This is the kit I bought which contains all the mounting stuff. I'm sure they will sell a single if asked.
TDA7294 monoBTL/stereo 100W amplifier kit two separarate boards !
TDA7294 monoBTL/stereo 100W amplifier kit two separarate boards !
Daniel,
I did, in fact pin4 becomes a secondary lower gain and lower quality non-inverting input in the mute state. It should tie to GND, it doesn't matter much where provided it's not polluted with supply ripple.
As for the supply section questions, please wait and see my upcoming post on an alternate approach to local decoupling.
Sergio, Sorry but still not understanding your question. Here are two choices.
1. Q. Where is the PBC that is the base of the kits discussed on the thread?
A. One of the members is currently developing a group buy at a great price. That hasn't started yet.
A. You can contact any of the eBay vendors listed in post #140 and request a bare board. Other members might have a source if they know what you are asking for.
2. Q. Where is the PCB for the advanced design being discussed?
A. I doesn't exist yet.
Please let us know which you are asking about.
Alternate Local Decoupling.
Instead of using two caps from either rail to GND one can also use two caps differently, one rail-to-rail (C+ below) and one from the neg rail to GND (C-) -- this also has been proposed by other DIYA members in the past. The common point of the caps is the negative supply, and all three feeds to the supply sport series elements (Z1 etc) shown as resistors here for simplicity.
Let's have a look of the current loop for a negative output current in the load :
Shown by the red loop, current flows into the output pin of the opamp and it leaves the chip at the neg supply pin. Then it returns back to the output though cap C-, GND attach and the load in counter-clockwise direction, the cap acting as a level shifter.
The green traces show current paths when the local caps become ineffective at LF down to DC, the slim green trace sees no DC, though.
Key points are that the caps (which are composite/arrayed caps) are close to chip (to benefit from their low ESR and ESL) and that the loops have low area for low inductance to neither emit nor reveice magnetic fields. Finally, nodes are located such that any I*R drop along the wires/traces have no consequences.
Assumed that Z1 can be made rather large (as we will see) as well as C-, there is almost signal current alone and very little AC supply recharge/balancing current. The rail-to-rail cap C+ is out of the picture mostly.
Now going to the postive load current :
The current leaves the chip via the output pin along the red path, goes through the load back to GND. From there the return path to the V+ supply pin is first through C- in opposite shifting level to V-, clockwise direction and then through C+, level-shifting voltage again to V+.
Again, the green paths show current at LF and LF/DC (thick green only).
The important thing now is the notion that cap C-, the GND-to-V- cap, is always in the loop, plain series to the load and will effectively see no balancing current to/from the supply at MF/HF. This means the V- pin (subtrate) is tightly bonded to GND and any variation of its voltage is small and basically linear, fully correlated to the signal output current (but with a different freq response). This leads to the lowest possible impact of supply rejection issues of the chip including nonlinear distortion.
C+, the rail-to-rail decoupler, feels only the half-rectified positive load current and therefore is a most delicate point where lowest loop area is paramount.
Finally looking at the feedback/input scheme :
Other than bias currents and some capacitive stray currents no currents flow in the opamp input pins IN+ and IN-. The arrangement is inverting, Rf/Ri determining the gain. Ci block DC (optional), Cf, Rng, Cng provide means for stability. Rofs minimizes input bias current offsets and Cofs shorts it for low noise. Rz and Cz is the output snubber.
Blue is the input current from the source and gets routed to the ouput pin while the signal current in the yellow GND line is basically zero until it meets star GND from where the return current finds its path back to the source. The yellow GND plane (or better, a shielded cage) must find a short and low inductance path back to the V- pin (and GND). This is not fully reflected in the drawing, also the exposed node area of the IN- network should be physically smaller than shown.
When several amp get connected (on the same PCB or elsewhere), a solid GND buss bar is going though the big yellow trace left of the star GND, and supplies lines would be fed with their own series elements, at least in parts of.
As some may have noted, the chip's frontend positive supply has its own, higher voltage supply but thats not strictly necessary, resp. not an option for chip which don't have seperate supplies.
I'm about to breadboard this asap...
Instead of using two caps from either rail to GND one can also use two caps differently, one rail-to-rail (C+ below) and one from the neg rail to GND (C-) -- this also has been proposed by other DIYA members in the past. The common point of the caps is the negative supply, and all three feeds to the supply sport series elements (Z1 etc) shown as resistors here for simplicity.
