Optimizing TDA7294 Output

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Both the PS and the amp LEDs on my build fade out. None of the four turn off instantly. The on-board LEDs are more sensitive and take longer but even they showed discharge after about 4 seconds . I did some timings this morning and found having just one amp with the new cap arrangement greatly influences the discharge time. The amp LED now fade first at ~ 4 -6 seconds. The modded amp side shows the presence of a charge up to 24 seconds. (on the PS only)

If you have two stock builds I would suggest swapping them to see if one of the amps is causing the discrepancy. A few posts back Daniel did remark he thought even my stock amp pair was a little slow discharging. Try the swap and let us know what you find.
 
I did some timings this morning and found having just one amp with the new cap arrangement greatly influences the discharge time. The amp LED now fade first at ~ 4 -6 seconds. The modded amp side shows the presence of a charge up to 24 seconds. (on the PS only)
Those times look appropriate but I'd need some more information.

It is important to compare when using the same model of speaker so that the load compared is the same.

Since the bass bandwidth is different between the two amplifiers, it is more useful to compare non-audio workload, and therefore we'd need to have the source connected, powered on, and the volume set to minimum (input dampening but no input signal).

In this condition, which amplifier takes longer to drain the power supply board?
 
The next step. . . Optimize Pin 2

Parameters:
1). Feedback resistor within the range of: 25k to 110k
2). Feedback-shunt resistor within the range of: 1.3k to 3.0k
3). NFB-shunt cap sized to fit Feedback-shunt resistor value so we don't accidentally make a bass blur/blocker.
4). Gain settings within the range of 20x to 42x--Simply divide the two resistor values and then add 1, to find your gain value. Gain also affects compensation so be sure to observe temperature and aim for cool.

If you need some gain for either fun dynamics and/or so that a PC or digiplayer doesn't strain, here's a couple of possibilities.
75k with 2.2k for a gain of 35x or
60k with 1.8k for a gain of 34x.
In these two examples, the gain factor is nearly identical but the feedback current is different so the audio will be different (while driving a real speaker).

Choosing by observation:
Maximize the soundcast width of a single (mono) speaker, since the bigger presentation usually represents higher resolution of micro detail.
Maximize the baritone range clarity, especially Cello/Piano, because this is the harmonic range that allows you to hear bass in normal size rooms.

Those two audible parameters are adjusted mainly with these two resistors: Feedback resistor and Feedback-shunt resistor.
 
Last edited:
Well there's just 5 things, some of which are location sensitive, so breadboard won't help. Just remember, you're also optimizing the sound of a given board. The easiest or fastest thing to do is actually use the printed circuit boards.

1. Feedback resistor trackside, onto the chip, from pin 2 to pin 14
2. Feedback-shunt resistor trackside (for easy swapping)
3. NFB-shunt caps and bypass caps mounted on pins (for easy swapping)
4. Input cap--mount helper/bypass cap trackside (for easy swapping)
5. Rail2Rail power filter cap (or RCR) at amp board V+, V- terminals

Suggestion: You've got a 2 amplifiers there, so modify whichever one you like the least, until it sounds superior. Well, the prospect is to mod one to beat the other, and then next, vice versa, round robin fashion, each better than before. Estimated timeframe when using printed board is from 2 to 6 hours depending on methodology with listening tests.
 
I want to thank Daniel and everybody else that contributed to this thread.

I have mentioned my "no name" chip amp before. Well I got it out because I intend to add a preamp (tone controls etc) to it so it is a truly integrated amp. I dusted off the chip and used my new illuminated magnifying glass (my eyesight is really terrible :( ) and found out that I have TDA7296 chips running off a 62 volt split supply.

I used the same board that was in the "no-fi" tabletop stereo. I removed the ripple caps (2200uF per rail is all it had) and installed 330 uF caps on the board. I used a capacitor bank like Daniel's with 4x2200 uF per rail and it made a noticeable improvement. Then I swapped the bootstrap and feedback caps (it used the data sheet 22 uF caps with smaller feedback resistors - really terrible!) with 100 uF caps and it made a huge difference. (I did not even know that the chip used a bootstrap - I just swapped it out.)

