Go Back   Home > Forums > >
Home Forums Rules Articles diyAudio Store Blogs Gallery Wiki Register Donations FAQ Calendar Search Today's Posts Mark Forums Read

Class D Switching Power Amplifiers and Power D/A conversion

What is wrong with TPA3255?
What is wrong with TPA3255?
Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 5th October 2018, 09:45 AM   #441
doctormord is offline doctormord  Germany
diyAudio Member
 
doctormord's Avatar
 
Join Date: Oct 2007
Quote:
No, of course its not 'proof', its too early days for that. Its an interesting result which calls for further experiments. I do intend to do some in respect of the 'too low PSRR hypothesis' because to do science, all hypotheses need to be testable.

As for whether the designers 'do it properly' - it hinges on what 'properly' means. To TI, 'properly' means achieving the target measurements (THD+N in particular). However to a large number of listeners here, 'properly' means sounding as good as they can get it. No contradiction, just differing aims. So what does 'properly' mean to you?
Well, what i mean is, that "cap rolling" does "improve" SQ for decades with (mostly) no scientific background/proof. SQ is so much subjective and biased that, to me, alot of people hear an improvement/benefit just because there must be something and it costs them some money. Some mods are really expensive for an subjective improvement. My personal opinion is, if it can't be measured in some way on properly interpreted afterwards, it is some kind of snake oil. But this is not meant to offend anyone who "believe in".

Quote:
Does 'everyone' in scare quotes here not really mean everyone? What benefit would a vendor/designer like to see ? I ask you because clearly you are both. Adding more capacitance on AVDD isn't going to degrade your measurements, merely give an increment on your BOM cost. But if you're a vendor defining a 'proper' design as one with certain measurements, you're not going to see any benefit of increasing your BOM.
With "everyone" i mean people who aren't into the details that much. Not to offend anyone here as well. I want to make clear that i'm not a vendor, but a hobbyist. To be on par, adding more capacitance on AVDD may degrade reliability due to not limited dI/dT at startup. It also affects loop bandwidth of the internal LDO which actually is unknown. Next, i can't understand, how stereo separation will get improved if there's no separation in the supply section where the capacitance is added. I have no problems adding costs to the BOM as hobbyist, but if they wont help/improve anything, they aren't needed. My actual 32XX design also does have some parts populated which don't add much benefit to the overall performance but are there "because textbook". I.e. output capacitance/buffering of the DCDC-converter rails with the help of electrolytics. As the converter runs at frequencies where electrolytics aren't much effective at all, the benefit is questionable. All they do is to influence loop-bandwidth at some extent. Taking PSRR and additional local decoupling into account, they most can likely be omitted. Next, the analog front-end does feature an HF-shielding, which does help on EMI interference from 3G/4G/WiFi coupled in and get demodulated for ranges closer than 10cm to the inputs. Benefit? yes! Real world problem? Most likely not.

Parts i "believe in" SQ improvement "because textbook" are .5% 25ppm thin-film resistors in the analog-frontend to keep the symmetry of the diff-stage as close as possible. Those i use: "RR0510P-104-D"

Edit:

Every additional capacitor is an additional (kind of) antenna which is prone to catch some HF from the switching notes. We've seen this i.e. here:

What is wrong with TPA3255?

Quote:
Quote:
Quote:
(Beside the fact that most “cap rollers” never do any measurements) PSRR measurements at AVDD shouldn’t be that hard to do on the EVM if adding caps is that simple anyway. if i don’t split the supply for strep sides, how can stereo image/separation get improved, the “influence” path is still there. All what happens here is to slow down the regulation bandwidth of the internal LDO and force electromigration to degrade the silicon. I wouldn’t call this a mechanism.
I get the impression you've not been following along. PSRR measurements I agree aren't hard to do, just time-consuming.
I still don't get the "mechanism" involved you're refering to.

Quote:
Quote:
Quote:
For the part of ripple reflection i mean it in a way that noise not only conducts this the output but also to the input (or source). That’s why switching power supplies got a filter also on the input, not only for power factor correction.
Its me who's not following you now. Input filters are for common-mode noise suppression (if we're talking about the same input filters). Can't see how CM noise is affected here.
I'm refering to effects like the ones shown on page 10ff:

https://flex.com/sites/default/files...ign_FlexTS.pdf

Last edited by doctormord; 5th October 2018 at 09:57 AM.
  Reply With Quote
Old 5th October 2018, 10:07 AM   #442
abraxalito is offline abraxalito  United Kingdom
diyAudio Member
 
abraxalito's Avatar
 
Join Date: Sep 2007
Location: Hangzhou - Marco Polo's 'most beautiful city'. 700yrs is a long time though...
Send a message via Yahoo to abraxalito
Quote:
Originally Posted by doctormord View Post
Well, what i mean is, that "cap rolling" does "improve" SQ for decades with (mostly) no scientific background/proof. SQ is so much subjective and biased that, to me, alot of people hear an improvement/benefit just because there must be something and it costs them some money.
I can't for the life of me see a point to cap rolling. I've not engaged in it, I'm unlikely to ever engage in it. But strange things might happen. What I'm suggesting here in terms of decoupling isn't to be confused with cap rolling. I don't care about particular cap brands, only measurements like ESR when it comes to decoupling caps.

