Bootstrap hookups are very clever that way!
Mute and standby are extremely low current--the mute trace can be small enough to fit in-between the pins.
Connecting Mute to Clip Detect takes a few more parts than shown (as shown with the board connections, that may cause louder, more frequent clipping), so let's just not connect them together; So, don't connect pin5.
Try to "scoot" the decouplers a little bit closer to Pin7 and Pin8. Every little bit counts. Also, consider double-size (thickened) trace between chip and C8.
C2 and C3 are 16v 470uF computer caps, and the same physical size as C5.
Mute and standby are extremely low current--the mute trace can be small enough to fit in-between the pins.
Connecting Mute to Clip Detect takes a few more parts than shown (as shown with the board connections, that may cause louder, more frequent clipping), so let's just not connect them together; So, don't connect pin5.
Try to "scoot" the decouplers a little bit closer to Pin7 and Pin8. Every little bit counts. Also, consider double-size (thickened) trace between chip and C8.
C2 and C3 are 16v 470uF computer caps, and the same physical size as C5.
Mark, I just remembered more stuff!
If you put a 4.7uF 250v cap (one cap, preferably cornell dubilier mallory SEK, 6 cents) across the V+ and V- on that screw terminal connector, the DC input, then the stability will go SO much higher that you'll have to increase the input load (slightly) and decrease the gain (slightly) as compensation (so as to return to "conditional stability" sweet spot for audio).
Actually, we didn't need the arc fault noisy slippery screw connector at all, but it could be fun to put on that cap.
Alternative is FB43's, same locale, and even though that is more elegant looking than a rail-2-rail cap, I believe that nobody would buy it; so, try the cap.
Anyway, see why I put a trimmer?
I also remembered that there is potential trouble if the gap between the bootstrap versus everything else isn't big enough. Perhaps double-up the insulating space for everything bootstrap related?
The speaker return should be at the 0v dc input connection at the amplifier board. So, I suggest that you don't provide extra connections for speaker return--the right spot is just the same spot where the dc umbilical cable provides the 0v contact for the board. No extra connection needed on the board. Large signal return point = Large signal return point--there's just not a better place for that. Unlike most of the other designs, or all of them, this thing has a heck of a lot of bass force. That current has to be sent to the power board, via the umbilical cable. The awesome recoil from a couple of these at max, is, in fact, enough to blow a window into the yard/garden. I tried it already. It has enough power. That much doesn't need to cross the board--it has to go to the umbilical cable.
Also, for DC tracker avoidance, R6 needs a 100n-ish cap paralleled with it, materials of either ceramic or polyester. This doesn't need extra vias--either surface mount pads or simply a footnote will suffice for it. That extra cap is to prevent treble droop at inverting input coupler. That's just exactly how the extra-looking tiny cap helps to remove the need of a DC tracker.
I suggest copperless, insulating, vias for pin5 and pin11.
If you put a 4.7uF 250v cap (one cap, preferably cornell dubilier mallory SEK, 6 cents) across the V+ and V- on that screw terminal connector, the DC input, then the stability will go SO much higher that you'll have to increase the input load (slightly) and decrease the gain (slightly) as compensation (so as to return to "conditional stability" sweet spot for audio).
Actually, we didn't need the arc fault noisy slippery screw connector at all, but it could be fun to put on that cap.
Alternative is FB43's, same locale, and even though that is more elegant looking than a rail-2-rail cap, I believe that nobody would buy it; so, try the cap.
Anyway, see why I put a trimmer?
I also remembered that there is potential trouble if the gap between the bootstrap versus everything else isn't big enough. Perhaps double-up the insulating space for everything bootstrap related?
The speaker return should be at the 0v dc input connection at the amplifier board. So, I suggest that you don't provide extra connections for speaker return--the right spot is just the same spot where the dc umbilical cable provides the 0v contact for the board. No extra connection needed on the board. Large signal return point = Large signal return point--there's just not a better place for that. Unlike most of the other designs, or all of them, this thing has a heck of a lot of bass force. That current has to be sent to the power board, via the umbilical cable. The awesome recoil from a couple of these at max, is, in fact, enough to blow a window into the yard/garden. I tried it already. It has enough power. That much doesn't need to cross the board--it has to go to the umbilical cable.
Also, for DC tracker avoidance, R6 needs a 100n-ish cap paralleled with it, materials of either ceramic or polyester. This doesn't need extra vias--either surface mount pads or simply a footnote will suffice for it. That extra cap is to prevent treble droop at inverting input coupler. That's just exactly how the extra-looking tiny cap helps to remove the need of a DC tracker.
