Delta 1010 Clock Mod

[mtl777]

Someone did a clock mod on my Delta 1010 PCI card by replacing the clock with his own improved design that uses a PI filter. The improved clock is on a small board that is attached with wires to the original card:

http://home.ca.rr.com/mtl777/Files1/Delta 1010 PCI Card BLA Mod.jpg

It uses a United Chemi-Con MVH series 2200uF 16V 125 deg.C aluminum electrolytic surface mount capacitor (see the big cap in the picture) which has been retrofitted with leads in order to attach to the board. My question is, why was a 125 dec.C cap used? Does it really get that hot in that area of the board? If not, what could have been the specific reason for this choice of cap when there are many other 2200uF 16V caps with lower impedance and higher ripple current rating that might be better. For example:

Nichicon HN - 8 mOhms impedance @ 20 deg.C / 100KHz; 3370 mA rated ripple
Rubycon ZLG - 12 mOhms impedance @ 20 deg.C / 100KHz; 2900 mA rated ripple
Panasonic FM - 15 mOhms impedance @ 20 deg.C / 100KHz; 3190 mA rated ripple

The MVH cap, in comparison, has 100 mOhms ESR @ 20 deg.C / 100 KHz; and only 1200 mA rated ripple.

I am very particular about this part of the soundcard as it has a huge effect on performance. Will I be able to improve it by replacing the MVH cap with a Nichicon HN, for example? What other caps would you recommend as replacement?

Thanks!

 

[kevinkr]

I think you might want to take a look at the way this thing has been installed instead - you'll likely net a bigger improvement by installing it with the shortest possible lead lengths, particularly the oscillator output and ground leads.

In the case of the capacitor there could be all sorts of reasons why that capacitor was chosen, none having anything to do with the temperature rating. ESR does not tell the whole story particularly in high frequency applications, ESL plays a large role, and the cap ESR may play a role in regulator stability not to mention damping high q resonances in the capacitor itself. I doubt ripple rating is an issue in this application - it does not appear to be a filter capacitor following a rectifier.

It may or may not have been randomly chosen with no thought to the issues mentioned above.

Best advice would be to try some of the alternatives mentioned after looking at supply noise with a good scope and see if you can improve it. Keep the original in case no improvement is found or it in fact performs worse.

 

[mtl777]

Thank you so much for your advice!

How many volts is the supply to a PCI card? Isn't it only 5 volts? If so, I can use an even smaller cap that's rated for 10 volts and it would be fine, right?

 

[Groundloops]

Hi man, glad to hear your 1010 is in working condition again. I totally agree with kevinkr. Low ESR is not per se improving things. Put the clock card as close as possible to the PCI card, that´s all what I would do. In the current state the wiring forms nice antennas picking up HF dirt.

For further improvements read this; it will not be thoroughly related to your PC but there are a lots of worthwile suggestions: http://imageevent.com/cics/v03theartofbuildingcomputertrnsp

You could also try something like my cookie tin shield:

 

[mtl777]

Well, my 1010 isn't totally fixed yet, but at least, damage to additional channels has stopped. I think it was a bad cable that was the culprit. I will need to replace SMD caps and resistors to fix it completely. Do you happen to know the type (tantalum, ceramic, etc.) and value of those cream-colored SMD caps in the 1010? They all look the same. I hope they're all identical.

Thank you so much for that article you referred me to! There's lots of useful info and good tips in there. My hands will be full trying them out.

That cookie tin shield you made looks awesome. How were you able to attach it securely to the card? It scares me to attempt doing something like that and risk the possibility of shorting something. I imagine, that metal needs to connect to ground somewhere. Where do you have its ground connection?

Thanks!

 

[Groundloops]

Time to go for reasonable LCR meter ...

In reading direction from underneath the PSU caps:

C25 0.44 µF
C20 0.90 µF
C21 1.15µF
C15 0.90µF
C16 1.15µF
C40 0.44µF
C57 1.15µF
C38 0.26nF
C56 1.15µF
C33 1.15µF
C6* 1.15µF (* unreadable)
C65 1.15µF
C32 0.10µF
C58 1.15µF

The cap values repeat in order to equal DAC and opamp structures. You may also do some reading in this: http://www.asahi-kasei.co.jp/akm/en/product/ak4393/ak4393.html
You will find that the 1010´s layout is quite close to what AK suggests.

