Optimizing TDA7294 Output

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
^ Yes, something similar to this. The xformer, rectifiers and main big reservoir can be shared while the sections after series resistors or other series elements (and in GND nets also) would be seperated between PCBs, being located on-board. Then a local GND bus connecting PCBs would establish a quiet and common GND. All this applies in case of mono PCBs (or multichannel >2CH), with stereo PCBs things would be rather simple; PSU (P2P or PCB), 3 wires, Amp-PCB.

I'm trying to "shrink" the simple version down to a size that would fit all on one board, but without omitting features.

I momentarily forgot what the series elements did. Fast is fast and dc is dc, and man I really forgot something. My DC (umbilical) power cable is 3-conductor, 20ga solid copper, insulated, twisted, and. . . with a knot in it. Well, that doesn't fit on 1-board compact builds and we wouldn't want to explain the knot. :) Assuming MBR1035 diode drops (instead of the cable) are the series elements at the output of the big power reservoir capacitance, I forgot the knot. Perhaps a very tiny ferrite bracelet on the leg of the MBR1035 can provide the same function as the knot?
FB43-226-RC J.W. Miller | Mouser
Eleven cents isn't a bad price for an audiophile part, and it is prettier than a knot. :)

And I forgot that the diodes had another job. The motivation to use them (or regulators) is that we could use large caps near the chip, without a dulling effect on audio. Panasonic Aluminum Electrolytic Capacitors - Snap In | Mouser Those large caps have very low ESR. Given the cable with knot or the diode with choke, is it then suitable to use a pair of these big caps instead of the multi-cap bank? High density capacitance would be so much easier to fit on a single board design.

It seems to be a bit difficult to make a 1-board design without compromising audio. The next step could be to get the bridge rectifiers on board in the fashion used for ham radio, tuners and integrated receivers. I've tried this with a real receiver at far fringe. MUR bridge rectifier utterly ruins the reception. KBPC1004's with approximately 3n3 polyester dip caps (lossy colorful bubble) parallel with each diode, didn't bother the reception. Am I on the right track?
 
Last edited:
. . . PCB were used for only the chipamp circuit, that board could be mounted on short standoffs, OVER one end of the cap-array PCB. Then, the chipamp's power leads could go down through the small board and connect directly to the power planes.
This okay if amp front end is regulated. H o w e v e r . . .
We could try something without regulators, knowing that the omission needs a fix.

Problem:
If a simpler design is mounted direct to the power board, there's an omission. Usually an cable serves a filter function as an inductive resistor (filter) from amplifier enclosure to power enclosure. If that was omitted, it would be a downgrade to the typical sound of a chinese kit amp. This needs fixed.

Supplies:
If you remove regulators, diodes, and inductive cables from the list of possible series elements to use, that leaves: Resistor.

Fix?
When trying for a fix that doesn't involve a schottky diode, I ran for the bell wire, cat 5 cables, and ohmmeter. Short length 8" 20ga or 5" 24ga, got in the neighborhood of 0.12 ohms to 0.15ohms. That's got the resistor value of the cable, which is only one feature. Perhaps the RF filtering could be replaced with the little ferrite bead. So, the choked non-inductive resistor can replace the missing cable?
Question:
Quiet, non-inductive, and inexpensive. Will it do?
Perhaps it will be easier to measure than the audiophile play of selecting the perfect umbilical cable (filter) to connect amplifier enclosures to power enclosure?
 