Let's have a look of the current loop for a negative output current in the load :
Shown by the red loop, current flows into the output pin of the opamp and it leaves the chip at the neg supply pin. Then it returns back to the output though cap C-, GND attach and the load in counter-clockwise direction, the cap acting as a level shifter.
The green traces show current paths when the local caps become ineffective at LF down to DC, the slim green trace sees no DC, though.
Key points are that the caps (which are composite/arrayed caps) are close to chip (to benefit from their low ESR and ESL) and that the loops have low area for low inductance to neither emit nor reveice magnetic fields. Finally, nodes are located such that any I*R drop along the wires/traces have no consequences.
Assumed that Z1 can be made rather large (as we will see) as well as C-, there is almost signal current alone and very little AC supply recharge/balancing current. The rail-to-rail cap C+ is out of the picture mostly.
Now going to the postive load current :
The current leaves the chip via the output pin along the red path, goes through the load back to GND. From there the return path to the V+ supply pin is first through C- in opposite shifting level to V-, clockwise direction and then through C+, level-shifting voltage again to V+.
Again, the green paths show current at LF and LF/DC (thick green only).
The important thing now is the notion that cap C-, the GND-to-V- cap, is always in the loop, plain series to the load and will effectively see no balancing current to/from the supply at MF/HF. This means the V- pin (subtrate) is tightly bonded to GND and any variation of its voltage is small and basically linear, fully correlated to the signal output current (but with a different freq response). This leads to the lowest possible impact of supply rejection issues of the chip including nonlinear distortion.
C+, the rail-to-rail decoupler, feels only the half-rectified positive load current and therefore is a most delicate point where lowest loop area is paramount.
Finally looking at the feedback/input scheme :
Other than bias currents and some capacitive stray currents no currents flow in the opamp input pins IN+ and IN-. The arrangement is inverting, Rf/Ri determining the gain. Ci block DC (optional), Cf, Rng, Cng provide means for stability. Rofs minimizes input bias current offsets and Cofs shorts it for low noise. Rz and Cz is the output snubber.
Blue is the input current from the source and gets routed to the ouput pin while the signal current in the yellow GND line is basically zero until it meets star GND from where the return current finds its path back to the source. The yellow GND plane (or better, a shielded cage) must find a short and low inductance path back to the V- pin (and GND). This is not fully reflected in the drawing, also the exposed node area of the IN- network should be physically smaller than shown.
When several amp get connected (on the same PCB or elsewhere), a solid GND buss bar is going though the big yellow trace left of the star GND, and supplies lines would be fed with their own series elements, at least in parts of.
As some may have noted, the chip's frontend positive supply has its own, higher voltage supply but thats not strictly necessary, resp. not an option for chip which don't have seperate supplies.
I'm about to breadboard this asap...
Attachments
Last edited:
GENIUS!!!
Klaus, I am so extremely happy to finally see someone approach a design with the right priorities and sequence of thinking, and who has the knowledge, skills and abilites to do proper analysis and design! I have been waiting for this for a very long time!
Just from your last few posts I have learned so much that I am giddy with excitement! I can hardly wait to see what comes next!
It's even better, knowing that you now have sophisticated test equipment that will enable getting real answers to questions that usually go unanswered, or result in many conflicting opinions and guesses. This could become one of the first truly-well-designed projects that I have ever seen, here.
Highest regards,
Tom
Klaus, I am so extremely happy to finally see someone approach a design with the right priorities and sequence of thinking, and who has the knowledge, skills and abilites to do proper analysis and design! I have been waiting for this for a very long time!
Just from your last few posts I have learned so much that I am giddy with excitement! I can hardly wait to see what comes next!
It's even better, knowing that you now have sophisticated test equipment that will enable getting real answers to questions that usually go unanswered, or result in many conflicting opinions and guesses. This could become one of the first truly-well-designed projects that I have ever seen, here.
Highest regards,
Tom
Last edited:
Klaus,
Do you have a way to quickly etch and drill PCBs, at home?
If not, you might find it to be a very useful capability. Back when I did a lot of prototyping, I eventually quit using all other methods and just made a PCB every time I wanted to test something. It was much more reliable, and I think that in the long term, it was easier and saved a lot of time.
I could have a board completed in less than an hour, from the time when it was only on my computer screen. So I could build many iterations in one day, if necessary.