For SMD resistor marking read this: http://www.talkingelectronics.com/projects/SurfaceMount/SurfaceMount-P1.html

Before you desolder those litte bastards you should compare your readings with mine. All the SMD parts are also fixed with kinda superglue and it´s a pain in the *** to get them off. Also the solder pads are not very stable on the PCB, they peel off easily when getting too much heat and then you are f****d.

I put some plastic sheet around my PCI card for insulation, so there is no possible shorting to the cookie tin sheet. That sheet is contacted to the PC chassis with a soldered litz and a screw.

 

[mtl777]

Awesome! Thank you very much, you're so very helpful! I'm so glad to find help on this project that I'm about to lose hope on.

Now if I could only desolder those little buggers. Why did the manufacturer put glue on them? It's an extra step in the mfg process, and what for? I hope not all 1010's are like that. BTW, do you know if those caps are ceramic or tantalum or what else?

Thanks again!

 

[Groundloops]

Sorry, I forgot: All the caps are tantalum. The glue is for fixing them before soldering.

I wonder how you made sure that the caps are defective ...

 

[mtl777]

I'm not really sure if the SMD caps are defective. I just want to try a "shotgun approach", i.e., replace all of them without regard to whether some of them are still good or not. I have already replaced all the large caps and still the problem persists. I want to try replacing the SMD components next. BTW, in this circuit, which component is more likely to fail -- SMD cap or SMD resistor?

Thanks!

 

[cuibono]

just a word of warning - boards like this are very delicate. I've been doing mods to soundcards for some time, and I can tell you they don't take well to fussing - and break very easily, and are quite difficult to fix. There is a pretty high risk/benefit ratio, and you have to be willing to lose the board. I have already lost one. I really suggest you repeat work someone else has done and who can spell out exactly what you need to do...

Good luck..

 

[cuibono]

On a more constructive note, why don't you try this:

http://www.diyaudio.com/forums/showthread.php?postid=199928#post199928
http://www.diyaudio.com/forums/showthread.php?s=&threadid=26597&highlight=

It is a well documented mod, and has a good chance of bringing improvements, with limited risk of losing your 1010 (you only have to remove a few parts). The only thing is you have to piece the instructions together from several threads - just search around this forum and diyhifi.org

The "shotgun" approach sounds like a sure-fire way to kill your 1010.

 

[mtl777]

Thanks very much for referring the Kwak clock! It looks pretty good. It will be great for a future project. For now, though, I just want to fix two damaged output channels in my 1010's breakout box. One of them is making some static noise by itself without anything playing. The other has a low output level, like 7 dBFS lower than normal.

Since the SMD components are so hard to desolder and remove, I've realized that the shotgun approach is not practical. I will have to buy a LCR meter and measure the cap values and replace only the ones that deviate too much from the values listed earlier by Gerd. And so, too, with the SMD resistors. The SMD resistors are easier to deal with since they are marked.

BTW, what good brand/series of SMD tantalum caps (approx. 5mm x 3mm size) would you recommend?

Thanks again!

 

[mtl777]

Sorry, I forgot: All the caps are tantalum. The glue is for fixing them before soldering.

Are the caps polarized or non-polar? By their looks, I'm guessing they're non-polar but I want to be sure. Thanks!

 

[kevinkr]

Are the caps polarized or non-polar? By their looks, I'm guessing they're non-polar but I want to be sure. Thanks!

I've never seen a non polar tantalum, doesn't mean they don't exist, but I suspect it is unlikely. Unlike conventional electrolytics tantalums are usually marked at the the positive end with a black band (yellow body), white band (black body), +++ or a dot, (very) rarely the negative end is marked instead. Check polarity in circuit with a meter if you are not sure.

Can't go wrong with kemet.