It works really great, but I know there are still gains to be had. I am not going to tweak the chip amp any more; I am just going to put the preamp in and put it back in the TV room. After reading this thread for probably the 50th time, I am motivated to build a parallel-bridge amp using 4 chips per channel. It will be optimised to the max (dual monoblock etc) and I will try the shottky diode trick to isolate the power supply for the low level input circuitry. I think that 100 watts RMS into 8 ohms is a good goal to shoot for.

I am truly inspired that such performance is available from such a generic chip by optimising design and layout. That's what this "hobby" and website is all about.
 
Thanks for posting! Kudos on the refit!
I'm grateful that your post shows that a good power circuit (big enough tank for good bass + working decoupling supporting high speed transient for high resolution treble) is an excellent head start for high quality audio. For TDA729X, the datasheet example application is like a prank, and that is used in retail amplifiers; so, improving performance begins with replacing the power circuit.

I'm a bit impressed. Normally, I would have risked 220u amplifier board power caps, and that's excellent clarity at some risk of shouty (but not an impossible amount) and then re-tune the whole amplifier board to keep the clarity, dump the shout. But, for Television amplifier with TDA729X, and a speedy refit, your 330u was the perfect choice for good clarity with no risk of shout. Good job!! I'll bet it runs cooler too.

P.S.
Did you check to see if the NFB-Shunt cap was a reasonable size for clear quality bass? It is possible to make a refit bypass to fit underneath the board, but I'd need to know the size of the existing cap and its associated series resistor. If I knew that info, I could make a speedy stick-on refit circuit for it, like those press on nails, except for different purpose of high resolution bass. :D
 
Thanks for posting! Kudos on the refit!
I'm grateful that your post shows that a good power circuit (big enough tank for good bass + working decoupling supporting high speed transient for high resolution treble) is an excellent head start for high quality audio. For TDA729X, the datasheet example application is like a prank, and that is used in retail amplifiers; so, improving performance begins with replacing the power circuit.

I discovered the capacitor bank concept by accident. I was fabbing up a headphone amp (DoZ type topology) from old salvage parts (some are real old- more than 30 years!) and it sounded tubby and flat. I was using some really old 2200 uF salvage caps and all I had on hand was some 220 uF caps from a capacitor "kit" I bought (I always buy this stuff to have on hand) and I used 10 per amplifier (single ended power supply sorry haters)(channels isolated by simple 10 ohm resistors) and that was the money shot! I knew there was something to it and so I googled around (finally googled the term "capacitor bank") and it brought me right to some of your posts. Since I am trained in electrical engineering I was not satisfied until I knew why it worked; and it makes perfect sense.

I'm a bit impressed. Normally, I would have risked 220u amplifier board power caps, and that's excellent clarity at some risk of shouty (but not an impossible amount) and then re-tune the whole amplifier board to keep the clarity, dump the shout. But, for Television amplifier with TDA729X, and a speedy refit, your 330u was the perfect choice for good clarity with no risk of shout. Good job!! I'll bet it runs cooler too.

Both amps are on the same board so 330 uF was a logical choice. I did not try different values or optimising the rest of the circuit; I just slapped them in there because it's what I had. I also put some 0.1 uF metallized poly caps (20 in each capacitor "kit" :D so use em up) at the caps and at the power supply pins of each chip.

I am surprised how cool it runs. When I tried it with no mods, it sounded terrible and ran hot. Is this maybe because of gross phase errors from the really terrible original circuit?

P.S.
Did you check to see if the NFB-Shunt cap was a reasonable size for clear quality bass? It is possible to make a refit bypass to fit underneath the board, but I'd need to know the size of the existing cap and its associated series resistor. If I knew that info, I could make a speedy stick-on refit circuit for it, like those press on nails, except for different purpose of high resolution bass.
j2xgwDMBDSsnVA69OtCogRQzID4UOTOFEwAHkGJSmhPzgRIAFU7lagXmBJGuXwEwEMPlz8AaBEagAjDrBQAkcUj5iCtQwK0FSzhciBsQADs=

I repalced the 22 uF shunt caps with 100 uF. It gives a pole of 7.9 Hz which isn't optimal but it's going to stay that way. I will build a highly optimised circuit from scratch and see what I get. Like I said, my goal is to bridge-parallel for 100 watts RMS/ 8 ohms so I have some homework to do.