I used the "toner tranfer" type of method, usually with two-sided PCBs. I eventually found out how to do it using locally-available supplies, as much as possible, so that I only had to ensure that I had PCB blanks on hand, because I never wanted to have to wait for anything to be shipped to me (and it was much cheaper). The only other real problem was making sure that I had the correct paper to use for the transfer. I found a very good-working one that was carried by the local Staples office-supply store. But they kept changing it, almost yearly.
Regards,
Tom
No. The price I quoted earlier was simply for the PCBs themselves. No chip or other mounting hardware. This was investigative pricing for those interested in taking a stock PCB ($4USD for the pair), and using some selective components. Whether or not there's sufficient interest to pursue at this point is unclear. My interest has changed a little, because if Klaus can come up with a new board relatively inexpensive, and the parts($$) are not to bad, I may go that route. However, I'm still questioning the overall value relative to the sonic gains of a total redesign, over the stock boards which I've already built. I know it is very subjective and will vary from member to member. But I may be willing to give it a go.
Hi Bob,
At the moment I'm using some near field Definitive StudioMonitor 350s with the 7294. In full disclosure, let me also say, that even given my age, I never had any tube gear, so I can't compare that 'sound' to my modest transistor/IC based amps and receivers I've used over the past 40 years. I also have a pair of Reference Connoisseur RC-30s that I intend to hook up and give a listen to also. Just haven't gotten around to it yet. The 350's sound pretty great though.
In all honesty, the sound with my current 7294 build amp has a lot of overall presence - which I like. While over the long haul it can be a bit fatiguing and somewhat harsh to listen to, I find I enjoy the dynamics of it's range over that objection. Music and vocals are crisp and clear across the range with a modest amount of bass from the 350's passive 8" woofer/baffle. I am NOT using a dedicated sub-woofer yet. I like it when I can hear the fingers slide on a guitar, or the vocalist take a breath! So, having said that, I may find it hard to distinguish a sonic signature with the mods that are being discussed here. But that doesn't mean I won't give it try. Nor does it mean there won't be any (improvement). I might (and hope) be pleasantly surprised! I would like to do some serious A/B comparison between the stock amp sound and a redesigned amp. I don't have any equipment that can measure any of it, so I'll just have to trust my ears!
Rick
Bob -
I just had a strange thing happen.... I was listening to my 7294 and source (my Zune) ended. I didn't immediately jump up and turn the amp off. After about 15 minutes I heard a loud pop, and then audible hum coming from the speakers. Didn't smell anything though. I took it around to my bench and opened it up to take a look. Nothing out of the ordinary. Nothing charred, no burnt odor. Haven't turned it back on yet, but will try and trouble shoot later tonight. Check PSU voltages, etc. This was the first time this happened. Odd. Is there some kind of standby mode the amp goes into automatically if it does not receive a signal after a preset time limit? Even so, it shouldn't make a loud pop when doing so!
Rick
I just had a strange thing happen.... I was listening to my 7294 and source (my Zune) ended. I didn't immediately jump up and turn the amp off. After about 15 minutes I heard a loud pop, and then audible hum coming from the speakers. Didn't smell anything though. I took it around to my bench and opened it up to take a look. Nothing out of the ordinary. Nothing charred, no burnt odor. Haven't turned it back on yet, but will try and trouble shoot later tonight. Check PSU voltages, etc. This was the first time this happened. Odd. Is there some kind of standby mode the amp goes into automatically if it does not receive a signal after a preset time limit? Even so, it shouldn't make a loud pop when doing so!
Rick
Tom,
At the moment I don't have PCB capabilities at home but I hope to change that soon, as I perfectly agree to start right away with a real PCB rather than bread-boarding. The more complex the total circuitry has grown (damn scope creep
) the more it pays off. I may do simple CNC-milled PCBs at work, occasionally, and did that before. Or just order, pay and wait for profesional proto boards, it's not that expensive.For this amp, though, I think dead-bugging on copper GND plane and the mentioned capacitor strips plus some p2p wiring will do and is quite quick and easy, too (I hope to get there by tomorrow). Impedance, resistance and loop characteristics are about the same and should transfer well to a real PCB. Time will tell...
Tom,
At the moment I don't have PCB capabilities at home but I hope to change that soon, as I perfectly agree to start right away with a real PCB rather than bread-boarding. The more complex the total circuitry has grown (damn scope creep
For this amp, though, I think dead-bugging on copper GND plane and the mentioned capacitor strips plus some p2p wiring will do and is quite quick and easy, too (I hope to get there by tomorrow). Impedance, resistance and loop characteristics are about the same and should transfer well to a real PCB. Time will tell...