For me cap-rolling is just a harmless leisure activity with no benefit to understanding or progress in audio design.

Quote:
Some mods are really expensive for an subjective improvement. My personal opinion is, if it can't be measured in some way on properly interpreted afterwards, it is some kind of snake oil. But this is not meant to offend anyone who "believe in".

So it would appear you're a believer - at least in audio snake oil I don't share your cynicism, whilst there is a whole heap of what I might call 'subjectivist nonsense' in audio, there's also such a thing as better SQ. That's not just my belief, its my experience too. Of course better SQ can be measured, just that the measurements aren't sufficiently developed to demonstrate it.

Quote:
With "everyone" i mean people who aren't into the details that much. Not to offend anyone here as well. I want to make clear that i'm not a vendor, but a hobbyist.
OK my mistake, I'd taken you as a vendor.

Quote:
To be on par, adding more capacitance on AVDD may degrade reliability due to not limited dI/dT at startup. It also affects loop bandwidth of the internal LDO which actually is unknown.
Yes, both good points.

Quote:
Next, i can't understand, how stereo separation will get improved if there's no separation in the supply section where the capacitance is added.
If by 'stereo separation' you mean a crosstalk measurement then sure, I can't understand that. But that's not what we've been talking about here, rather the extra capacitance gives subjective improvements. Which presumably you take to be snake oil (based on what you wrote above) - that's your choice and I have no issue with you making that choice.

Quote:
I have no problems adding costs to the BOM as hobbyist, but if they wont help/improve anything, they aren't needed.
Does 'anything' include subjective impressions? I take it from what you've written so far your answer is likely to be 'no', in which case you'd be daft to add extra capacitance to your BOM.

Quote:
I still don't get the "mechanism" involved you're refering to.
Let me outline it again see if more words helps in any way. With a classAB output stage (as used internal to the TPA3255 signal processing), the draw from the power supply is modulated with signal. If the PSRR of the other circuitry sharing the same rail is sufficiently high then this won't matter but since I (and others) hear improvements from lowering the supply impedance it tends to indicate the PSRR isn't sufficient for good subjective results. I hear the biggest improvement from lowering the impedance on high crest factor music - orchestral violins for example become so much more life-like. For those who only listen to heavily compressed material I'd guess the improvement will be smaller. But that's just conjecture.
__________________
I know you think you understand what you thought I said but I'm not sure you realize that what you heard is not what I meant - Alan Greenspan
  Reply With Quote
Old 5th October 2018, 10:38 AM   #443
doctormord is offline doctormord  Germany
diyAudio Member
 
doctormord's Avatar
 
Join Date: Oct 2007
Now i got it. So for the mechanism to further explain regarding PSRR informations regarding OTA implementation is needed, i.e. like shown here:

http://rincon-mora.gatech.edu/resear...c/gtac_s3f.pdf

The usual links given:

http://www.analog.com/media/en/techn...es/AN-1120.pdf


Regarding supply seperation:
I remember TI seperated VDD and GVDD with 10uH inductors initially but then went back to simple resistors as they showed better performance and where less prone to EMI pickup.

Quote:
Quote:
Quote:
I have no problems adding costs to the BOM as hobbyist, but if they wont help/improve anything, they aren't needed.
Does 'anything' include subjective impressions? I take it from what you've written so far your answer is likely to be 'no', in which case you'd be daft to add extra capacitance to your BOM.
This isn't restricted to capacitance, but let me show you my decoupling path for the AFE:

Decoupling at LM5017 input/outputs:
What is wrong with TPA3255?-lm5017-png

This is done locally with a concentrated GND plane connected to global GND at the GND-pad of the input supply decoupling capacitor. LM5017 is equipped with LC-filter input. fSW is 620kHz.

in distance of 3cm there is local decoupling via 47uF at the AFE connector - this cap is of questionable value as it isn't effective at switchers fSW but in the audio range.

What is wrong with TPA3255?-connector-png

On the AFE module itself there's a 2 stage RC-filter with 2nd-stage feed-trough cap (X2Y) and finally local decoupling at the opamp itself.

