Help: new truepath screeching noise

[ikoflexer]

I finally put together the truepath (after one set of fireworks in which the mosfets and ta3020 were gone), but although everything seems to measure fine, I got this screeching noise (sounds similar to a cicada, and loud) that I can't get rid of, so any suggestions would be greatly appreciated.

The setup:
- power supplies built by myself, CRC filters of high capacitance all
result in very tiny ripple (verified on the scope); the 56V PSU is the earth/ground point
* 56V PSU = ~ 38v-0-38v toroid + rectifier bridge + CRC filter
* 10V PSU = ~ 15v + rectifier bridge + CRC filter + 7810 regulator
* 5V PSU = ~ 12v + rectifier bridge + CRC filter + 7805 regulator

- changes from stock configuration
* 4 mosfets STW20NM60, since the original were fried (I am using these ones until the new stock arrive); these have quite high Rds(on) but audiophool mentioned in a different thread that I could use them; should any changes be made to the circuit? -- resitors of different value perhaps?
* 4 1N4148 diodes as D102,D103,D202,D203 since the original fried as well (until I get the murs120t3); the stock actually did not include the murs120t3, but some diodes that looked like regular switching diodes (little red ones)

The interesting thing is that the screeching noise starts about .5 seconds after un-muting. In that .5 seconds I can actually hear music and it sounds ok.

I should mention that I have checked and rechecked the soldering. The noise is on both channels, is very loud, and the mosfets get hot very fast.

Thanks in advance for any pointers.

[ikoflexer]

I added some more capacitance to the 56V supply and now the noise is more of a hisssss, like a tuner in between radio stations.

Checked the bias cap and almost all the other capacitors and they are fine. So, I'm out of ideas.

http://soconfused.wordpress.com

[BddB]

Hi ikoflexer,
perhaps you could provide a schematic or some pictures of your power supply setup.
Does the noise stop immediately when the Trupath is muted?
What BBM-setting are you using?
(Just a gut-feeling, I'm very doubtful about using the 1N4148 diodes.)
/B

[ikoflexer]

I will supply the schematics soon (need to leave the house now).

Yes, the noise stops immediately when muting. But does not start immediately after un-muting. In fact I can hear clear music for a few milliseconds when I switch from mute to active.

Tried all bbm settings, same thing. Now I left it to 80ns.

I already ordered murs120t3 and am waiting for them, as well as the original mosfets.

What worries me most is that when muted, there is no voltage on pin 32 (REF1). Also 0V on pin 19 (BIASCAP). It'd suck if this tripath is fried too.

[ikoflexer]

Here are the psus I'm using. All capacitor values in mF.

http://www.cs.toronto.edu/~relu/misc/tp_psu.jpg

http://soconfused.wordpress.com

[BddB]

The schematics seems to make sense, DGND and AGND separated, 10VN referenced to VNN
...and you double and triplechecked the polarity of diodes and capacitors?

Does anything get hot? (so that you can't touch it for more than a second or two)

What voltage do you [u]measure</u> for VDD/VNN floating (not connected) and connected to Truepath?
(I know you wrote 56V, but I can't get that you have more than 50-52V, with losses over rectifier and the 50 Ohm resistor in count)

But, perhaps you should wait until you get the MURS120's, so that you don't fry anything.[B)]

Perhaps you've got some pictures, preferably in high resolution?

/B

[ikoflexer]

Hi again. Just want to say that I really appreciate your help.

Today the murs and fets arrived. So, other than the power supplies, the rest is stock.

For the record, it seems that when unmuting the voltage on the REF pin is 1.1V and on the BIASCAP pin is 2.5V. Phew, maybe the chip isn't fried after all.

I got the pictures, in as high resolution as my camera could do. As they are too large to show here, it's probably better if I only give links to them:

pic1 (http://www.cs.toronto.edu/~relu/tpassembled/DSC00471.JPG)

pic2 (http://www.cs.toronto.edu/~relu/tpassembled/DSC00472.JPG)

pic3 (http://www.cs.toronto.edu/~relu/tpassembled/DSC00473.JPG)

pic4 (http://www.cs.toronto.edu/~relu/tpassembled/DSC00474.JPG)

pic5 (http://www.cs.toronto.edu/~relu/tpassembled/DSC00475.JPG)

pic6 (http://www.cs.toronto.edu/~relu/tpassembled/DSC00476.JPG)

[Scratchy]

First thing to do is to clean off the flux using ISO, as per this video.
http://tangentsoft.net/elec/movies/tt05.html
Some problems have been resolved by simply cleaning the PCB.

