I briefly compared the following chip amps, TDA7294 and LM3875 against a tube amp ASL AQ-1003 With EL34 output tubes ( push pull ,class AB ).
The tube amp sounds better overall. However the difference is not huge though very audible. The TDA7294 comes closest . Wonder what will happen if I remove the dc decoupling cap in the feedback loop. Maybe worth trying. While the LM3875 sounds generally nice on it's own , when you throw complex sounds at it it gets noisy....you want to turn the volume down ! That didn't happen with the tube amp and also the 7294 amp. Both chip amps have pcb's bought off the Net. They also run without input capacitors though both have dc decoupling caps in the negative feedback loop. That is one thing that needs to be removed and checked. Will have to check how much the offset rises. I'm not too keen on using a dc servo. All amps were playing below 10 watts peak on music. Must have been a watt average !
I used a Chinese saxophone music disc which can really sound nasty at times with some amps ! I don't know what the disc says as all of it is written in Chinese. I got it in Hong Kong a few years ago. Good disc to check how 'shouty' the system can get !
Signal was direct from source ( Bluray player ) to Alps volume pot to amp. Looks like the LM3875 will now go in for ( low power ) 'sub' use !
The decoupled input stage on a TDA 7294 is supposedly even better than the stock circuit though it needs protection to prevent it from being killed on a power up/down cycle. I haven't done this yet. Need a new board to do that. No one offers such a board so I'll have to make one myself. No time right now !
The tube amp sounds better overall. However the difference is not huge though very audible. The TDA7294 comes closest . Wonder what will happen if I remove the dc decoupling cap in the feedback loop. Maybe worth trying. While the LM3875 sounds generally nice on it's own , when you throw complex sounds at it it gets noisy....you want to turn the volume down ! That didn't happen with the tube amp and also the 7294 amp. Both chip amps have pcb's bought off the Net. They also run without input capacitors though both have dc decoupling caps in the negative feedback loop. That is one thing that needs to be removed and checked. Will have to check how much the offset rises. I'm not too keen on using a dc servo. All amps were playing below 10 watts peak on music. Must have been a watt average !
I used a Chinese saxophone music disc which can really sound nasty at times with some amps ! I don't know what the disc says as all of it is written in Chinese. I got it in Hong Kong a few years ago. Good disc to check how 'shouty' the system can get !
Signal was direct from source ( Bluray player ) to Alps volume pot to amp. Looks like the LM3875 will now go in for ( low power ) 'sub' use !
The decoupled input stage on a TDA 7294 is supposedly even better than the stock circuit though it needs protection to prevent it from being killed on a power up/down cycle. I haven't done this yet. Need a new board to do that. No one offers such a board so I'll have to make one myself. No time right now !
Last edited:
The killer with the TDA7294 is that the substrate is connected to the signal -ve pin, not the output stage -ve pin. So this means if you RC decouple the signal -ve pin from the power -ve pin, it's no longer the most negative potential in the circuit. Result is a quickly destroyed chip.
If the TDA7294 is bridged then there's no need to decouple the -ve signal rail, just run from a single supply with decoupling only on the +ve side. Or use the dreaded output capacitor in series with the speaker and take the speaker return from the -ve supply rail.
If the TDA7294 is bridged then there's no need to decouple the -ve signal rail, just run from a single supply with decoupling only on the +ve side. Or use the dreaded output capacitor in series with the speaker and take the speaker return from the -ve supply rail.
A few years ago Ilimzn had explained the decoupling problem. Here there was a zener and protection diode at the two supply pins ( +in/+out and -in/-out )to keep the substrate within a safe voltage level . Did anyone try that ?
I wanted to ,but things just kept getting delayed. It should work .Ilimzin knows his fundas quite well !
Look it up.
http://www.diyaudio.com/forums/chip-amps/3857-tda7293v-4.html
I wanted to ,but things just kept getting delayed. It should work .Ilimzin knows his fundas quite well !
Look it up.
http://www.diyaudio.com/forums/chip-amps/3857-tda7293v-4.html
I suspect if you bandlimit all the amplifiers to exactly the same frequencies and you do not overload any of them during the comparison, you will find they all sound almost identical.