Last edited:
Depending upon which we road the group takes here... If we want to experiment with the original amp PCB that both Bob and I used, I got a price/pair of $6.50USD from my vendor in China. I'm requesting a discounted price if we, say, order 20 pair - depending upon the level of interest. If we go with just the boards, we can further optimize performance using select components. I believe we could still make these PCB work with the PSU changes proposed in the earlier posts. I have not read Daniel's latest posts yet. However, if we want to go with a single, stereo PCB things obviously change. :)

Rick
 
For the reservoir caps, I would like to have consideration given to something similar to Terry Given's cap bank board. It uses a non-etched two-sided PCB. All you need is a drill. Use a slightly larger drill to remove copper from edges of holes for connections to bottom side. Cut cap leads, bend parallel to copper, and solder. Unless two PCBs were used, it would have to have all ground plane on one side and the power plane side would be split into two planes, for the power rails. The chipamp could be located over the line between the power planes. Check out the link and see what impedance Terry measured with his network analyzer!
That's a good an easy way to go. I have a scheme that is even more simple : Use a strip of double-sided blank PCB like 1" wide and 5" long and solder the caps directly along the two long edges,. This also is as low as it gets in term of inductance.

A network analyser is the weapon of choice to make a PSU that has low and consistant impedance, now that I have my own I learned so much about this in only a few weeks.... For example : an array of some 30 pieces of 47uF/50V of a moderate ESL type (Pana FC) with smallest lead spacing gives a completely flat profile up to 20MHz. No way I could achieve that with a couple of low ESR bigger caps with paralleled film caps, it always looked more ragged albeit a bit lower impedance in general.

---------:---------

As for the questions of etched PCBs or not, a well designed PCB always can cope with any other type of contruction and has the advantage that it is easy to recreate the original results for any builder even at low skill level. Also, cost is moderate if a batch of a dozen or so is made and if we don't use fancy options. The success of the MyRef FE is a good example, wiring up that thing on perfboard or P2P would be a hassle and problematic.
 
My vote is for both directions. The cost/scope creep reality, if fully adopted, will not work for some budgets (and some with self control and common sense :rolleyes:) and limit their participation. The kit, current PCB, is remarkable even in a stock build and IMHO, deserves as much attention/improvement as it can attract. That may be all some builders need or want. Those board prices look very attractive an would encourage me to participate in a ground-up revision of the components.

Let me suggest the possibility of a split - two separate threads if the interest and need develops.
 
My vote is for both directions. The cost/scope creep reality, if fully adopted, will not work for some budgets (and some with self control and common sense :rolleyes:) and limit their participation. The kit, current PCB, is remarkable even in a stock build and IMHO, deserves as much attention/improvement as it can attract. That may be all some builders need or want. Those board prices look very attractive an would encourage me to participate in a ground-up revision of the components.

Let me suggest the possibility of a split - two separate threads if the interest and need develops.
Sounds like a wise approach to me. Even when doing a new PCB layout, the designer needs to account for the size of the components, if not the exact ones he or she recommends. Though the external components are few, it's still something that whomever does the layout needs to be aware of. If we can use the stock PCBs and populate with better components then it seems to me a very viable approach with a minimum of risk and additional time involved with a new PCB layout, prototyping, fabrication, etc from the ground up. But I'm just one member speaking. :)

Rick
 
bcmbob said:
The kit, current PCB, is remarkable even in a stock build and IMHO, deserves as much attention/improvement as it can attract. That may be all some builders need or want. Those board prices look very attractive an would encourage me to participate in a ground-up revision of the components.
Bob, I think you can determine whether the kit boards will do or not, after you try some small signal options on the amp boards and also some power board options.

At some point, hardship in component selection might indicate that upgrading to excellent design onto a dual layer PCB with an excellent layout is both a better quality and a considerably less expensive way to go.
KSTR said:
A network analyser is the weapon of choice to make a PSU that has low and consistant impedance, now that I have my own I learned so much about this in only a few weeks.... For example : an array of some 30 pieces of 47uF/50V of a moderate ESL type (Pana FC) with smallest lead spacing gives a completely flat profile up to 20MHz. No way I could achieve that with a couple of low ESR bigger caps with paralleled film caps, it always looked more ragged albeit a bit lower impedance in general.
Separate power board it is then. Thank you! That answered my capacitor question thoroughly.
 
