Hey all,
I have basic question...I have a single TDA1543/CS8414 setup and quite happy. I noticed a number of designs that have nore than one TDA1543 (up to eight) to improve sound quality and wonder if it would be worth it to literally Stack another one or two on top of the existing one I have...the same pins would be soldered t the same pins of the TDA1543 below it...these TDA1543 chips can be had quite cheap, but wonder if anyone done it, or worth it. One such design is attached from this site...
My power supply is 12V and 1.5A, so I think I have enough "juice" for another 5V TDA1543...
I have basic question...I have a single TDA1543/CS8414 setup and quite happy. I noticed a number of designs that have nore than one TDA1543 (up to eight) to improve sound quality and wonder if it would be worth it to literally Stack another one or two on top of the existing one I have...the same pins would be soldered t the same pins of the TDA1543 below it...these TDA1543 chips can be had quite cheap, but wonder if anyone done it, or worth it. One such design is attached from this site...
My power supply is 12V and 1.5A, so I think I have enough "juice" for another 5V TDA1543...
Attachments
john65b said:
AFAIK they draw 60mA per chip, so you could get away with 4 or more pieces... watch out for cooling issues (maybe stick pieces of copper sheet between them). Additional cooling is mandatory with more than two stacked.
Should be ok, but you need to change Vref and the IV resistors as the Iout would add.
Here's a nice tool I discovered on another forum for calculating the resistors.
Wow, that was a quick response...
I have ordered a few more TDA1543...and found a newer TDA1543 calc that corrects the Vout...Thanks lucpes...your ver was 24, newer ver is 28... I think...
Anyway, I have read that a few people run their TDA1543 at voltages above 5V (mine is 5V)...upwards of 8 - 9V with heatsink to increase the output to thier pre...I have a low gain tube amp and can use the input boost...
So I noticed my Vref and IV resistors are both 1K (single TDA1543) ....and according to the calc, if I stack one more TDA1543 and leave the 1K resistors as is, I have the same effect as running one TDA1543 at 8V (same increase in the output), and with two chips, it should sound "better"?
Is this correct and recommended?
To use a total of four stacked (I bought three more), I understand I would definitely need to modify the resistors, but I decided to try one first without changing any resistors...if recommended...
Here is a link to what I have...
http://us.hifidiy.net/Article.asp?ArticleId=152
Pretty simple, cheap, and basic, and I am liking it with my stepped attenuator on a few Gainclones, Tripaths and UCD180 amps...
Thanks all
I have ordered a few more TDA1543...and found a newer TDA1543 calc that corrects the Vout...Thanks lucpes...your ver was 24, newer ver is 28... I think...
Anyway, I have read that a few people run their TDA1543 at voltages above 5V (mine is 5V)...upwards of 8 - 9V with heatsink to increase the output to thier pre...I have a low gain tube amp and can use the input boost...
So I noticed my Vref and IV resistors are both 1K (single TDA1543) ....and according to the calc, if I stack one more TDA1543 and leave the 1K resistors as is, I have the same effect as running one TDA1543 at 8V (same increase in the output), and with two chips, it should sound "better"?
Is this correct and recommended?
To use a total of four stacked (I bought three more), I understand I would definitely need to modify the resistors, but I decided to try one first without changing any resistors...if recommended...
Here is a link to what I have...
http://us.hifidiy.net/Article.asp?ArticleId=152
Pretty simple, cheap, and basic, and I am liking it with my stepped attenuator on a few Gainclones, Tripaths and UCD180 amps...
Thanks all
Uh, you can't just willy-nilly add stacked chips. You have a fanout issue here. First, you take care of the obvious power supply needs.
But then you have to drive the logic inputs. Will need some sort of heavy duty buffers to drive them properly without totally screwing up the rise/fall transitions. I'd suggest using a small series resistor for each receiver pin. Can be 100 ohms or so. Also helps separate the distributed capacitive load. Anyway, bottom line is you'll have to redesign the BCK, DIN, WCK drivers too. Proper fanout and clean terminations are in order.
Regardless, one problem you will have is different propogation delays. The output from each DAC will switch at a slightly different time. They're all close, but if in parallel, you will now add some jitter to the output currents. Or basically, not-so-clean edges. Filtering can help this problem.
Anyway, just paralleling devices can get you into a world of trouble. You have to appropriately design everything around the converter (on either side) to accommodate parallelism.
Of course, if all you are looking for is more volume, then go for it.
jh
But then you have to drive the logic inputs. Will need some sort of heavy duty buffers to drive them properly without totally screwing up the rise/fall transitions. I'd suggest using a small series resistor for each receiver pin. Can be 100 ohms or so. Also helps separate the distributed capacitive load. Anyway, bottom line is you'll have to redesign the BCK, DIN, WCK drivers too. Proper fanout and clean terminations are in order.
Regardless, one problem you will have is different propogation delays. The output from each DAC will switch at a slightly different time. They're all close, but if in parallel, you will now add some jitter to the output currents. Or basically, not-so-clean edges. Filtering can help this problem.
Anyway, just paralleling devices can get you into a world of trouble. You have to appropriately design everything around the converter (on either side) to accommodate parallelism.
Of course, if all you are looking for is more volume, then go for it.
jh
everything in this thread is covered on my website. You can read everything about cooling, voltages, bias, paralleling etc etc
just have a look at www.dddac.de
doede
just have a look at www.dddac.de
doede
Hi Dan, no I haven't ever actually stacked DAC chips in real life. Only in my head. It's an intellectual exercise in thinking of what the problems would be and how to tackle them.