Also inspect the SMT soldering with a Loupe, you will be surprised at what you can see.
This is what I bought:
http://www.leevalley.com/wood/page.aspx?c=1&p=51092&cat=1,43456,43351&ap=1

These pictures were taken through this 15x Loupe with my digital camera:

http://scratchy.kicks-ass.net:81/Amp1-B/PCB-bottom_zoom.jpg

http://scratchy.kicks-ass.net:81/Amp1-B/PCB-top_zoom.jpg

BTW, this AMP is working.

***If it was easy, everyone would be doing it***

[Rephlex]

the mute delay cap (see sticky mute thread) is backwards...
(the pcb is wrong)

[thomaseliot]

What are the small white wires between ohmite resistors and the wire jumpers? The ohmite resistors would be soldered lifted by 1 cm.

When the problem is hiss in 99 on 100 cases it is caused by bad solderings: usually litz wire, input caps and comnnections.

[BddB]

ikoflex,
informative pictures ;)
There's a couple of things to start with.

Excessive amount of solder on the surface mounted devices (risc of solder bridges).
- remove with "soder wick" or similar and resolder.

Too little solder on the hole mounted devices (risc of a "cold" solderjoint and weak mechanical withstand)
- the solder has to wet all the way through the board and "bubble up" on the other side. This requires a fairly hot iron as the board is multilayer and dissipates the heat faster.
Keep the soldering iron in contact with both component pin and the copperplating at the mountinghole.

The toroid soldering looks suspicious. Make sure that every strand of the Litz-wire is uninsulated at the solderjoint.

The mosfets needs resoldering (see pt. 2)

You had some accident with this board and used wire to connect. That's OK, [u][b]but</u> for the sensesignal on bottom (long white wire) as this is matched with the corresponding signal for the other channel. The routed length is 52mm (the width is 0,5mm and thickness is 0,07mm) Use a wire of suitable gauge and length.

Clean the PCB as per the link in Scratchy's post.

Examine the board to make sure there aren't any solderbridges or poorly soldered joints.

If no component is defective, you should pretty much be back in business[8D]

Good luck
/B

[ikoflexer]

Scratchy, good suggestion. Not that I haven't cleaned up the board before, but I've resoldered the components so many times that I just became careless at some point and didn't do the general cleaning. Anyway, just did it last night. Noise was still there. Inspected again all parts with a 8x loupe. Honestly, I don't think I like pcbs where the traces are so closed to each other. Especially that I found on the first day a spot where the lacquer covering was off and there was a short from a diode pin to ground.

thomaseliot, the white wires, as BddB was right on the money, are my fixing after the "accident" when the first chip was fried. Several pcb traces got too hot and basically came off the board.

rephlex, AARGH!!! Is it only me who thinks this is insane? Even after a couple of looks at C311 it wasn't obvious on which of its sides the potential is higher/lower. The assembly instructions should at least mention to be careful about the polarity there. The funny thing is that while taking out the caps to check for polarity, my daughter, looking over my shoulder, said: "papa, pull that cute little one" and I said, "nah, that one's fine." Moral: always listen to your 7 years old.

Mr. B, I will do all that you said. For the sense signal wire, are you suggesting that put a wire on the other channel as well?

Do you guys have any idea how helpful it is to get such informative input and support, at this level of frustration? Tons. My wife thinks I'm turning into Van Gogh and cut my ear off... [B)]

Cheers, many thanks; I'll write another update tonight.

[BddB]

quote:Originally posted by ikoflexer

...
Mr. B, I will do all that you said. For the sense signal wire, are you suggesting that put a wire on the other channel as well?
...

Not needed. You can glue the new wire the same path as the original was, so you get the correct length automatically.
/B

[ikoflexer]

Well, there goes nothing. Done all that was suggested above and it still does the same thing. I noticed that the hissing becomes more of a hiss rather than a lower freq sound when adding more capacitance to the Vnn and Vpp. The oscilloscope show high amplitude and frequency oscillation on the output (speaker output); and lower amplitude oscillation on Vpp and Vnn, on V10 and on V5, only when unmuted. It's as though the tripath chip switching somehow gets spread to the rest of the circuit somehow.

I'm thinking the oscillations start either in the chip, or input/feedback circuit, so I'll take it a part at a time.

Does anyone know why C107 is 150pF and C207 is 270pF? In the ta3020 data sheet application example both channels have a 150pF cap in that position. Just curious.

[col_s]

quote:Originally posted by ikoflexer


Does anyone know why C107 is 150pF and C207 is 270pF? In the ta3020 data sheet application example both channels have a 150pF cap in that position. Just curious.