Last edited:
Let me suggest the possibility of a split - two separate threads if the interest and need develops.
Agreed, the thread here should stick to this stock PCB/kit in question, which I think is quite a good layout for a single-sided design.

After finishing a 2p2/deadbug proto, I'll open a new thread on my take of a simple but excellent 7294 (stereo) amp which will have some design features that are not seen typically. Since double-sided PCBs are only a few % higher price (if at all) than single-sided, single-sided design makes no sense to me. Further, with double-sided the integration of the PSU section(s) is far easier.
 
Agreed, the thread here should stick to this stock PCB/kit in question, which I think is quite a good layout for a single-sided design. Since double-sided PCBs are only a few % higher price (if at all) than single-sided, single-sided design makes no sense to me. Further, with double-sided the integration of the PSU section(s) is far easier.
I just bought a single channel kit on that exact same single sided board. It was unfortunately $3 more than a nicer dual layer board. But, I wanted to try the same board that Bob has to see what that board will do.
After finishing a 2p2/deadbug proto, I'll open a new thread on my take of a simple but excellent 7294 (stereo) amp which will have some design features that are not seen typically.
No idea why the board designers think the mute and standby circuits are more important than the power circuit--The P2P build helped me mainly by getting some 220u (or 270u) caps for power as close as possible to pins 7 and 8.
And, I'm eagerly anticipating seeing your new features! :D
I've been wondering about doing lead lag compensation (shunt comp) and/or anti-miller in order to get some more gain out of this chip so that modern sources can push it to 65 watts without having to do a preamp. Currently my efforts are foiled by having to set low gain to get higher quality except for the insertion loss of an inconvenient preamp. Help?
 
I've been wondering about doing lead lag compensation (shunt comp) and/or anti-miller in order to get some more gain out of this chip so that modern sources can push it to 65 watts without having to do a preamp. Currently my efforts are foiled by having to set low gain to get higher quality except for the insertion loss of an inconvenient preamp. Help?
I'm not a fan of letting chip amps run at any gain higher than needed for stability (about 20x), and if any possible with even lesser gain for the audio range. With ~30V supplies and 20x this results in an input sensitivity of 1.5V(peak), or 1.1Vrms which any home-hifi source should be able to deliver. For devices with lesser output (or if you need really high gains for 16Ohms speakers and such) a preamp is in charge, something like an inveting x5 stage built around an opamp. Another option would be a composite chip amp that has high initial gain capability to start with, but that's gonna be a topic for a future, advanced level project after my bare-bone TDA7294.
 
Update!!! I can get 20 pairs of stock PCB for $4USD/pair with postage of $8USD. Maybe even a lower unit price if we order more than 20. That's an incredibly good price for 2 PCBs! Food for thought.

Rick

Clarification.... That's $8USD for all PCBs shipped from China. Not to individual members whom want to take part. I think 2 of these boards can go First Class mail for less than $1USD. At least in the states.
 
I'm not a fan of letting chip amps run at any gain higher than needed for stability (about 20x), and if any possible with even lesser gain for the audio range. With ~30V supplies and 20x this results in an input sensitivity of 1.5V(peak), or 1.1Vrms which any home-hifi source should be able to deliver. For devices with lesser output (or if you need really high gains for 16Ohms speakers and such) a preamp is in charge, something like an inverting x5 stage built around an opamp. Another option would be a composite chip amp that has high initial gain capability to start with, but that's gonna be a topic for a future, advanced level project after my bare-bone TDA7294.
The chip likes about 23x, but there was a problem:

Ran out of computer before I ran out of power amp.
Computer was straining and clipping itself, which isn't fun to amplify.
Same for mp3 player
Same for multi-function phone
Same for SD player
Same for digital tuner
 
This okay if amp front end is regulated. H o w e v e r . . .
We could try something without regulators, knowing that the omission needs a fix.

Does it have to have regulators? If so they could be added.