I suggest copperless, insulating, vias for pin5 and pin11.
It is disappointing that so many compromises are necessary. Moving parts just makes it more difficult to build and then you are better doing a P2P build.
For C2-3 I would use Nichicon ES bi-polarized (green) caps and they are slightly larger.
I don't see the advantage of placing the caps close to the chip pins and then placing the return at the other end of the board. The total loop includes the path from load return to the capacitors. I want the smallest loops possible.
For C2-3 I would use Nichicon ES bi-polarized (green) caps and they are slightly larger.
I don't see the advantage of placing the caps close to the chip pins and then placing the return at the other end of the board. The total loop includes the path from load return to the capacitors. I want the smallest loops possible.
Last edited:
Long routes for the current are among the most common failing of pcb layouts.
We have absolutely superb layout technicians on this Forum. Unfortunately they sometimes don't understand why circuit route length and circuit loop area are important to performance.
Typical layout error are power traces along opposite sides of a PCB and a power ground trace that runs around three sides of the PCB. The loop areas and the route lengths are enormous.
These large distances then ruin the performance of local decoupling.
This and your next post are riddled with errors. But I'll comment on your first statement.
A bootstrap using a capacitor is a positive feedback system.
It does the opposite to what negative feedback can achieve.
Bootstrapping can exaggerate performance errors, whereas NFB can reduce performance errors.
It can make stability margins smaller.
In the worst case it can cause oscillation.
Yes, bootstrap hookups are very clever!
Nichicon ES (fw, pw) worked great at bootstrap.
Random recycled polar computer caps worked better at C2-3.
Everything on the board returns to small signal, except for the speaker, which is large signal. Just sayin. The tweeter needs decoupler power, the woofer needs power board power, and the exact spot that does both at once is where the dc umbilical cable joins the amplifier board. That is the uncomplicated spot to do all of those jobs simultaneously. So, that is the logical speaker return point.
The higher parts count of quad rail design does provide more flexibility in that matter, but does not provide higher audio resolution than the simplified style done with care. Both are extremely similar on return points. Unfortunately, simplifying the quad rail design does also happen to obfuscate a lot of design, via averaging.
I have already split the small signal ground from the power/speaker ground with R6.
Consider that the smallest caps are charged by mid caps and they are in turn charged by the largest caps. Each supply and charge.
And, R6 needs to connect to the speaker return. Connecting anywhere else adds (a voltage) error to the feedback signal.
Consider that the smallest caps are charged by mid caps and they are in turn charged by the largest caps. Each supply and charge.
And, R6 needs to connect to the speaker return. Connecting anywhere else adds (a voltage) error to the feedback signal.
That location--those vias at R6, is close enough to X1-2 to work just fine for speaker return. No problem there.
So, then physical size of C2-3 (16v 470u caps are a lot smaller than pictured) and using a really skinny trace for mute so that pin5 isn't directly connected to mute.
Need copperless (insulating) vias for Pin5 and Pin11
So, then physical size of C2-3 (16v 470u caps are a lot smaller than pictured) and using a really skinny trace for mute so that pin5 isn't directly connected to mute.
Need copperless (insulating) vias for Pin5 and Pin11
Last edited:
Yes it can cause oscillation! That is sure to happen if other traces are too close to bootstrap traces.
Thanks for the reminder!
Mute and Standby circuits
So, what about just disabling these two circuits by feeding them directly from the V+ source.... it's a bit of overkill, since the functions turn off at 3.5v according to the datasheet, but it saves space on the board and a few resistors and capacitors - any opinions?
Sixto -
Minne-no-snow-ta.
So, what about just disabling these two circuits by feeding them directly from the V+ source.... it's a bit of overkill, since the functions turn off at 3.5v according to the datasheet, but it saves space on the board and a few resistors and capacitors - any opinions?
Sixto -
Minne-no-snow-ta.
If you connect them directly together, the audio quality is hindered. I think that could be because the mute circuit is at input, but the standby isn't. There's also possibility of an accident if wrong-order startup.
So minimum parts looks like this:
Mute a resistor;
Standby a resistor and a cap load for delay.
or, here's a different possibility (not tested yet)
Mute a resistor;
Standby a resistor series to a zener.
Theoretically, if that zener is worth at least 2/3rds of 1 rail's voltage, then the power supply reservoir's time to charge, has also provided for standby delay. However, a zener has to be "biased" strongly enough to prevent stuttering (that generates unwanted signals), so the resistor value may have to be adjusted.
notes:
At least two separate resistors are needed.
You can connect a 10k~22k smd (or 1/8W) resistor from pin10 (mute) to pin7 (v+).