It is so that the two channels switch at different frequencies.

From the datasheet:
" CFB removes very high frequency components from the amplifier feedback signals and lowers the output switching frequency by delaying the feedback signals. In addition, the value of CFB is different for channel 1 and channel 2 to keep the average switching frequency difference greater than 40kHz. This minimizes in-band audio noise."

It is interesting that you say you see the same problem on both channels. There are many parts of the circuit that are duplicated for each channel and a fault in one channel may not affect the other (eg. input, output and feedback sections). On that premise, perhaps it would help to concentrate first on the common parts of the circuit. Probably the main one is the power supply and its various connection points and paths through the circuit. Of course, it's quite possible the fault has been duplicated on both channels...

[BddB]

Hmm,...
sounds like feedback, that the signal from output is fed back to the input and then is boosted beyond control. Try to keep input signals from output signals as far as possible. You must also consider the mosfets and toroids as part of the output signal.
Question: Why are the jumpers for BBM-setting that long? (they surely act like receiving antennas.)
Q2: Do you have a dummyload or speaker on the outputs (and have the inputs shorted) when you fire up the amp.
Q3: In your psu-schematic you have 2x18,8mF for VDD/VNN. I hope this is mF, like in milli Farad and not micro Farad(uF). (Otherwise you need ~1000 times more capacitance. In the PSU1-PS, I use 20000uF (=20mF) per rail.)
/B

[ikoflexer]

It appears my reply from an hour ago didn't make it to the forum.

OK; col_s, thanks for the note, somehow I missed this part in the data sheet. Indeed, noise of same amplitude and frequency on both channels is puzzling.

BddB, A1) Honestly it didn't occur to me that the bbm jumper length would somehow affect anything, but you're right, no reason for them to be long. I'll make them as to reach from hole to hole.
A2) I have tried resistor dummy load (had a couple of 15ohm 5W on hand), speakers, and no load; all with same result
A3) Yes, all milli Farad. I could add some more capacitance to the VNN, VDD rails.

I feel quite frustrated that I don't understand more about the tripath chip, kind of like working on one of these new cars, where there is so much of a black box and so little you can do on the outside.

If you guys have any suggestions what to test, and on which pins/parts of the circuit. I do have an oscilloscope and multimeter and such. This is what I tested last night:
- noise amplitude on the speaker output is about equal to rail voltage. It shows as high frequency spikes (I guess I should post some pictures)
- same noise appears going back in the circuit, on both pins of the toroid inductor; taking the toroid out does not change anything, the noise appears at the input of the inductor (which is no longer there)
- noise appears on the 5.6ohm resistor and the diode next to it, therefore on LO1 (and mirrored circuit in the other channel and hi/lo part
- IN1 and IN2 do not show noise, at least not that that amplitude
- MOSFETS do get warm after about 20 seconds of switching from mute to active; did not try to go longer
- tried replacing the 56V psu with a HP regulated variable power supply that I had around; tried different voltages between 36 and 56 volts, same thing

I'm pretty much out of ideas now.

[BddB]

Which frequency do you have on the noise and is it steady or does it vary in frequency?
How about some new pictures on your reworked, checked for short circuits and cleaned PCB, perhaps? ;)
/B

[ikoflexer]

I'll take some pics tonight. The noise at the output looks like a fuzzy line with spikes at about .25 mS, which would be roughly 240Hz (hm, 4*60Hz mainfreq)? Though I think that there are some higher frequencies in the noise. The amplitude of the spikes is about 2.25 * 20V/div, which is close to rail voltage.

The noise frequency does not seem to change over time.

[Scratchy]

According to some of the tripath whitepapers posted in the downloads section, these amps show a lot of noise above 30khz, but since this is above audible range, it only shows up on an O-scope. I have not check my AMP1B yet, but I suppose I should.

Also, the toroids should be filtering out this lower frequency noise.
If you are seeing the same before and after the toroids, then they are not working correctly.

***If it was easy, everyone would be doing it***

[col_s]

I stumbled across this via google today:

***
I have removed the links. It appears that you should pay for these articles. They are available for download from http://www.elektor.com for a small charge. There are three main articles plus a follow-up article detailing the output filter configuration. Search for "clarity" on the web site. There is some valuable information and some interesting circuits in there...
***

It's another TA3020 based implementation. It has a detailed description of the operation of the amplifier. Although it's probably not identical to the Truepath, I find the schematic particularly easy to read.

It has a description of the modulator stage of the TA3020 chip and the vboot circuitry. Maybe it might help with understanding how it works. Looking at the area where the pcb was previously fried and the symptoms of the problem, perhaps something in this area is amiss.