It's their behaviour in overload that shows up sonic differences.
It's the different frequency response that "colours" the sound.
It's their behaviour in overload that shows up sonic differences.
It's the different frequency response that "colours" the sound.
Building a basic inverted Gainclone chip amp using hard-wiring Try that particular LM3875.
Of course they are bound to be great, you are making them.
I don't sell them and will say the Folsom amp sounds great.....
amps can sound different for understood and tweakable reasons - the biggest being frequency response, including amp output Z interaction with speaker load
The Carver Challenge | Stereophile.com
The Carver Challenge | Stereophile.com
I did a more simplified approach that is able to challenge a tube amp:A few years ago Ilimzn had explained the decoupling problem. Here there was a zener and protection diode at the two supply pins ( +in/+out and -in/-out )to keep the substrate within a safe voltage level . Did anyone try that ?
I wanted to ,but things just kept getting delayed. It should work .Ilimzin knows his fundas quite well !
Look it up.
http://www.diyaudio.com/forums/chip-amps/3857-tda7293v-4.html
That annotation is still not quite on. So, it is a very good thing that the trimmer is present so that you can fine tune as you like. Truly, it will run best, if undervolted as shown. And, actually, it is possible to substitute that 25k feedback resistor for 27k, if you like. Just comp with the trimmer afterwards. 6A01 and 10a01 diodes also work fine. It is possible to refit that schematic onto a pre-made board. And, the very smallest boards tend to work best so that the decoupling isn't too far away from the chip.
For a "dial-a-shout" feature, replace the 25k input resistor with a 50k trimmer (configured as a variable resistor, applicable range 18k~27k), and adjust as you like for midrange tone.
Last edited:
That is at post #5.
It is not convenient. It does work very well though.
My suggestion and schematic succeeds in crossing the bridge between those needs. That was not done without costs; however, they are not difficult to manage and optimal application of either certainly does take more careful reading and a lot more effort than the datasheet suggests--the text is great but the schematics have no known use. For sure, Ilimzn's approach will do slightly higher resolution than mine. And for sure, mine is more doable and does, subjectively *almost* as well. Choose for need. It is great to have two good options. Willing to make a lot of custom power stuff for higher quality, then choose Ilimzn's; or not willing, then choose mine. You got a win either way. Neither have any waste. They both do audio entertainment relevantly. That is not commonplace.
You may want to try out any of the nested chipamp circuits using the LM1875 or LM3886, with an outer loop having a high-performance opamp. Examples include Mauro Penasa's MyRef Rev C and derivatives, the Modulus-86, and my own MiniRef-1875 v1.04.
Don't just take my word for it - try one out and you'll find that the audible sonics do indeed sound much better than stock gainclone/chipamp designs. The main limitation is in output power (~80W with a dual LM3886 configuration), but that's not an issue for typical indoor audiophile listening with efficient speakers - 1W should be sufficient.
My efforts now are concentrated on making the outer small-signal opamp as refined as possible, usually with the use of Class-A biasing and/or a Class-A buffer.
Yes, it is a matter of implementation. They don't use the same power supply, how can they sound similar!?
I don't compare amps like this. If I want to compare 3875 with 7294, first I will find the best implementation for 3875 then the best implementation for 7294, then I compare...
There is optimum supply voltage for each chip (in term of THD). But too often the different in sound is caused by power supply. Change the amount of capacitance, change the capacitor brand, change the diodes (bridge), all will bring very audible differences.
Diodes is the one often overlooked. The best ones are usually the discrete axial ones. Most packaged diode bridge sound terrible.
BUT here we are comparing mosfet output versus bipolar output. They have different characteristics that will remain different when built into amplifier circuit. They will sound different, and that affect listeners preference or taste.
So far I prefer the bipolar characteristics of 3875. Mosfet characteristics as in 7294 is very "impressive" (replace that with many audiophile words of your choice) but I prefer to judge my preference based on my enjoyment when using the amp to listen to music all day for several weeks, months or years.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.