When building a current limited chip bridge amp, you want to cut down on the voltage to bring the power output right in line with twice the output. In other words, choose a voltage for a 4 ohm single ended amplifier for a bridge driving 8 ohm loads. Then double the current = double the power. This raises efficiency and produces far less heat.
If you want to go hog wild, you could bridge-parallel and get 4 times the power. I always wanted to do this and I just might.
You've got that right.
Has anyone ever built a bridge amp that uses a balanced differential driver and precision resistors? Would there be a disadvantage to this, other than complexity and the expense of 0.1% resistors?
So, a 22+22 transformer instead of a 25+25 transformer?
You could sharpen up the soldering iron to a really fine point and try the OP275 bridge adapter for hi-fi. It is possible to spread the pins a bit so the very tiny thing can be soldered. For DIY, the precision resistors don't cost more money, they just cost more time, with the ohmmeter.
Bridging chip amplifiers has the disadvantage of putting on-chip limiters series with more on-chip limiters.
Can block the limiter by paralleling more:
Parallel amp to 8 ohm speaker, or.
4 chips parallel to 4 ohm speaker.
Now the limiter doesn't work because threshold is higher than output.
Likewise, 5 chips parallel amplifiers can help make the bridged chip amp that doesn't set off a doubled limiter.
Question:
If we build an ordinary TDA7293 with bootstrap, mute and standby and after that, do the input section exactly like the Slave mode configuration (disable the onboard front end) so that we could drive it with any amplifier (and get regulated front end). . . I wonder if the limiter is still engaged?
Can block the limiter by buffering the output:
There's a wide variety of ways to put high end, yet inexpensive, output devices on TDA7294 and TDA7293. Then the chip amplifier doesn't see a high load and doesn't set off its limiter. Before considering that, maybe consider that the LME chip amplifiers are higher resolution.
Or, a more streamlined alternative:
However, with that much effort, you might as well build the Honey Badger discrete amp instead. Since the PCB boards are availabile in the diyaudio.com store, building the huge amp is almost as easy as soldering practice. And it has enough power to X-Max most speakers.
If you can find TDA7294S, with all 15 pins active, just like TDA7293.
The Master/Slave parallel method is easy, low cost and extremely low loss. It is for use with speakers that are very difficult to drive and also for when you need a sturdy amplifier with zero output impedance.
Otherwise, the normal TDA7294, can be be paralleled very similar to LM3886's PA100. This method of paralleling uses matched resistors (via ohmmeter) and sturdy resistors at the output--that is the hi-fi method because output device noise goes into ballast resistors instead of into speakers.
That sounds like an uninspired guess to me. I will show you a simple "cookbook" method of determining optimum supply voltages for your bridge amps.
For illustrative purposes let's say we want to build a bridged 3886 amplifier. A look at the datasheet reveals these specs that are relevant to our discussion.
68W Cont. Avg. Output Power into 4Ω at VCC = ±28V
38W Cont. Avg. Output Power into 8Ω at VCC = ±28V
50W Cont. Avg. Output Power into 8Ω at VCC = ±35V
If we build a bridge amp with Vcc = ± 28 volts, our bridge amp will deliver 136 watts into an 8 ohm load with optimum efficiency and lowest heat generated. Raising Vcc will not raise the output of the amplifier, but it will significantly raise the operating temperature of the chips.
If we build a bridge amp with Vcc = ± 35 volts, we can deliver 100 watts into a 16 ohm load with optimum efficiency.
If we build a parallel-bridge amp (4 chips per channel) with Vcc = ± 28 volts, we can deliver 272 watts into a 4 ohm load with optimum efficiency.
All of the above configurations are subject to the usual caveats and represent absolute best case scenarios. Prudent design would dicatate that we set our goals a little below what these figures suggest. A safe "fudge factor" would be to reduce Vcc by 5-10% which would reduce generated heat but also reduce available power.
It is really up to the designer to juggle all the factors.