[BddB]

From the TA3020 datasheet:

"The audio input signal is fed to the processor internal to the TA3020, where a switching pattern is generated. The average idle (no input) switching frequency is approximately 700kHz. With an input signal, the pattern is spread spectrum and varies between approximately 200kHz and 1.5MHz depending on input signal level and frequency.
Complementary copies of the switching pattern are level-shifted by the MOSFET drivers and output from the TA3020 where they drive the gates (HO1 and LO1) of external power MOSFETs that are connected as a half bridge. The output of the half bridge is a power-amplified version of the switching pattern that switches between VPP and VNN. This signal is then low-pass filtered to obtain an amplified reproduction of the audio input signal."

The output filter forms a second order LP-filter, with F3 at about 100kHz.
BTW, .25ms is corresponding to F=1/0,25e-3 =&gt; F=4000Hz, (which is unaffected by the LP-filter).
Examine the circuitry before the output filter.
Is this 4kHz present at inputs, modulator feedback circuit,...?

Did you try another PSU?

/B

[ikoflexer]

scratchy, good point there. Interesting link, col_s, I downloaded the article and pcb to have a look at it but haven't had time yet.

bddb, yes, you're right about the frequency, I got it wrong. In any case, yes, I have tried a different psu (have a variable regulated psu) with same results.

However, I did have some progress. I got so sick of it last night that I replaced all the smd caps. Then by a fluke I measured the resistance across R118, R119, R218, and R219. All showed around 507k except for R218 which showed only a few ohms. I took out all the parts in that area of the board, cleaned up, put them back with attention to potential shorts. Plugged it all in and surprise, I got clean sound in one channel. I also found at least one source of the buzzing noise. Luckily L100 was a bit loose at its input side (coming from FB1) and when I moved it I got the buzzing noise. It turns out that there is high freq signal entering the toroid, as you guys mentioned, which was passed all into the pcb (probably ground) area around it when the two touched. It beats me why with so much space in that pcb area the spacing between the toroid contact and the rest is so small!!! Considering that this project is intended for hobbyists as opposed to industrial ultra-precise robots, the pcb could be made, IMHO, much better to minimize the possibility of mistakes. I'm just saying that this project is not inherently difficult; rather it is made much more difficult because it is so easy to make errors while soldering.

I thank you all for your help, it is much appreciated. The amp is not yet working; parts of the pcb started to simply come apart on me. In quite a few places the through the hole part of the pcb simply came off together with the unsoldered part (L100, a few caps, a diode).

It is not my intent to be negative, even though this project frustrated me very much. I do think though, if we do not express the shortcomings of a DIY project, then I do not see how any improvements be made. In my opinion both the documentation and the pcb can be greatly improved. I will send detailed comments to the 41hz, for what they may be worth.

I guess I'll just concentrate on making this amp work for now.

Regards.

[audiophool]

quote:Originally posted by ikoflexer



...

I do think though, if we do not express the shortcomings of a DIY project, then I do not see how any improvements be made. In my opinion both the documentation and the pcb can be greatly improved.

...


ikoflexer,

Congrats on getting as far as you did. The first one is pretty tough. I haven't been posting on your thread because plenty of others were helping you out, but mostly because I was busy working on Truepath assembly guidelines which may be of interest to you (although perhaps a bit late for you now!) I will start another thread about the guidelines.

Rgds.

[BddB]

quote:Originally posted by ikoflexer

...
It is not my intent to be negative, even though this project frustrated me very much. I do think though, if we do not express the shortcomings of a DIY project, then I do not see how any improvements be made. In my opinion both the documentation and the pcb can be greatly improved. I will send detailed comments to the 41hz, for what they may be worth.
...

This is NOT a beginners kit. You need to know how to solder properly and mount the components in right order.
An experienced person probably will find this board well structured and easy to work with, while beginners might have some difficulties.[}:)]
I'm sure that Jan will appreciate all kind of feedback, that helps to improve the documentation :)
Glad that you made progress, anyway ;)

Cheers
/B

[col_s]

quote:Originally posted by ikoflexer

Considering that this project is intended for hobbyists as opposed to industrial ultra-precise robots, the pcb could be made, IMHO, much better to minimize the possibility of mistakes. I'm just saying that this project is not inherently difficult; rather it is made much more difficult because it is so easy to make errors while soldering.