I am intrigued by your use of small resistors and/or high impedance wire to isolate the power supply board from the amp board. My training says this is completely wrong, but your explanation makes sense. I will try it.

I've been following your work for a while, Daniel. You and I think alike on many topics. They don't teach this stuff in engineering school and it is painfully obvious to me that many commercial designers don't know squat about audio or else they just don't care. Keep it up.
 
Last edited:
I am surprised how cool it runs. When I tried it with no mods, it sounded terrible and ran hot. Is this maybe because of gross phase errors from the really terrible original circuit?
The hot temp and bad quality was from the 2200u from the datasheet schematic, very close to the chip, disabling fast transients because of outwaiting the cap. That problem also cripples the negative feedback a bit, and makes much heat.
You fixed that problem by replacing it with the 330u. Much faster!
The amount of capacitance is unique per each amplifier design because you need only just enough to dampen the midrange for a palatable tone. So, generally, we'd first try to find out how much is insufficient (maybe 100u) and then install progressively larger caps until a suitable compromise is found. For example, I like your idea of 330u for TDA729X and Television source, since a smaller cap might be a bit shouty. But you could try 270u, 220u to see if the clarity increases while the tone remains tolerable. After that's all maxed, you can then apply the rail to rail cap to recoup the pretty tone.
I am intrigued by your use of small resistors and/or high impedance wire to isolate the power supply board from the amp board. My training says this is completely wrong, but your explanation makes sense. I will try it.
I wouldn't use small resistors for that. Actually we need stuff capable of big current. The bell wire umbilical (and filter) is because constructors would have omitted the correct part, which is schottky, like MBR1035 plus an FB73 ferrite. I tried this. It looks too weird. Nobody would use it. So, I substituted 7" bell wire, 20ga solid copper. The substitute does work and has bonus peak filtering and bonus RF filtering if you tie a knot in the umbilical, close to the edge of the amp board (like a $0 ferrite). Still, it would probably sound better with the schottky (at amp board power taps) either instead of or in addition to the bell wire. The series element at this point prevents the amplifier predrive small transients from outwaiting the power tank/reservoir capacitance. So, a good series element blocks that error. Regs or diodes would be good. But, a rather fine tuned cable might do a portion of that task (while looking conveniently orthodox).
If using a discrete parts amplifier, the decoupling technique could be more elegant if done only for the voltage stage as in Hafler DH-220.

And then there's the matter of charge/recovery speed, like this: Inexpensive little CRC supply running the big Honey Badger discrete amp.
 
Last edited:
I replaced the 22 uF [nfb]shunt caps with 100 uF. It gives a pole of 7.9 Hz which isn't optimal but it's going to stay that way.
There's a couple of quick options there:
You could parallel on another, identical, model 100u (to form a high end 200u), or you could decrease the input cap size until the amp plays clean.

P.S.
Thanks for the compliments. I wish I was trained as an engineer, but my calc skills are too poor and there's no nearby educational facilities for EE. Gosh I wish I knew how to use a scope. Making handmade tuners is a bit slow going without a scope. Oh the darned coils! But, as a benefit of a few wars with that, aligning already built tuners is super easy. It was really good practice for power filtering too.
 
Last edited:
I wouldn't use small resistors for that.


I meant the small value resistors you use in your capacitor arrays. It makes sense.

I have used resistors for decoupling input and voltage gain stages before (you always should), but never to decouple the power supply from the output stage. I can see where it would help recovery time a little by "holding back" the discarge of some of the capacitors.

My present research mirrors some of your concepts. Apparently such line of thinking is unorthodox; I am unaware of any commercial equipment that uses capacitor arrays or the tweaks we discussed. But the arrays work better than a big cap for reasons that are so obvious once you analyze it. And I got the idea of "right sizing" the on board caps from you. It all makes sense.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.