An additional option on that supply is to bypass the pi-filter resistors with MBR1645 schottky. That will maximize the "bass slam" during high volume playback and the crc filter does still work well during lower volume playback.CAN YOU HEAR THAT ????
I almost forgot what all the power supply stuff was about.The amps are playing, the bulb tester is out and everything is at room temps.
We had a problem with that. Some people got real chips and some got fake chips. One person even had a real on the right and a fake on the left. Most of us buying from eBay, got the fake TDA7294 chips.
The chips that we bought can be very useful so long as we don't try to use them like a TDA7294.
The solution is at the Radio Shack.
It is the 12+12 2a transformer.
Yes, much lower voltage is the solution.
The fake chips are stable at very low voltage, tolerate low gain, and in that combination, may do fantastic quality. You'll get enough output power for some relaxed level listening or for good use with desktop and full range speakers. The chip's behavior is typical of a die-shrank clone.
It will very likely make a charming system. . . at a great deal smaller scale than you wanted. Probably, the smaller scale end results will go along in direct proportion to the die-shrank. I have no idea if some mystery cloners did this or if ST may have die-shrank their own product? So far, a chip matching TDA7294's datasheet is not available new on the market anywhere (except NOS).
Or, if you want to use a chip with a too small die at too much voltage and too much load, that requires lower current at small signal, such as 110k feedback resistor and its partner at 2k7. Or with your slightly lower voltage (22+22vac transformer), 100k feedback and 2k7 feedback-shunt. That compensation will correct the tonality.
Now, that was worth a mention because if you keep going, lower the voltage and likewise lower the feedback resistor value (for lower gain), then by the time you get down to a 14+14vac transformer, and 68k fb with 2k7fb shunt (we kept the 2k7 for the convenience of having the in- coupling cap smaller than a Winnebago), then the gain is now low enough for rather involving imaging.
But, keep going down to the 12+12vac transformer and see if you can change the feedback resistor down to 56k (with 2k7) and have the amp still stable at that gain setting of 22x. That will need checked a bit. That low gain at such low voltage gives penultimate efficiency and imaging realism performance, which are excellent features, but they will cost most of the output power. Even so, I think that's what it takes to find a use for it.
To get the gain down that low, you may have to flip the board over and stick the 220u caps on the pre-drive power pins directly. At that required low voltage, 25v caps can be used, and they're easier to fit. For a little more bass, you can put 220u caps on all four of the chip's power pins.
It is probably necessary to run the feedback resistor directly across the chip and run the feedback-shunt resistor from in- directly by shortest route to the in- coupler cap (that is also 220u, 25v; however, the gain is so low that a wrong size smaller cap will still work if the in+ coupler is 1uF).
So, that under-board area may get a bit crowded.
The good news is that to fix up a stereo pair, 10 of 220u 25v caps (5 each board) may just get to the volume discount pricing.
The resulting small scale hi-fi, is excellent for use with miniature full range speakers (and a subwoofer).
I use them on purpose in my table radio projects. The output is really nice and the heatsink expense is minimal. A flattened beer can is far more than enough, or you could use an old northbridge cooler. Yeah, it runs that cool. The only actual need of a heatsink that I noticed is for keeping the caps cool.
Even though the datasheets don't work for them, I had wished to be able to buy both chip varieties on purpose. But I wasn't prepared to get what I wished for. The label on the outside of the package is inconvenient! Want the little imaging champ small scale table radio variant? Just buy almost any TDA7294. Or did you want a real TDA7294? The easiest way to do that is buy a TDA7293.
Want a TDA7296? Well, I've no idea how to buy one on purpose, but I'm sure we've bought them by accident a few times. P.S.
Four different versions of TDA7293 have been seen, with all 15 pins working, no application troubles and no dramatic differences other than the one generic that I spotted is a little slower, easier to stabilize and can use slightly lower gain. No problems so long as the board supports the TDA7293-specific bootstrap cap hookup (33u per chip). These TDA7293 work much closer to the datasheet specs. The eBay part is identical to the Mouser part. So, if you'd like to reboot this concept using the TDA7293, I'm up for it.
with all these fake/clone/pull tda72xx chips to sort out, you know what id did? i finally gave up on them and went all-in with TI Overture series parts; avoided the Spike issues by paralleling enough of them for any given application to run well within the safe operating areas shown on the datasheets. have yet to hear a peep out of Spike.
Such as a 1v source with enough headroom for clean output? Of course, but that's not the main point.
Tone/stability would be the main point.
The gain setting is the main compensation for a chip amp.
TDA7294 gets very sketchy near max power voltage. So your option for lower gain and still have a useful tone, can be done by much lower power supply voltage. The amp behaves far better at 20+20vdc (and less). Then you can set lower gain.
The TDA7293 is better behaved and doesn't need quite so much gain. Even at near maximum power voltage, I got it down to 38x easily, and was able to trim it down to 36x. With a modest amount of de-rating (a bit lower voltage), you could get a lower gain amp, with 40-ish watts. That is favorable.
Also, go look at the Honey Badger's 42X.
