and the possible impact on the sound quality…

[AndrewT]

The question on sound quality is irrelevant if the speakers cannot survive the amplified DC your sound card is feeding them.

[Leglandu]

Hi AndrewT, don't worry, I understood I have to stop the DC at the CG input.
Just my interrest is concerning the choice between a capacitor or tranny... unless it's an infamous question

[EWorkshop1708]

Just use a big capacitor on the input to keep the bass. Instead of the standard 2.2 or 4.7 uf caps on the input, try 10uf, 22uf, or bigger depending on your input impedance. I used 10uf/100V cap on my 63K impedance input on my subwoofer amp and get bass that goes really low.

Don't go too big, so it don't take forever to charge the input cap, and you get a short DC offset in your speakers on startup.

[Leglandu]

Hi Eworkshops1708… 22uf, really ?!

I just done an extensive search in the forum (better to be done before…hmm..) and you’re right, usual values are 2.2 or 4.7uf, depending of the impedance of the CG.

Lower values don’t pass the LF, and phase error increase… but higher values present a risk of DC escape (the price escape also )

In my circuit (50K pot + 20K shunt + LM3875), depending of the volume setting, I estimate the input impedance to be 20K to 25K, so a 4.7uf would pass 1,5Hz @ –3dB (16Hz @ 0dB) with a low phase error...

http://www.tubes.mynetcologne.de/roe.../rc-filter.xls

4.7uf seems correct as the amp is dedicated to 6” FR drivers, not subwoofers.

But I keep in mind the idea of increasing the value to boost the bass if necessary… with cheap parts, before breaking the bank for that :

http://www.riviera-acoustics.com/cat...sort=3a&page=2

[Conrad Hoffman]

Remember that a lot of PC stuff runs single supply, and may already have caps. If you measure with a high impedance meter or scope, you might see a significant offset. Try loading the thing down with about 10kohms. I've also seen this on inputs, where if you don't sink the leakage current, there will be a large offset and even shut down the input circuitry. If you really do have 300mV, follow the cap advice.

[danielwritesbac]

I have that very same sound card.

Panasonic ECW-F2335JB is a 3.3uf polypropylene capacitor that will let everything through except the DC.
Even an electrolytic 3.3uf can work (+ goes towards the amp, - towards the source).

I believe that, as illustrated, the "load" is the 22k resistor that came with the kit. So, 3.3uf allows the amplifier to attempt to amplify non-audio signals.

So, a cap at 0.68uf will be "down by 3" at 10hz. Of course, that's misleading. Unlike engineers, most people would draw the line before it has any audio effects. That's flat to 21hz, and its the same 0.68uf cap.

If related to speaker designers, the "speaker" (load) is 22k (or whatever you're using for input load) and the cap is to decrease bass below 10hz.

A mighty fine layout for Soundblaster X-Fi in combination with LM chips is a layout of entirely 100k and 2.2k resistors (not including speaker zobel about 4 ohms).
According to National's design spreadsheet, this matches every part for 100k layout with 100k pot, with the Rb and Ri at 2.2k.
Somebody please correct if that's not right.

See my latest tinkering of LM1875 over at its thread, because here's a design that works great with X-Fi--and currently in service rocking down the house.
http://www.diyaudio.com/forums/showt...28#post1412428

EDIT: My main reason for suggesting that layout is the 2.2k in-series resistor (Rb) provides additional isolation (armor) for the amp to use an X-Fi source. That will protect the new input filter cap as well.

[Leglandu]

Conrad Hoffman, as per your advice, I tested with a 10K shunt : it doesn't kill the 0.300 VDC !

The Xfi has a very low output impedance (+/- 50 Ohm, I’m not sure) so you can plug in directly a headphone.

There is a lot of polarised caps near the female plugs but I was unable to find any continuity between one of them and the stereo output plug. Are some caps plugged in reverse ? Anyway I’m afraid that is too much sophisticated for my 40W soldering iron

The really strange thing is this concerns both channels, exactly the same DC on right and left ! Does it mean this is a "per design" phenomena !?