What is wrong with TPA3255?-afe_filter-png
Attached Images
File Type: png lm5017.png (14.7 KB, 252 views)
File Type: png afe_filter.png (12.8 KB, 446 views)
File Type: png connector.png (14.9 KB, 247 views)

Last edited by doctormord; 5th October 2018 at 10:44 AM.
  Reply With Quote
Old 6th October 2018, 01:55 AM   #444
IVX is offline IVX  Ukraine
diyAudio Member
 
IVX's Avatar
 
Join Date: Jul 2003
Location: south of China
Guys, it is a really bad idea to use roll-type Wima as LC filter or as decoupling for class D. Sometimes even MLCC X7R(with 3-5 times higher voltage rating, 1210 1u/100V for 25VDC is ok) could be better regarding THD/Noise/EMI(of course if the THD+N related to EMI, that's pretty common thing). The optimal cap for output LC is stacked film cap, ideally if SMD one. I see rolled film caps like those glamour-pinky Wima in many wrong cases such as SMPS/classD/DAC 3.3V rails etc no idea who made that mistake so popular.
__________________
Let's do smth nearly impossible!
E1.
  Reply With Quote
Old 6th October 2018, 06:43 AM   #445
doctormord is offline doctormord  Germany
diyAudio Member
 
doctormord's Avatar
 
Join Date: Oct 2007
Well, stacked film exist also in THT variant, i use TDK which is supposed to be in stacked configuration. SMT film is very delicate to solder and can/will make trouble if convection reflowed with components from much higher thermal mass like (huge output) inductors. To overcome this, vapor phase reflow soldering is needed. Biggest problem is shifting capacitance and tand. I prefer THT film or 250V MLCC at this place.

@IVX, are you referring to the much lowered dV/dT rate for most Wima caps thatarent pulse capable and the fact that roll-type is prone to „suck in“ HF/RF/EMI?
  Reply With Quote
Old 6th October 2018, 07:03 AM   #446
IVX is offline IVX  Ukraine
diyAudio Member
 
IVX's Avatar
 
Join Date: Jul 2003
Location: south of China
I think the issue is ESL, it is high enough to create problems for self-oscillating amps.
__________________
Let's do smth nearly impossible!
E1.
  Reply With Quote
Old 6th October 2018, 04:23 PM   #447
voltwide is offline voltwide  Ireland
diyAudio Member
 
Join Date: Jan 2010
Quote:
Originally Posted by IVX View Post
Guys, it is a really bad idea to use roll-type Wima as LC filter or as decoupling for class D. Sometimes even MLCC X7R(with 3-5 times higher voltage rating, 1210 1u/100V for 25VDC is ok) could be better regarding THD/Noise/EMI(of course if the THD+N related to EMI, that's pretty common thing). The optimal cap for output LC is stacked film cap, ideally if SMD one. I see rolled film caps like those glamour-pinky Wima in many wrong cases such as SMPS/classD/DAC 3.3V rails etc no idea who made that mistake so popular.
I wonder where you see rolled wima tht caps. Which series do you refer to?
  Reply With Quote
Old 6th October 2018, 04:32 PM   #448
irs20957 is offline irs20957  Russian Federation
diyAudio Member
 
Join Date: Dec 2017
Location: Vladimir-Moscow
Quote:
Originally Posted by voltwide View Post
I wonder where you see rolled wima tht caps. Which series do you refer to?
in my constructions may by?
  Reply With Quote
Old 6th October 2018, 05:51 PM   #449
voltwide is offline voltwide  Ireland
diyAudio Member
 
Join Date: Jan 2010
Quote:
Originally Posted by irs20957 View Post
in my constructions may by?
who knows?
  Reply With Quote
Old 6th October 2018, 08:13 PM   #450
doctormord is offline doctormord  Germany
diyAudio Member
 
doctormord's Avatar
 
Join Date: Oct 2007
I remember using CBB22 style capacitors without an issue in the output filter, they where rolled for sure. (Rolled foil dipped in resin)
  Reply With Quote

Reply


What is wrong with TPA3255?Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
TPA3255 - all about DIY, Discussion, Design etc. lms Class D 913 18th October 2018 06:15 AM
My 1st Class-D Amp -TPA3255 with PFFB JohnnyInBris Class D 19 13th August 2018 02:26 PM
FS: TPA3251 TPA3255 2.0 BTL Amplifier doctormord Swap Meet 11 13th August 2017 07:14 PM
TPA3255 Amplifier for Car Audio System GABRIELITEDEVIL Class D 6 1st January 2017 09:37 PM
Wrong Voltage / current on BoZ... or Wrong Measuring ? gionag Pass Labs 8 14th May 2008 11:24 AM


New To Site? Need Help?

All times are GMT. The time now is 08:38 PM.


Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 15.00%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Copyright ©1999-2018 diyAudio
Wiki