Problem:
If a simpler design is mounted direct to the power board, there's an omission. Usually an cable serves a filter function as an inductive resistor (filter) from amplifier enclosure to power enclosure. If that was omitted, it would be a downgrade to the typical sound of a chinese kit amp. This needs fixed.

Daniel, getting rid of inductance and resistance is the point of a good power supply. The whole goal is the lowest-possible impedance, as seen by the power pins. Impedance is directly proportional to inductance and resistance, and inversely proportional to capacitance.

If you for some reason had to add inductance, then we would just have to try to cancel it with the decoupling caps. But that's an imperfect solution because we would then also induce unwanted resonances.

Supplies:
If you remove regulators, diodes, and inductive cables from the list of possible series elements to use, that leaves: Resistor.

No. That would leave: "nothing", which would be perfect.

Fix?
When trying for a fix that doesn't involve a schottky diode, I ran for the bell wire, cat 5 cables, and ohmmeter. Short length 8" 20ga or 5" 24ga, got in the neighborhood of 0.12 ohms to 0.15ohms. That's got the resistor value of the cable, which is only one feature. Perhaps the RF filtering could be replaced with the little ferrite bead. So, the choked non-inductive resistor can replace the missing cable?

How about using: Nothing!

Instead of your 0.12 to 0.15 Ohms, how about 0.010 Ohms, but 0.00 ohms would be perfect. Of course, maybe we need a little resistance, but only BEFORE the reservoir caps, so that it actually forms a low-pass filter with the caps. But that would already be in the windings, wiring, and fuses.

If everything is configured well, with minimized enclosed loop area (see Faraday's Law), then RF should not be an especially-bad problem. And that's what the RF filters are for, but they should be right before the input pins (and the power supply pins, and the output pin... Everything's an input, for RF.)

Question:
Quiet, non-inductive, and inexpensive. Will it do?
Perhaps it will be easier to measure than the audiophile play of selecting the perfect umbilical cable (filter) to connect amplifier enclosures to power enclosure?

We'll see.
 
Hi Tom,

apart the physical problems (size) for the cap array I understand that the (much) lower inductance is desirable for an 'ideal' power supply but I'm not sure it's a real improvement for sound quality.

For example stacked film capacitors, according Bateman measurements, have usually higher distortion than wound film caps...

And an array of capacitors is terribly similar to a stacked film cap in my mind...

Wouldn't a different approach like this one be better?
 
Hi Tom,

apart the physical problems (size) for the cap array I understand that the (much) lower inductance is desirable for an 'ideal' power supply but I'm not sure it's a real improvement for sound quality.

For example stacked film capacitors, according Bateman measurements, have usually higher distortion than wound film caps...

And an array of capacitors is terribly similar to a stacked film cap in my mind...

Wouldn't a different approach like this one be better?

Hi Dario!

Actually, there is almost no similarity to a stacked film cap.

A small 120mm x 120 mm array of 10x10 (qty 100) 1000 uF electrolytic capacitors, if constructed properly, has a super-low impedance, with an inductance component of about 1 nH. A stacked film cap is a single cap with some known "bad" issues for audio, and more inductance than the entire 100000uF array. How could there even be a comparison?

The main features of a cap array are provided by the closely-spaced power and ground planes (NOT a good antenna), the very LOW inductance of the planes, and the paralleling of many caps, which DIVIDES their individual ESL and ESR figure by the number of caps!

It's OK to be "not sure it's a real improvement for sound quality". The circuit WILL operate MUCH better. We just don't know (YET) if that will be enough to make any audible difference. But it's so easy to make! So why not do the BEST we can? (I am, by nature (or maybe by "aging"), a skeptical nerdy engineer/scientist type of guy. But, understanding the math and seeing that every different parameter moves in the most-desirable direction, even _I_ am excited by this concept, which I did not invent; excited-enough to at least test it, once. Why do you not simply test it for yourself?)