If you do get a chance to read that article you will notice that the author goes into great detail about the measures that *must* be taken in circuit layout, construction and component selection to prevent interference, parasitic self-inductance, grounding issues and to handle the peak power and voltages that the amplifier is capable of. This means using SMD components where possible, mounting them as close as possible to the tripath chip and keeping distances short between other key components, and addressing the various aspects of grounding. Even the thickness of the insulation used between the mosfets and the heatsinks is very important.

Given the constraints involved, I really don't think you can criticise the PCB layout itself, but it would be nice to have the component identifiers printed on the PCB.

[ikoflexer]

I have a suspicion that my second chip is dead. Could anyone who's got a ta3020 please tell me what resistance should be between pins 47 and 48 (L01 and LO1COM)?
:(

http://soconfused.wordpress.com

[col_s]

quote:Originally posted by ikoflexer

I have a suspicion that my second chip is dead. Could anyone who's got a ta3020 please tell me what resistance should be between pins 47 and 48 (L01 and LO1COM)?
:(

http://soconfused.wordpress.com


I have one right in front of me :)

I measured 18.4 Mohms between pin 47 and 48. This is out of circuit.

Between pins 2 and 3 (lo2 and lo2com) I measured about the same.

[ikoflexer]

quote:
I have one right in front of me :)

I measured 18.4 Mohms between pin 47 and 48. This is out of circuit.

Between pins 2 and 3 (lo2 and lo2com) I measured about the same.


Many thanks! Mine measures around 3ohm, on the board, but with most parts connected to those pins out already. I pray for a short on the board, but it's unlikely. I'll pull the tripath out now.

I have to say, for the amount of grief the truepath has caused me, it better sound like gold.

[BddB]

quote:Originally posted by col_s


quote:Originally posted by ikoflexer

Considering that this project is intended for hobbyists as opposed to industrial ultra-precise robots, the pcb could be made, IMHO, much better to minimize the possibility of mistakes. I'm just saying that this project is not inherently difficult; rather it is made much more difficult because it is so easy to make errors while soldering.


If you do get a chance to read that article you will notice that the author goes into great detail about the measures that *must* be taken in circuit layout, construction and component selection to prevent interference, parasitic self-inductance, grounding issues and to handle the peak power and voltages that the amplifier is capable of. This means using SMD components where possible, mounting them as close as possible to the tripath chip and keeping distances short between other key components, and addressing the various aspects of grounding. Even the thickness of the insulation used between the mosfets and the heatsinks is very important.

Given the constraints involved, I really don't think you can criticize the PCB layout itself, but it would be nice to have the component identifiers printed on the PCB.



Spot on mate! [8D] That's exactly what it's all about.

The comp.id's are intentionally left out, to keep "solderhappy wannado's" away.
OTOH, thorough and systematic craftsmanship will be greatly rewarded :D.

Cheers
/B

[ikoflexer]

I can find no references to the fact that the distance between a solder pad and ground plane should be as small as possible, or that it would in anyway relate adversely to emi/emc. In fact if you look at modern smps pcb boards you will see comparatively large spaces between solder pads or traces and the surrounding ground plane.

Really, if assembly can be facilitated by making it a lot less error prone, of course, without any performance effects, then why not do it?

Am I missing anything?

[ikoflexer]

I can find no references to the fact that the distance between a solder pad and ground plane should be as small as possible, or that it would in anyway relate adversely to emi/emc. In fact if you look at modern smps pcb boards you will see comparatively large spaces between solder pads or traces and the surrounding ground plane.

Really, if assembly can be facilitated by making it a lot less error prone, of course, without any performance effects, then why not do it?

Am I missing anything?

[BddB]

quote:Originally posted by ikoflexer

I can find no references to the fact that the distance between a solder pad and ground plane should be as small as possible, or that it would in anyway relate adversely to emi/emc. In fact if you look at modern smps pcb boards you will see comparatively large spaces between solder pads or traces and the surrounding ground plane.

Really, if assembly can be facilitated by making it a lot less error prone, of course, without any performance effects, then why not do it?

Am I missing anything?


It's not so much the distance to groundplane that matters and in some cases it's even harmful to have groundplane between signals. In some cases it's better to just maximize distance between signals, to avoid crosstalk- (and potential EMC-) problems.

What DO matter, for EMC-related problems , is to keep loopareas to a minimum (as the signal and it's returnpath) as these act like antennas (both radiating and emitting).
The most affecting parameter is the trace-length, which is solved [u]only</u> by placing components, with sensitive signals, close to each other.

Another parameter is to return the signal-current in the groundplane without interrupts. Any obstacles, such as cutouts and splits, in groundplane, will force the returnpath to expand from the ideal (under the signalpath) and thus will the looparea expand.
There are some other factors, (such as tracewidth, layerstackup, dielectric material, amongst others) that matters but more at higher frequencies (mega and giga ;)).