I like those as parallel amplifiers! Sounds great! There's no squeaking!
Kudos!I can certainly appreciate that your phase margins are off, but even with a 1 Vrms source on 34 Vdc rails, you need a maximum gain of ~25 to drive the amp into clipping. You're suggesting double that amount of gain.
That reeks of bad engineering/layout, especially considering the 80 db of DC gain. Use lead-lag to hit an intercept at Gv=50 instead--at least then your signal gain structure is better set and you're applying more feedback over most of the signal range. And not sure what on earth you mean by "tone".
Assume, using the recommended parts (Gv=32), that you have a phase margin of 45 degrees and a GBW of 1 MHz (half the minimum value of the "equivalent" 3886). Gv_min is 16 (24 dB). Where does that put you?
That reeks of bad engineering/layout, especially considering the 80 db of DC gain. Use lead-lag to hit an intercept at Gv=50 instead--at least then your signal gain structure is better set and you're applying more feedback over most of the signal range. And not sure what on earth you mean by "tone".
Assume, using the recommended parts (Gv=32), that you have a phase margin of 45 degrees and a GBW of 1 MHz (half the minimum value of the "equivalent" 3886). Gv_min is 16 (24 dB). Where does that put you?
Honey Badger is designed to drive 150 W/8 ohms. And is designed around Vpp_in ~1.2V. Most equipment can overdrive this, but it's sensible enough. I'd be aiming to design around 2Vpp, knowing a fair bit of equipment can actually output 2.8Vpp. I have little desire to waste that gain.
35 V rails for a 7294 give us ~70W/8 ohms (before the thing entirely falls apart, and I'm being generous). Using that same Vpp_in ~1.2V, we'd need Av = 28 to do it, which is pretty much datasheet suggestion. Not Av = 50.
35 V rails for a 7294 give us ~70W/8 ohms (before the thing entirely falls apart, and I'm being generous). Using that same Vpp_in ~1.2V, we'd need Av = 28 to do it, which is pretty much datasheet suggestion. Not Av = 50.
So. . .Parts list:
Per each TDA7293 amp
4 of 220uF 50v* decoupling (individual power decoupling caps not larger than 220u)
2 of 330uF 16v in- coupler (or a dc tracker)
a 1k multi-turn cermet trimmer for feedback-shunt
25k for feedback resistor (range 25k~27k)
25k//470p for input load (range 22k~25k)
10u and 22k for standby
10k (range 10k~15k) for mute (no cap!)
1u for input cap
33u, per each chip, for bootstrap, using TDA7293-specific hookup (pin6, pin12)
*There are 4 power pins, so there are 4 of 220u power decoupling caps. For layout, you can either situate them as close as possible to their respective pins, or you can gang them up at pins 7 and 8. Either way is fine, so long as the first pair is very very close to the chip, especially at pins 7 and 8.
See also, post#30 on my TDA7293 Parallel amp thread, since that power circuit hack allowed me to set lower gain (with the trimmer on the list above).
The TDA7293 will perform best on the smallest possible board.
If when the transmitter's signal suits the receiver.
It goes well, so long as you avoid the datasheet schematic, which is for making cheap mass market black box audio for sale at best buy and wal-mart.
Therefore, I've put more useful values on that list above.
Neither you nor I should decide this without asking the amplifier. The means to do so is a trimmer at the relevant spot (on the list above) and then some measuring.
On this thread here, if you scroll through and find that big photo of the old Memorex ad (the guy almost blown free from his chair), you can see where I've soldered every 10% resistor value for the gain divider, and took notes about the differences observed. Yes, a whole pile of resistors were all tried. That section got down to 47k. . . and then a few posts later I picked it up again, with all of the rest of the resistors. One particular combination with 27k feedback was most favorable of the lot.
After adding the power circuit tweak, that allowed the gain to be set lower.
And then I added the trimmer, which also allowed the gain to be set lower.
Then I try slightly lower power voltage, which allowed me to change the feedback resistor to 25k, also lower gain then too.
I've been working on it, very slowly. If you arbitrarily set the gain down too low, the output is too similar to a hungry basset hound at dinnertime. So, the matter will take a little more care and a bit of doing. The trimmer serves to balance the relevant factors most conveniently and the process of setting it will be highly informative. Then you'll find out why I set the gain higher than suggested by the datasheet.
Last edited:
Thanks, Daniel.
Very thorough and helpful reply.
I'm going to take it a bit different direction, as I kind of want to have fun with the engineering side of things, and have a specific purpose intended for the system. It's been a long time since I've really done up a circuit (and never a layout, so that'll be good "fun"). Or I blow up a bunch of chips and learn some things along the way. Given I've had these parts just laying here for years, it's a win-win either way.
Recommendations as far as decoupling/mute/standby/bootstrap are certainly appreciated, however!
(Another) Daniel
Very thorough and helpful reply.
I'm going to take it a bit different direction, as I kind of want to have fun with the engineering side of things, and have a specific purpose intended for the system. It's been a long time since I've really done up a circuit (and never a layout, so that'll be good "fun"). Or I blow up a bunch of chips and learn some things along the way. Given I've had these parts just laying here for years, it's a win-win either way.