-
Last edited:
An 18+18 dual secondaries transformer /w dual schottky bridge rectifiers will probably make that 28+28vdc or very close to it. However, that's the unloaded rails. Yes, that will run very cool with Bridged TDA7294. I'm not sure about the 136 watts.
P.S.
There's actually some advantages to this. You get 6db more without having to increase the gain. And, subjectively speaking, I though the chip sounded a lot better at lower voltage.
P.S.
There's actually some advantages to this. You get 6db more without having to increase the gain. And, subjectively speaking, I though the chip sounded a lot better at lower voltage.
Best case scenario with 3886. I don't know about TDAxxxx.
You're getting it.
A chip amp run on lower voltage will be subject to less instantaneous thermal modulation. This is not necessarily the only dynamic at work here.
Now try one of those bridge adaptors and see if it sounds better. It is a simple way around a short list of caveats.
Well, almost. But if you buy some real 0.1% resistors, you usually also get a much-better temperature coefficient, so that they will stay within 0.1%, or stay better matched, even when their temperature changes.
Some standard 1% metal film resistors that I just looked up (e.g. Xicon) vary their values by up to 50 ppm (parts per million) PER DEGREE Celsius. Some others are listed as +/-100 ppm.
Some 0.1% 1/8th-Watt metal film resistors that I just looked up (e.g. the Vishay/Dale PTF series) vary their values by 5 ppm or 10 ppm (depends on which particular value) per degree C, i.e. 1/10th or 1/5th as much. Their cost in my 2006 Mouser.com paper catalog was $0.81 for qty 1 of the 10 ppm values and $1.30 for qty 1 of the 5 ppm values.
The 0.1% ones also tout their very low noise and very good high frequency characteristics, and they note that they are epoxy-sealed for superior moisture protection.
You don't have to choose - why not both? At least that's my ideal solution.
You did omit some important advantages of bridged which are related to the power supply. Running bridged is akin to full-wave rectification in terms of the supply caps, it doubles the frequency of the supply ripple for a given output frequency. In my experience, power supply noise has most often been the limiting factor in chipamp SQ, not THD caused by heavier loading of the output stage .
@Fast - why do you want to go to 0.1% resistors? They're needed for gain setting when paralleling amps or they'll fight one another, but I can't see another application...
Last edited:
The only place that I've ever had trouble with a resistor is metal film at feedback-shunt. And, I won't ever do that again. Conversely, I find that the metal film resistors work delightfully for input load. I wonder why there's a difference. Question: Are metal film resistors slightly capacitive?
I think that running the power supply as low as possible by cutting the available recovery time to half is not promotable as an advantage to audio quality. However, it may be more important that a borderline size supply, running out of charge with an 8 ohm speaker and then clipping early, does not have that problem with either a 4 ohm speaker or a bridge amp with an 8 ohm speaker. Of course there's a lot of other factors, but when it comes to clipping, the bridge amp does a lot less clipping. For sure, that practicality easily swamps most of the other factors.abraxalito said:
Miller amp with LTP throws the baby out with the bath water so that power noise rejection is an excellent feature we hope never to use.abraxalito said:
Last edited:
good explanation. thx
I think it's similar with tda7294, from data sheets:
Vcc = 35 Volt Po=70Watt into 4Ω
Vcc = 27 Volt Po=70Watt into 4Ω
If using an 18+18vac transformer, which is totally advisable for bridged, output falls to about 25+25vdc or less when loaded well, such as with a bridge amp.
You're looking for a 20+20vac transformer, if you actually want your 27vdc or 28vdc while the bridge amp is running.
Currently, I'm looking for ways to make the amp clear enough to tolerate bridging, and I think that could be done by cutting off the crosstalk while preventing charge loss at small signal power. We probably need something more effective than bell wire, and I think that the replacement will look worse and sound much better. Meanwhile, back to the soldering iron. . .
You're looking for a 20+20vac transformer, if you actually want your 27vdc or 28vdc while the bridge amp is running.