Very good.
But that reminds me of another option: Since not many people make their own PCBs, and since professionally-manufactured PCBs are pretty cheap if a lot of them are ordered at once, then if this project eventually has a PCB made available, through group buys or whatever, why would we not want to use a multilayer board with at least four layers? (Or, maybe that could just be another future version.)
........In all honesty, the sound with my current 7294 build amp has a lot of overall presence - which I like. While over the long haul it can be a bit fatiguing and somewhat harsh to listen to, I find I enjoy the dynamics of it's range over that objection. Music and vocals are crisp and clear across the range with a modest amount of bass from the 350's passive 8" woofer/baffle..........I like it when I can hear the fingers slide on a guitar, or the vocalist take a breath! ............I might (and hope) be pleasantly surprised! I would like to do some serious A/B comparison between the stock amp sound and a redesigned amp. ........... so I'll just have to trust my ears!
Rick
So glad to get a matching subjective review. You called out the exact features that impressed me. That dynamics/presence trade-off is what I've been calling the need to clean up the top end. I have some side sound absorbency panels along both walls and I set the blinds as to not allow reflections. Opposite the speakers (~ 12 ft.) is a long overstuffed couch and as a result experience little or no ear fatigue with the stock build.
The V1.2 MyRef experience was similar but I am totally thrilled at how Dario, linuxgury, Uriah and others brought the entire presentation to a level far above my expectations with their mods and parts selection. It looks like that same level of development is going to take place here.
I built a sub (rather two subs - both sealed and reflex) using a Hi-Vi driver with a 68 pound magnet and a 500W bash sub amp. I haven't even turned it on once the improvements to the MyRefs were completed. I'm sure the "gurus" here will be able to tune in the bottom of the TDA7294 to everyone's satisfaction. It's my personal view that a woofer surface area equaling at least 8" is necessary to produce what high quality chip base amp is fully capable of.
I have never experiences anything like you describe as a standby shutdown. There have been some strange noises, but in my case are due to hiccups in the JRiver player. Maybe what you heard was the Zune going to sleep. Hope you have some junk speakers around for the next power-up.
Thanks again for the review.
Attachments
Last edited:
I think it was mentioned that the MyRef was originally designed with TDA7294 but later found that 3886 was more suitable. I do think that the TDA needs the same approach as MyRef to be improved.
Probably the Zune went off.
But you could check the amplifier's speaker jack for DC, just in case.
P.S.
I always use speaker protection of some sort with my chip amplifiers. Some output caps are most fun for bass extension tuning, but speaker protector kits work too.
Really interesting, similar to AN-202 regarding feed forward decoupling.
A stacked film cap in practice is a parallel of hundreds of plain caps.
I don't know if it were the TDA7294 but for sure from the TDA series.
I'm not sure the TDAs are well suited as current pumps.
I think that might have been what happened, unless it was possibly heat related. But with no signal, I would suspect that the amp would not continue to heat up. I checked the voltage on both channels. 1.5 and 1.6mV respectively. I've never used coupling caps on the outputs of any of the amps I've built other then what's included on the standard schematic. (Should I be?) I've always felt they've 'messed' with the sound a bit.
Dario,
That AN-202 is a classic and my oldest reference to that idea.
TDAs did well in some standard "current pumps" and mixed output impedance amplifiers I designed. Howland type is critical because common mode voltage is high (67% of the output voltage in the MyRefs) and could exceed chip limits, which are unknown for the TDAs other than that the input stage cuts off when pulled to V-. True usable input CM range is something I want to find out in detail sometime, in comparison to the LMs. Both have not specified this parameter in their datasheets so we can only assume a guaranteed range of no more than about 1/10th of the ouput voltage plus some unknown headroom, depending on supply voltages and their balance.
That AN-202 is a classic and my oldest reference to that idea.
TDAs did well in some standard "current pumps" and mixed output impedance amplifiers I designed. Howland type is critical because common mode voltage is high (67% of the output voltage in the MyRefs) and could exceed chip limits, which are unknown for the TDAs other than that the input stage cuts off when pulled to V-. True usable input CM range is something I want to find out in detail sometime, in comparison to the LMs. Both have not specified this parameter in their datasheets so we can only assume a guaranteed range of no more than about 1/10th of the ouput voltage plus some unknown headroom, depending on supply voltages and their balance.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.