Recommendations as far as decoupling/mute/standby/bootstrap are certainly appreciated, however!
(Another) Daniel
defective cap (way below marked value) on the positive rail.
so it charged to full potential long before the negative rail.
chip blew immediately. it had been working ok on an older PSU.
after that incident i used LEDs on each rail, to have a visual on the start times.
(although this seems to be inessential with the TI chips i now use).
another way to blow a 72xx chip, if direct-coupled:
asymmetric clipping -- if very severe -- may slam the output into a rail.
very loud deep bass plucked instruments (ex: bass line in the intro to "Time" by Pink Floyd--i've never heard that intro bass guitar sound, all the way up and down the scale, convincingly like a real live bass guitar, with any capacitively coupled amp)
asymmetric clipping -- if very severe -- may slam the output into a rail.
very loud deep bass plucked instruments (ex: bass line in the intro to "Time" by Pink Floyd--i've never heard that intro bass guitar sound, all the way up and down the scale, convincingly like a real live bass guitar, with any capacitively coupled amp)
Last edited:
A different way to mute
Although I would do it like post 631 (only one resistor, no cap, for the mute circuit), here's a different way:
10k series to zener
That will make the mute switch the last event during startup, and also the first event during power down. You won't need a cap for that either--the power circuit already has plenty and the power ramp up/down delay times are sufficient.
Although I would do it like post 631 (only one resistor, no cap, for the mute circuit), here's a different way:
10k series to zener
That will make the mute switch the last event during startup, and also the first event during power down. You won't need a cap for that either--the power circuit already has plenty and the power ramp up/down delay times are sufficient.
Daniel--nothing top secret. As I said, I'm scavenging the amplifier section of an old, dead, Logitech z680 system (controller died, so friend got a replacement ages ago) for parts and will be doing this on the "cheap" by reusing as absolutely as much as I can from that (PSU + heat sinks/case + 7293's being the biggies). Hopefully a few op-amps, some caps and resistors (and ultimately pcb's) will be necessary. Famous last words, though.
This will be to drive a 2.1 setup that reuses the subwoofer from said system and some "simple" corner horns driven by full rangers. I'm using an old desktop with a decent sound card (that does 2Vrms) to serve as the media server + DSP.
Ultimately, I'm looking at a composite setup driving parallel chips and DC servo correction. Probably the LM49710 like everyone else and a TL072. (Why mess with success?!) Might try to bias the parallel chips against each other to pull them up (just a hair) and mitigate some of the crossover distortion. Will start with just the basic 7293 with point-to-point, though, and try to figure out how much lead/lag it needs to get the gain down closer to 12-15, then add the servo, then use the 49710 as a unity-gain buffer to drive the parallel chips (think this will be a much better overall solution than using the 7293's intrinsic paralleling, but that remains to be determined), then finally composite.
If things are working well along the way, I might skip some steps. Or, I might stop halfway and call it "good enough." Then again, 3/4 of the fun will be in the engineering, as I doubt a composite design will really help.
More or less porting Tomchr's good work on the 3886 over the the 729X series as much as I can. Short of buying his kit I don't think there's much he's left hidden on layout/bypassing. Props to him on that.
But there's no good model for these chips in SPICE and a LOT of missing data on gain/phase margins to work out how to stabilize these chips and I lack the hardware (at least that I own, will be bugging some of my friends in analog circuits labs to sit on some of their test equipment) to be really digging into these chips, so it's going to be very, very conservative design.
Steve--thanks. Good information to keep in mind.
This will be to drive a 2.1 setup that reuses the subwoofer from said system and some "simple" corner horns driven by full rangers. I'm using an old desktop with a decent sound card (that does 2Vrms) to serve as the media server + DSP.
Ultimately, I'm looking at a composite setup driving parallel chips and DC servo correction. Probably the LM49710 like everyone else and a TL072. (Why mess with success?!) Might try to bias the parallel chips against each other to pull them up (just a hair) and mitigate some of the crossover distortion. Will start with just the basic 7293 with point-to-point, though, and try to figure out how much lead/lag it needs to get the gain down closer to 12-15, then add the servo, then use the 49710 as a unity-gain buffer to drive the parallel chips (think this will be a much better overall solution than using the 7293's intrinsic paralleling, but that remains to be determined), then finally composite.
If things are working well along the way, I might skip some steps. Or, I might stop halfway and call it "good enough."
Then again, 3/4 of the fun will be in the engineering, as I doubt a composite design will really help.More or less porting Tomchr's good work on the 3886 over the the 729X series as much as I can. Short of buying his kit I don't think there's much he's left hidden on layout/bypassing. Props to him on that.
But there's no good model for these chips in SPICE and a LOT of missing data on gain/phase margins to work out how to stabilize these chips and I lack the hardware (at least that I own, will be bugging some of my friends in analog circuits labs to sit on some of their test equipment) to be really digging into these chips, so it's going to be very, very conservative design.
Steve--thanks. Good information to keep in mind.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.