Indeed, I am correct in assuming the propogation delays through each DAC chip will be different. This is purely based on experience, not any measurements. This info might even be listed in the data sheet. They are common parameters in all other digital chips. However, after some more thought, I am wrong in using the word "jitter". The result will appear as though it is jitter, but the misaligned and staggered edges are not random, but fixed. Hence, no actual jitter component is added. The propogation times, however different, are reasonably fixed. The output current step will thus be an "average".
I have done many other circuits where gates or chips or whatever are paralleled. You do have to take into account a number of things. My main concern is the input capacitance of each receiver (the DAC inputs). This dynamic and nonlinear loading gets multiplied. Your driver circuits need to be designed to handle the increased load. That should be obvious.
There is yet another problem with stacking I hadn't mentioned. You end up with a lot of crosstalk in the power and ground legs of the chips. Any decoupling is down on the circuit board. The transient currents (which are not switching at the exact same time) are coupled from one chip into another via the stacking wiring. Paralleling 2 DACs is not a big problem. But stacking 20 of them is really asking for trouble. Just because it works does not mean it works well.
Are you using an output buffer after the stacked DACs? Or just going directly into a resistor and the outside world? Without a buffer, the effects of interconnects and the loading of a subsequent preamp will affect operation. Indeed, the more DACs in parallel, the lower the output impedance and the better it can drive the external loads. So is the improvements you hear because of parallelism? Or because of better output drive? Perhaps a proper output buffer can accomplish the same thing without multiple DAC chips. Do you really want an unknown external load sitting on your DAC? Such would make the sound very sensitive to each individual installation.
So what is the benefit of parallelism? You get theoretically 3dB better SNR for each doubling of chips. But is noise floor really a problem? I don't think so. The one useful improvement is that of averaging. If, and this may not be true, the differential nonlinearity of each DAC is random, then paralleling will tend to average this out and provide more linearity and less distortion. Basically, a better DAC. The more, the merrier.
But stacking seems like such brute force. You might want to consider a more gentle approach, like a distributed amplifier. Split inputs via transversal filters and then couple the outputs again via inverse timed filters. Check out schematics of old Tek scopes (like a 545) for great examples of a distributed amplifier. Do the same thing but using DACs. This would allow you to combine DACs without the problems of stacking. Each could be layed out side-by-side with their own supply decoupling. The delay lines provide proper timing of inputs and also combine outputs whilst providing a nice HF filter rolloff. Hmmm - maybe this is too much for DIY.
jh
hagtech said:
correct, that is the issue here .... too many intelectual excercices, too many speculations and too less people who actuall listen and measure .... and I mean the combination !!! not only measuring like some others .... BOTH !!
indeed, on paper Hagtech has a point, but in real life the paralleled dac sounds beter for most of the people building this concept... so what is most important ??
OK, I know, we had this discussions probably a hundred times here, but I cannot let go reacting on it
doede
Thanks for the reply JH.
I'm pretty much in agreement with what you say. I'm more interested in the improved linearity than noise, plus the better output Z. My DAC uses a 68 ohm resistor for I/V, so is pretty much load insensitive.
I might try adding an active output stage as you suggest, but at the moment I'm looking at building a DAC based on the AD1896 and PCM1796 which should give an interesting comparison between too different approaches.
Dan
I'm pretty much in agreement with what you say. I'm more interested in the improved linearity than noise, plus the better output Z. My DAC uses a 68 ohm resistor for I/V, so is pretty much load insensitive.
I might try adding an active output stage as you suggest, but at the moment I'm looking at building a DAC based on the AD1896 and PCM1796 which should give an interesting comparison between too different approaches.
Dan
Classic. The run and hide defense. Why not just address some of the points I make with a reasonable discussion?
Look, nowhere did I suggest that paralleling DACs was bad or could not sound better. I was only pointing out possible problems with certain implementations and some ideas on how to address them. Perhaps you did your design from the start knowing you had to deal with parallelism. Fine. Others may not. I was merely cautioning against stacking chips in designs not prepared to handle the load. It helps to engage your brain and consider potential pitfalls.
Hiding behind an I - listen - therefore - nothing - you - say - matters response gets us nowhere. It is also insulting to many who contribute here.
jh
Where is the Aloha here? No need to start a ******* match over this.
In all fairness, "You cannot just stack the chips and expect it to work" advice is well taken - It was exactly what I was going to do...with some critical listening...After all, if it sounds great or better than a single chip, what do I care if all these other "technical minutia" issues appear...as long as it doesn't damage my equipment...
I will not just stack three more cihps on without correcting the three resistors, as is what is recommeded. All the other Buffer, Rise/Fall, Driver issues and other stuff I will leave to those more electronics savvy than I.
No offense, but I do not want to turn a $40 DAC into a $4000 University Electrical Engineeering Thesis.
I am getting off my soapbox now.
In all fairness, "You cannot just stack the chips and expect it to work" advice is well taken - It was exactly what I was going to do...with some critical listening...After all, if it sounds great or better than a single chip, what do I care if all these other "technical minutia" issues appear...as long as it doesn't damage my equipment...
I will not just stack three more cihps on without correcting the three resistors, as is what is recommeded. All the other Buffer, Rise/Fall, Driver issues and other stuff I will leave to those more electronics savvy than I.
No offense, but I do not want to turn a $40 DAC into a $4000 University Electrical Engineeering Thesis.
I am getting off my soapbox now.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.