Note that tantalums aren't the best cap to use in the signal path due to linearity concerns. (See some of the old Walter Jung & Richard Marsh articles on passive components for a thorough discussion of the whys.)

 

[mtl777]

The SMD caps look the same on both ends, no markings that I can see. Here's the PSU section somewhere in the D/A board of my 1010 where you can see them (for example, C56 and C57):

An externally hosted image should be here but it was not working when we last tested it.

 

[cuibono]

what makes you think the caps are the source of your trouble? It seems like there could be a lot of possibilities...

 

[mtl777]

Well, there are 8 analog output channels. Only two of them have a problem. The other six are perfectly fine. In the two channels that have a problem, I have already replaced all the large caps. The only components that have not yet been replaced are the DAC chips, opamps, and SMD caps and resistors. I hope the DAC's are OK, and I think they are, because the channels are still playing music from wav and mp3. I imagine that if the DAC's are gone, the channels wouldn't play at all. That leaves the opamps, SMD caps and resistors. It could be any of these, but basically what I'm trying to do for now is to research the SMD caps because they are the only ones I could not identify -- they are unmarked! I'm not really saying that they are the ones that are defective. All I'm trying to do is gather all the info and help I need to diagnose the problem.

So my strategy is to test the SMD cap values in the defective channels and replace the ones that are deviating too much from what the values are supposed to be. Then I'll do the same with the resistors. If, after all these, the problem persists then I'll replace the opamps.

 

[cuibono]

the problem started when you changed out the larger caps, didn't it?

so doesn't it seem likely that the problem has something to do with the caps you changed, and not the untouched smd caps?

I would check all those caps very carfully. You could have dry solder joint, reversed one, or possibly torn a track or via up (which would be hard to spot).

You said you had scope, right? They are pretty easy to use, and will give you tons of good, real information. Hook up your probe to channel one, set the time/div to 1ms (2ms will make it easy to see 60Hz hum, .1ms will let you see lots of high frequency hash), make sure you can see your trace on the scope, connect the aligator clip to ground on the card, and start probing. Vary your volts/div from 5mV to 1V - for the noise on your output, try 5mV/div first. Try and spot the noise on the output first. Then compare it to a clean channel - with nothing playing on the cpu, make sure the clean channel has nothing on it (less than 1mV), while the problematic channel clearly has something (probably above a few mV). Take pictures and post them here. Trace up the problematic channel (in the analog section, around the opamps) seeing if the noise comes from all the way up to the dac output pin (look up the pinout on the datasheet, available on the manufactures website). You'll need the datasheets of the output opamps too.

It is really likely the noise is caused by something you've done.

Troubleshooting is fun!

 

[mtl777]

the problem started when you changed out the larger caps, didn't it?

No, it started after I changed the 1N4001 diodes in the PSU to MBR1100 Schottky's. Here's the thread depicting the history of my troubles ...

http://www.diyaudio.com/forums/showthread.php?s=&threadid=109365&pagenumber=4

At first, for a few days, my unit worked fine after the diode replacements. But one day a channel started having noise, then days later another channel started having the same problem. Initially I thought it was because of the diode replacements that I did. But later, I remembered that during the time when the problems started, my mouse would act funky and malfunction. I traced this to a bad breakout cable (connecting the 1010's breakout box to the PCI card) because my mouse worked fine again after I replaced the breakout cable. So I now strongly suspect that the bad breakout cable was the original source of my troubles.

 

[kevinkr]

I've never seen a non polar tantalum, doesn't mean they don't exist, but I suspect it is unlikely. Unlike conventional electrolytics tantalums are usually marked at the the positive end with a black band (yellow body), white band (black body), +++ or a dot, (very) rarely the negative end is marked instead. Check polarity in circuit with a meter if you are not sure.

Can't go wrong with kemet.

Note that tantalums aren't the best cap to use in the signal path due to linearity concerns. (See some of the old Walter Jung & Richard Marsh articles on passive components for a thorough discussion of the whys.)

The reason those caps have no polarity markings is that they are ceramic, and unless overvoltaged aren't likely to be the cause of the problems you are currently having.