Pin 5 and Pin11 don't require a via at all for this application.
Pin 6, pin12 and their traces must not be crowded.
Standby, pin9, is "on front" and is therefore not difficult to route.
Last edited:
I've been using both circuits disabled for a long time, because an speaker protector is taking care of the mute on power-on and power-off.
I'm not using stand-by either...
I've designed an small stereo PCB using TDA7294 with mute and stand-by disabled and it works ok.
Attachments
Options, options!
Hi Ezavalla, thanks for posting the images!
When you say speaker protector, do you mean a separate PCB just for that, or a capacitor in series with the driver line out? I like the stereo configuration, I am assuming it could be adapted to run 2-way mono as well. The photos look great! I'd like to learn more about your build: Did you etch the boards yourself? What process did you use to transfer the layout? What voltage are you feeding it? How does it sound?
Sixto
Hi Ezavalla, thanks for posting the images!
When you say speaker protector, do you mean a separate PCB just for that, or a capacitor in series with the driver line out? I like the stereo configuration, I am assuming it could be adapted to run 2-way mono as well. The photos look great! I'd like to learn more about your build: Did you etch the boards yourself? What process did you use to transfer the layout? What voltage are you feeding it? How does it sound?
Sixto
Hi Sixto!
Yes... the speaker protector is another PCB which has been designed using the well known uPC1237.
About the boards:
I've etched the boards drawing them using the "toner transfer" technique (I guess that is the name of the process). It´s necesary to be careful enough to align both sides, but it isn't that difficult . Anyway, the boards have been designed to have a very low count of "vias" and to use the component leads to swap connection between both sides of the board (yes.. I know it isn't the best design, but my country has almost closed every way to import duties and I cannot access to PCBs at low cost... I hope this situation change by now).
The testing have been done with +/-25V DC but it's possible to take it to +/-31V DC because the transformers have a pair of additional taps.
And the sound... is good . I'm not used to describe the sound using words, and my system is a fully multiamplified 3-way one, so tests have been done on a limited frequency range.
I'm currently building the chassis to mount all the amplifiers (this is an 8 channel box) and heatsinks are "big", so it's neccesary to design a box big and rugged enough...
Kudos!!! Very good!!!
That (under-volting) is exactly how to get the current-production TDA7294 chip to run stable. And, it can be really rewarding when it does.
It (under-volting) costs much of the output power; however, in trade for that, lower gain becomes practical (without tone/stability foobar) and high fidelity resolution becomes available.
In my experiments, I got down to +-19.5vdc before the datasheet minimum gain figure was available. Rather the using it for an under-powered home hi-fi, instead I used it for a nicely overdone table radio. For sure, the bass power isn't enough to flap the garage door; however, the imaging/realism/resolution is of maximum quality, and easily broadcasts very clear audio throughout a rather large room.
That does exist! And I like it.
Also noteworthy, is that the expense for the power, chip, and heatsink, were all tiny.
It did encroach Class D on efficiency. Some difference remained, but it (severely under-volted TDA7294 class aB) was *approximately* as efficient with an 8 ohm speaker as a Class D would be with a 4 ohm speaker. Truly, the heatsink expense was minimal.
On that basis, I guess that it could also do a useful job as a midrange&treble amplifier, so long as there are passive crossover parts sufficient to reduce the load. With my 8 ohm full range, I did use 3300u series, a BSC (notch with resistor bypass) and I had a 220nF little polyster dip cap, located at the speaker and in parallel with it, so as to reduce cone breakup (a concern of full range speakers).
Similar amplifiers include LM1875, TDA7295, TDA7296. Of that lot, the TDA7295 (probably labeled "Singapore" since most TDA7295 are authentic), would probably love your +-25vdc supply.
The observations in this post were done with point-to-point builds (previously, I made a thread on building it easily), and, importantly, the amplifier that was observed couldn't have either the advantage of a good board nor the problems of a bad board, since there wasn't any board involved.
P.S.
If we wanted higher voltages to run practical amplifiers, then I think we'd better use the TDA7293 chip. It is a lot easier for that. And, yes, enough power to flap the garage door. The range of applications does differ surprisingly.
Last edited:
I've tested the amplifiers with +/-31V DC on a 4 ohms load and they run pretty well, but the temperature starts to increase very fast. The datasheets shows that behaviour anyway, but as the only 4 ohms load I'm going to use is a pair of tweeters, I preferred to take +/-25V DC in order to live a little in the "safe" side.
The gain I'm using is about 27 dB, which gives a 1V sensibility for full output at +/-25V DC.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.