Currently, I'm looking for ways to make the amp clear enough to tolerate bridging, and I think that could be done by cutting off the crosstalk while preventing charge loss at small signal power. We probably need something more effective than bell wire, and I think that the replacement will look worse and sound much better. Meanwhile, back to the soldering iron. . .
i have read nearly four pages and quite impressed by bob and daniel
I am gonna build three amplifiers using ST tda7294. I have also got couple of la3161 ics- might be required for preamp.
Pls provide me in a nutshell summary of guidelines for building it for
1. tuning it solely for low frequency response (for woofer)
2. tuning it for a wide range frequency response (i will then attach a crossover)
Also mention if i could use 18-0-0-18 transformer i have from my old STK amp. I highly doubt that it could fit worthy anyway in 1. or 2. configs.
I desire an output of at least 50W across my 8ohm woofers
As i m not much deep in these amp-things so it wud b better if an OPTIMISED SCHEMATIC.
I am gonna build three amplifiers using ST tda7294. I have also got couple of la3161 ics- might be required for preamp.
Pls provide me in a nutshell summary of guidelines for building it for
1. tuning it solely for low frequency response (for woofer)
2. tuning it for a wide range frequency response (i will then attach a crossover)
Also mention if i could use 18-0-0-18 transformer i have from my old STK amp. I highly doubt that it could fit worthy anyway in 1. or 2. configs.
I desire an output of at least 50W across my 8ohm woofers
As i m not much deep in these amp-things so it wud b better if an OPTIMISED SCHEMATIC.
The BIG effort, solder every one of the feedback+feedback shunt values and compare.
There's no bode plot in the datasheet, but a practical substitute is available: Leave no stone unturned. If you want it to do something, make it do that. 1930's style engineering, full steam ahead!
So, there's 3 grades, Fail, Borderline and Valid, plus audio notes, and a * for settings with merit enough to be worth re-check.
The described behaviors vary by operating voltage--I'm using 35+35vdc, full power of course. I'm using the power supply from post 116, post 117, the MBR schottky are plugged into V+ and V- at the amplifier board and the respective DC cables are connected to the MBR schottky (the full implementation). There's simple 220u per rail on the amp. No rail2rail cap, no buffer, no blarebuster, no lightspeed, no EQ, no potentiometer, no preamp, no accessories, no distractions for these tests.
So here goes with the soldering and taking notes while testing and soldering. . .
Format is:
Feedback resistor vs Feedback-shunt resistor, practical benefit
First group /w 440u nfb-shunt cap:
100k vs 2k7* Valid, beautiful mids and treble, insufficient bass, high res, no clip
100k vs 2k2, Fail, coarse/clinical treble, gain capacity exceeded
75k vs 2k7, Fail, muffled treble
75k vs 2k2, Fail, strong metallic tone, acutance, adeq res, higher temps, no clip
68k vs 2k7, Fail, muffled treble
68k vs 2k2, Borderline, bright mids not match bari, good bass, good img, no clip
68k vs 2k0, Fail, muffled treble
56k vs 2k2* Valid, near level sound, good bass, good img, high res, no clip
56k vs 1k8, Valid, near level sound, big bass, adeq img, adeq res, choked@75%
47k vs 1k8, Borderline, metallic tone, good bass, adeq res, big img, no clip
47k vs 1k5, Valid, near level sound, big bass, adeq res, big img, choked@85%
39k vs 1k5* Valid, near level sound, adeq bass, high res, big img, no clip
Second group /w 660u nfb-shunt cap:
(to prevent LTP damage/thumps, antiparallel diodes are added to NFB-shunt cap)
47k vs 1k5, Valid, near level sound, good bass, high res, good img, choked@95%
39k vs 1k5, Valid, near level sound, big bass, adeq res, good img, choked@95%
39k vs 1k2* Valid, tolerably forward, loud bass, good res, big img, no clip
47k vs 1k2* Valid, level sound, big bass, high res, big img, armstrong power, no clip
Full stop! Document! Panic! Research! Whiskey Tango Foxtrot???