Cheers
/B

[BddB]

quote:Originally posted by ikoflexer

I can find no references to the fact that the distance between a solder pad and ground plane should be as small as possible, or that it would in anyway relate adversely to emi/emc. In fact if you look at modern smps pcb boards you will see comparatively large spaces between solder pads or traces and the surrounding ground plane.

Really, if assembly can be facilitated by making it a lot less error prone, of course, without any performance effects, then why not do it?

Am I missing anything?


It's not so much the distance to groundplane that matters and in some cases it's even harmful to have groundplane between signals. In some cases it's better to just maximize distance between signals, to avoid crosstalk- (and potential EMC-) problems.

What DO matter, for EMC-related problems , is to keep loopareas to a minimum (as the signal and it's returnpath) as these act like antennas (both radiating and emitting).
The most affecting parameter is the trace-length, which is solved [u]only</u> by placing components, with sensitive signals, close to each other.

Another parameter is to return the signal-current in the groundplane without interrupts. Any obstacles, such as cutouts and splits, in groundplane, will force the returnpath to expand from the ideal (under the signalpath) and thus will the looparea expand.
There are some other factors, (such as tracewidth, layerstackup, dielectric material, amongst others) that matters but more at higher frequencies (mega and giga ;)).

Cheers
/B

[ikoflexer]

Thanks BddB, for your patience and detailed explanations. I think it helps other users too to have answers such as these, I know it helps me.


quote:Originally posted by BddB
It's not so much the distance to groundplane that matters and in some cases it's even harmful to have groundplane between signals. In some cases it's better to just maximize distance between signals, to avoid crosstalk- (and potential EMC-) problems.


Then, it would not hurt from a emc avoidance standpoint to have a little more distance between groundplane the solder pad of a pin (for instance the FETs, or the inductors, but many other places would apply), right? It would help tremendously by minimising the probability of a shortcircuit when soldering components to the board.

Basically, what I am trying to get at, is that I recognize that it was my own mistakes that resulted in two fried ta3020s. However, I also think that I the board could be made such that it would minimize the probability of such errors. I searched a bit throught the 41hz forums and it is a fact that there are many reports of fried components due to soldering errors. Perhaps the situation could be improved.

I say this not as a criticism; only trying to help others, if possible.


quote:Originally posted by BddB
What DO matter, for EMC-related problems , is to keep loopareas to a minimum (as the signal and it's returnpath) as these act like antennas (both radiating and emitting).
The most affecting parameter is the trace-length, which is solved [u]only</u> by placing components, with sensitive signals, close to each other.

Another parameter is to return the signal-current in the groundplane without interrupts. Any obstacles, such as cutouts and splits, in groundplane, will force the returnpath to expand from the ideal (under the signalpath) and thus will the looparea expand.
There are some other factors, (such as tracewidth, layerstackup, dielectric material, amongst others) that matters but more at higher frequencies (mega and giga ;)).

Cheers
/B


I have been reading in the meanwhile the technical bouletins from tripath and I agree with you on all accounts, meaning that I am convinced that you have put a lot of thought, expertise, and experience into the truepath pcb.

If the pcb could be made a bit easier to assemble without affecting performance, perhaps even more people would be buying truepath kits, no? :)

[ikoflexer]

Thanks BddB, for your patience and detailed explanations. I think it helps other users too to have answers such as these, I know it helps me.


quote:Originally posted by BddB
It's not so much the distance to groundplane that matters and in some cases it's even harmful to have groundplane between signals. In some cases it's better to just maximize distance between signals, to avoid crosstalk- (and potential EMC-) problems.


Then, it would not hurt from a emc avoidance standpoint to have a little more distance between groundplane the solder pad of a pin (for instance the FETs, or the inductors, but many other places would apply), right? It would help tremendously by minimising the probability of a shortcircuit when soldering components to the board.

Basically, what I am trying to get at, is that I recognize that it was my own mistakes that resulted in two fried ta3020s. However, I also think that I the board could be made such that it would minimize the probability of such errors. I searched a bit throught the 41hz forums and it is a fact that there are many reports of fried components due to soldering errors. Perhaps the situation could be improved.

I say this not as a criticism; only trying to help others, if possible.


quote:Originally posted by BddB
What DO matter, for EMC-related problems , is to keep loopareas to a minimum (as the signal and it's returnpath) as these act like antennas (both radiating and emitting).
The most affecting parameter is the trace-length, which is solved [u]only</u> by placing components, with sensitive signals, close to each other.