And my intuition is telling me that the soundstage imaging might be optimized by it, because IF we take the planes ALL the way to the chip power pins, the planes will also behave like ideal decoupling caps, and so, then, all of the frequencies in the transients will be reproduced with the correct (or best-acheivable) phase angles and gains (i.e. correct TIMING, and correct amplitude vs time, combined, giving very-accurate reproduction of edges and changes in the signal), which would mean MINIMAL distortions of the temporal (time-related) cues that our brains use to create the image.

And there are a lot of golden-earred people here. I'm sure that if there is an audible difference, someone will tell us, very soon. So you can simply wait. I myself am getting ready to build one, maybe this weekend if I don't have to spend too much time helping my friend hang concrete board in her bathroom, for the new ceramic tile she is going to put on the walls. But she is a very close friend so if she calls I will be there. Anyway, my ears are actually still pretty damn good, even though I have lost some high-frequency hearing, because, I am assuming, I have somehow gotten relatively old, despite doing almost everything I was ever told not to do, which was supposed to have killed me a long time ago. I am smoking and drinking as I type this. (But at least I quit most of the really-bad habits I used to have.) <grin>

I have to say that I don't think I like the idea of 100000uF (the 10x10 1000uF cap array). So I might scale it down, probably by using lower-valued caps. I just don't want the huge inrush-current peaks that such a low-impedance and high-capacitance array would cause, and would like to try to avoid inserting an inrush current limiter, if possible. I also don't believe that 100000 uF of reservoir capacitance is necessary, almost ever, for audio amps of up to a couple hundred watts at least.

I might try 100uF to 330 uF, in a 10x10 array. 330 uF x 10 x 10 might be very good. And maybe it will be half the size, or less, of the 1000uF array. 60mm x 60mm doesn't seem too large, does it? That would be something on the order of twice the dimensions of a 33000uF cap's footprint, I think, with probably less than 1/10th to 1/20th as much inductance. That seems like a good trade, since inductance is the main enemy. (And really, I am not worried about the size at all, compared to the inductance/impedance.)

I tried that approach you gave a link to, several years ago or more, in LT-Spice simulations, but could never see much of an effect from it. Maybe I should re-visit it, because my simulation skills and abilities are far better than they were back then. Still, it looks a little bit like a trick. I didn't do the math, back then, either, to analyze it. Maybe I will try that, first. (I need the practice.)

Cheers,

Tom
 
Last edited:
And my intuition is telling me that the soundstage imaging might be optimized by it, because IF we take the planes ALL the way to the chip power pins, the planes will also behave like ideal decoupling caps, and so, then, all of the frequencies in the transients will be reproduced with the correct (or best-acheivable) phase angles and gains (i.e. correct TIMING, and correct amplitude vs time, combined, giving very-accurate reproduction of edges and changes in the signal), which would mean MINIMAL distortions of the temporal (time-related) cues that our brains use to create the image.

[...]
I might try 100uF to 330 uF, in a 10x10 array. 330 uF x 10 x 10 might be very good. And maybe it will be half the size, or less, of the 1000uF array. 60mm x 60mm doesn't seem too large, does it? That would be something on the order of twice the dimensions of a 33000uF cap's footprint, I think, with probably less than 1/10th to 1/20th as much inductance. That seems like a good trade, since inductance is the main enemy. (And really, I am not worried about the size at all, compared to the inductance/impedance.)
Hi Tom,

Your ideas are spot on, IMHO. A cap array of approbriate 'lytics on coplanar planes brought to the chip directly also looks best to me. Good caps for this are the tall and thin ones with the smallest pin spacing, 2.5mm or 3.5mm.
Panasonic FCs with 3.5mm and 35V, 50V resp. would be 150uF, 82uF resp.
Several dozens of these in parallel will give really low ESR and ESL at mid and high frequencies up to 10MHz or so, real world, right at the chip. Then we might add a couple of big high capacitance types to beef up the low-frequecy domain. Experimentation and measurements will show which combinations, location points (off-board vs on-board) and connection methods are winners.