That description is not enough, but there is this:
A bit stunned at the moment. My apologies; due to the huge surprise of TDA7294S// at 47k vs 1k2 in combination with a full implementation of the power circuit, I cannot concentrate to make more comparisons at this time--too distracted. The remainder of the second group and all of the third group are scheduled for comparison later. Five * items need re-checked with added input fine tuning support, later. On emergency basis, beer and concert sound have temporarily preempted the soldering. This amp, I desire to use. I'm a bit humbled by having succeeded and not know why. I didn't know it could do this much, but of course doing better will be fun.
A perfect day. I love that.
There's no bode plot in the datasheet, but a practical substitute is available: Leave no stone unturned. If you want it to do something, make it do that. 1930's style engineering, full steam ahead!
So, there's 3 grades, Fail, Borderline and Valid, plus audio notes, and a * for settings with merit enough to be worth re-check.
The described behaviors vary by operating voltage--I'm using 35+35vdc, full power of course. I'm using the power supply from post 116, post 117, the MBR schottky are plugged into V+ and V- at the amplifier board and the respective DC cables are connected to the MBR schottky (the full implementation). There's simple 220u per rail on the amp. No rail2rail cap, no buffer, no blarebuster, no lightspeed, no EQ, no potentiometer, no preamp, no accessories, no distractions for these tests.
So here goes with the soldering and taking notes while testing and soldering. . .
Format is:
Feedback resistor vs Feedback-shunt resistor, practical benefit
First group /w 440u nfb-shunt cap:
100k vs 2k7* Valid, beautiful mids and treble, insufficient bass, high res, no clip
100k vs 2k2, Fail, coarse/clinical treble, gain capacity exceeded
75k vs 2k7, Fail, muffled treble
75k vs 2k2, Fail, strong metallic tone, acutance, adeq res, higher temps, no clip
68k vs 2k7, Fail, muffled treble
68k vs 2k2, Borderline, bright mids not match bari, good bass, good img, no clip
68k vs 2k0, Fail, muffled treble
56k vs 2k2* Valid, near level sound, good bass, good img, high res, no clip
56k vs 1k8, Valid, near level sound, big bass, adeq img, adeq res, choked@75%
47k vs 1k8, Borderline, metallic tone, good bass, adeq res, big img, no clip
47k vs 1k5, Valid, near level sound, big bass, adeq res, big img, choked@85%
39k vs 1k5* Valid, near level sound, adeq bass, high res, big img, no clip
Second group /w 660u nfb-shunt cap:
(to prevent LTP damage/thumps, antiparallel diodes are added to NFB-shunt cap)
47k vs 1k5, Valid, near level sound, good bass, high res, good img, choked@95%
39k vs 1k5, Valid, near level sound, big bass, adeq res, good img, choked@95%
39k vs 1k2* Valid, tolerably forward, loud bass, good res, big img, no clip
47k vs 1k2* Valid, level sound, big bass, high res, big img, armstrong power, no clip
Full stop! Document! Panic! Research! Whiskey Tango Foxtrot???
That description is not enough, but there is this:
An externally hosted image should be here but it was not working when we last tested it.
A bit stunned at the moment. My apologies; due to the huge surprise of TDA7294S// at 47k vs 1k2 in combination with a full implementation of the power circuit, I cannot concentrate to make more comparisons at this time--too distracted. The remainder of the second group and all of the third group are scheduled for comparison later. Five * items need re-checked with added input fine tuning support, later. On emergency basis, beer and concert sound have temporarily preempted the soldering. This amp, I desire to use. I'm a bit humbled by having succeeded and not know why. I didn't know it could do this much, but of course doing better will be fun.
A perfect day. I love that.
Last edited:
thanx for quick review
could you please advice me on input load and use of preamps LA3161 for both configs 1. low freq. 2. wide range freq.(will use crossover)
As i said i need at least 50W at 8ohm.
Yeah, they all have a voltage coefficient, too. The precision ones I was looking at had a significantly lower voltage coefficient than the non-precision ones.
The temperature coefficient will also cause AC distortion, especially if the power rating of the resistor is not a lot higher than the average power it will actually dissipate with AC. The AC current can actually modulate the resistance, by varying the temperature. To avoid that, some designers use oversized resistors.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.