Another parameter is to return the signal-current in the groundplane without interrupts. Any obstacles, such as cutouts and splits, in groundplane, will force the returnpath to expand from the ideal (under the signalpath) and thus will the looparea expand.
There are some other factors, (such as tracewidth, layerstackup, dielectric material, amongst others) that matters but more at higher frequencies (mega and giga ;)).

Cheers
/B


I have been reading in the meanwhile the technical bouletins from tripath and I agree with you on all accounts, meaning that I am convinced that you have put a lot of thought, expertise, and experience into the truepath pcb.

If the pcb could be made a bit easier to assemble without affecting performance, perhaps even more people would be buying truepath kits, no? :)

[BddB]

I think it's about choosing the right tool for the job. I have different tips on my soldering irons for different assemblies. Sometimes a hotter tip is needed, sometimes a narrower or wider.

Well, me and Jan had a discussion today about manually assembled and tested Truepath modules.
The question is, if anyone is interested and if it's worth the effort.

Cheers
/B

[BddB]

I think it's about choosing the right tool for the job. I have different tips on my soldering irons for different assemblies. Sometimes a hotter tip is needed, sometimes a narrower or wider.

Well, me and Jan had a discussion today about manually assembled and tested Truepath modules.
The question is, if anyone is interested and if it's worth the effort.

Cheers
/B

[col_s]

quote:Originally posted by BddB

I think it's about choosing the right tool for the job. I have different tips on my soldering irons for different assemblies. Sometimes a hotter tip is needed, sometimes a narrower or wider.


The thing is, at first I found surface mount soldering hard, but now I really prefer to solder the SMD components over hole mount, which I now find tedious. It definitely comes down to using the right tools. In addition to soldering iron tips, I find a good set of tweezers designed for this task helps a lot.

I use these:

http://www.altronics.com.au/index.asp?area=item&id=T2379

The tips are angled flat and are very thin, so it makes it really easy to align components that are in close proximity. They are also flexible so you are a lot less likely to crush or launch one of the components across the room if you squeeze it too tight (if you have done this, you will know what I mean). These are made by Xytronic. Excelta also have a good range:

http://www.excelta.com/marketplace/browse.asp?category=324

The 108-SA model would be similar to the one I use.

Maybe Jan could offer some tools like these for sale on his site?

[audiophool]

ikoflexer,

Well done persisting with your question about PCB pad size - I imagine it must be getting as frustrating as building your first Truepath. Keep trying and you will get an answer eventually.

rgds.

[audiophool]

quote:Originally posted by BddB



...

Well, me and Jan had a discussion today about manually assembled and tested Truepath modules.
The question is, if anyone is interested and if it's worth the effort.
...



Fully assembled and tested Truepath is already available. It is the BFA package at the Optimizing Truepath thread. BFA stands for Base truepath (ie no optimization), Full assembly and test, along with some Accessories (such as screws to affix heatsinks). Refer here:

http://www.41hz.com/Forums/topic.asp?TOPIC_ID=1597&whichpage=3

rgds.

[BddB]

quote:Originally posted by ikoflexer

...
Then, it would not hurt from a emc avoidance standpoint to have a little more distance between groundplane the solder pad of a pin (for instance the FETs, or the inductors, but many other places would apply), right? It would help tremendously by minimising the probability of a shortcircuit when soldering components to the board.

Basically, what I am trying to get at, is that I recognize that it was my own mistakes that resulted in two fried ta3020s. However, I also think that I the board could be made such that it would minimize the probability of such errors. I searched a bit throught the 41hz forums and it is a fact that there are many reports of fried components due to soldering errors. Perhaps the situation could be improved.
...


The isolation distance between signals and groundplane should definitely be enough for manual soldering. It's about using the right amount of solder, right tool for the work and to be thorough. Soldering is both science and an art form ;)
I visually and electrically check every solderjoint I make and in fact I measure every passive component before soldering.

Cheers
/B

ps. Link about soldering: http://www.circuitrework.com/guides/7-0.shtml

[ikoflexer]

I too found soldering the tiny surface mount components easy, that's not where I had problems. The two shorts I identified later were with through the hole components.

Well, I tried identifying something that could be improved; it may or may have not fallen on deaf ears. As far as I am concerned, I've learned a valuable lesson. There are high risk and low risk projects. I would venture to say that this is a higher risk project than many others. Just have a look at the number of failures (some even reported by very experienced hobbyists) in these forums and you'll see what I mean. There is no need to repead that this is not a beginner's project, I think I got that.