-----:-----

When thinking about decoupling of chip amps I find extremly useful to abandon any notion of a big low impedance lumped GND (plane or whatever) as the live saver.

We must look from the chip's view instead, and it doesn't know of GND. It sees voltages on 4 inputs : +IN, -IN, V+, V-.
We need to stabilize and lock V+ vs. V- first, with a low impedance path, and then place the INs at levels where they move little and linear with respect to the supplies, notably against the negative supply which is the chip substrate that should be quiet as all internal nodes have (non-linear) stray capacitance to it. Any unavoidable deviation of the INs potential (vs V-) should be either slow (low freq) or small and linear (== a linear copy of the input signal) for higher freqs.

GND is not a part of the picture. At the chip decoupling point, the current path just simply runs through it but not alongside, there is mostly cross current and only little load current. Load current is injected at the point the return wire from the speaker hits GND and where we should take our GND reference from, on the way back to the first cap after the diodes.

For anything else but a monoblock we have the "dilemma" that we need GND both at the chip's supply pins to be able to route "through it" as well as at a common reference node which is further away from the chips. The situation is further complicated in that the traces returning GND from the chips to the common will be polluted with supply recharge currents causing I*R drops, depending on the PSU strategy. That's a reason why small local caps may fare better than big ones, notably when the board's GND reference is derived locally at the GND entry point. We can imagine that it is not trivial to fulfill the mentioned conditions for the four input pins for each chip in a multi-chip setup (stereo and/or several PCBs) in order to avoid any possibility of unwanted noise pickup.

There are alternatives to this standard scheme making things easier to get factored out, I'm working on it right now...
 
Amplifier PCB problem: The majority of the boards that I've seen assume that the mute and standby circuits are more important than the power circuit, with the unfortunate results of the power caps ending up farthest possible from the chip.

Question: If a new design was made, would you suggest an order of importance, most important first, like 220u power cap as close as possible to pin8, 220u power cap as close as possible to pin 7, little power bypass cap close to pin15, little power bypass cap close to pin 13, RCR from pin 15 to pin 13, bootstrap cap close to pin 6, feedback resistor trackside without via, and the rest TBD?

Question: I do mute, pin10 with only a 10k resistor but I'm curious about an alternative: If pin10 is not powered, then does pin4 become the non-inverting input as the datasheet suggests, or is there a caveat? I didn't try it--longevity worries.

P2P speed/ease: I like having the power pins, 7, 8, 13, 15 bent upwards near the chip face, and I like KSTR's suggestion of double-sided copper board for more compact rails. If I were to make a new P2P publication, for increased workspace, I'd position the TDA7294 pins 1, 2, 3, 4, 6, 9, 10, 14 hanging down below the heatsink so that both front and back of pins can be used for much easier faster build.
Clarification.... That's $8USD for all PCBs shipped from China. Not to individual members whom want to take part. I think 2 of these boards can go First Class mail for less than $1USD. At least in the states.
Does it come the chip, thermal pad, and shoulder washer so that the board can be used by anyone who has a capacitor and resistor collection?
Board+ChipAmp is interesting, but I have noticed a lack of interest in just the board by itself.

I wonder which of the available boards have the amplifier board's power caps as close as possible to the chip?

TDA7294 65W Mono Amplifier Board Kit 28 | eBay
1 PC x Assembled Finished 85W TDA7293 Mono Power Amplifier Board Support BTL | eBay
TDA7294 85W 85W Mono Audio Stereo Power Amplifier Board 1 1 Two Channel Y199 | eBay
YJ Mini 85W 85W TDA7293 x 2pcs Amplifier Board New | eBay
Mono TDA7294 Audio Power Amplifier Kit ?One Channle Kit | eBay
2pcs TDA7293 60W 60W Amplifier Kit LJM 1 | eBay
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.