I'm not sure I will continue this project, unless I get another ta3020 from some very cheap source. I have to draw a line somewhere as far as sinking money into one particular amplifier, however special it may be. Just to make it crystal clear, I'm not blaming anybody else but me for not getting the truepath working. I do thank all of you for your kindness and attention.

Best regards.

[Rephlex]

I have just solved a similar problem on my truepath - I had a static sound or sometimes a high frequency hiss on the output (like the sound a kettle makes when it is boiling).
I managed to solve it by increasing the current limit on the VN10 bench supply. I had set it to 250mA, but this obviously wasnt enough and I was getting a HF ripple on VN10 that spread to other parts of the board. I increased the limit to 500mA and it now works perfectly, and the current reading on the PSU is around 200mA, which is what it is meant to be.

I am sure that this noise on your trupath is due to a power supply fault (5v or 10v supplies), either a filter capacitor not connected properly or too much resistance / not enough power coming into the board.
I hope this helps

[audiophool]

quote:Originally posted by Rephlex



....


I managed to solve [static sound] by increasing the current limit on the VN10 bench supply. I had set it to 250mA, but this obviously wasnt enough and I was getting a HF ripple on VN10 that spread to other parts of the board. I increased the limit to 500mA and it now works perfectly, and the current reading on the PSU is around 200mA, which is what it is meant to be.

...



Your statement is not clear. On which PSU is the current reading around 200mA? Are you intending to say that there is something wrong with the current limit on your VN10 bench supply because when its current limit was set at 250mA then less than 200mA flowed in it?

[Rephlex]

I meant to say that although the average VN10 current was approximately 200mA, there must have been some high frequency spikes greater than 250mA, causing the bench supply to go into constant current rather than constant voltage mode. Either that or the bench supply current limiter is inaccurate.
Anyway the noise was caused by a hf oscillation on VN10

[audiophool]

I think Okham's Razor clearly points toward your bench supply current limiter being inaccurate.

[ikoflexer]

Thank you reflex, it might very well have been my problem, initially. However, after desoldering and soldering most parts (quite) a few times, in search of bad connections, I probably introduced the shorts I mentioned earlier, and the second ta3020 chip got one channel fried. I need a break from this project, as it frustrated me too much. If I ever get the urge to work again on the truepath...

[audiophool]

quote:Originally posted by ikoflexer

Here are the psus I'm using. All capacitor values in mF.

http://www.cs.toronto.edu/~relu/misc/tp_psu.jpg

http://soconfused.wordpress.com


ikoflexer,

Your problem is the 100 ohm resistor in your VN10 power supply. Replace it with the piece of wire...

Rgds.

[audiophool]

How annoying is this?

In my previous post I politely told Ikoflexer what his problem was - it was his design of his VN10 power supply and nothing to do with Truepath or 41Hz - and how to fix it. I even updated Audiophool's Truepath Assembly Guidelines to include an additional standalone VN10 test (http://www.maedos.com/Guideline.html#Test_VN10) to specifically catch this problem in future.
Not only did this ungrateful guy ignore my advice and not even bother to reply to my post, but he actually went over to diyaudio, and had a big grizzle about Truepath and 41Hz. http://www.diyaudio.com/forums/class-d/155974-choice-ta3020-boards-2.html (I'm assuming that both Ikoflexers are the same person):


My experience with the 41hz truepath board was quite bad. One of the problems I had was that while searching for the kit assembly documentation I found several versions on their website. Some of if was very confusing. The board is very well made in the sense that an assembly robot is ideal to stuff it. The distance between traces and pads is very very small and you need to work with a magnifier and make check your soldering very well. Look at the forum and check out the number of failures with the truepath. When I mentioned the fact that the board isn't very well suited to human soldering and that tiny changes could improve the success ratio the response I got was "this is for experienced people only." So there, after agonizing days and nights I got one channel working, and fried two tripath chips in the process. There were people in the forum that are very helpful, including a 41hz guy. They do try to help, that's for sure. I didn't know about the [cheap TA3020 based amp whose name starts with "con"] before, but in retrospect, that's the one I'd go for now. Just my 2c.:mad:

[V-bro]

Sharp as always Tim! Very sharp!

They also don't see the Connex boards are really commercial money making crappy modules where many parts are used on the edge of their ratings. Still, this is another example of how little people really care about quality. They just want something that works and preferably without breaking a drop of sweat. Even when it is just a quick and dirty solution. It's like there are cars that are a tin on four wheels, will bring you anywhere, just as a Porsche. Only a handful understand this, you get the Porsche for a bargain price here! Only at the price of investing a little more of your own time and trouble...