I know what you mean, and that is true. It is just not that kind of bass. Somebody else may express it as "unnatural" bass, or may be "clinical" bass, or "cold" bass...
That is why some of us avoid chip amps, even opamps... which is a subject for hot debate
Jay, I'm always open to trying something new. What do you like that produces the kind of low end response you like?
At the risk of being repetitive and or argumentative or pushy, have you heard or built any of the LM3886 designs like those from Chipamp.com or the MyRef series? They have come a long way since the "Gainclone" years.
At the risk of being repetitive and or argumentative or pushy, have you heard or built any of the LM3886 designs like those from Chipamp.com or the MyRef series? They have come a long way since the "Gainclone" years.
I don't know. I just know it when I hear it. Chips usually sound cold, clinical, unnatural. If you look at their specs, you can see that they have no problem in having very good spec at HF. But rarely have good spec at LF. May be that explains the bass I found inferior than the ones I found in some discrete designs.
I have built many times. But the Fremen one I have not.
I would lean closer to Andrews view as I have all the low end I can use. Careful cap and resistor selection has greatly enhanced the upper registers. IMHO, one of the benefits of the entire MyRef series is that the design allows numerous options for tuning to match associated equipment and personal tastes. Some builders may not want to go through that process.
Again, I'm just reporting what I like as the thread title suggests. Since it's been thirty years since I owned a tube amp, I will probably build one in the coming year for the enjoyment of building it and for comparison. I have several good "transistor" power amps but they have all been placed on the shelf since I entered the LM3886 world. The Pass F series has a lot of devotees and has always attracted my eye. Once the Freemem is finalized - who knows???
P.S. linuxguru is close to releasing a new two chip MyRef design that should address several LM3886 issues.
Again, I'm just reporting what I like as the thread title suggests. Since it's been thirty years since I owned a tube amp, I will probably build one in the coming year for the enjoyment of building it and for comparison. I have several good "transistor" power amps but they have all been placed on the shelf since I entered the LM3886 world. The Pass F series has a lot of devotees and has always attracted my eye. Once the Freemem is finalized - who knows???
P.S. linuxguru is close to releasing a new two chip MyRef design that should address several LM3886 issues.
Last edited:
Yeah, I think you're right. But what is the different with discrete of similar topology? In case of the overture series, I think I haven't heard any discrete quasi amp that is better than the chip?
Bass
The lack of low bass, in my experience, can come from several locations:
Generally applicable errors:
Undersize NFB cap
Undersize Transformer (why use 1a transformer for 4a chip?)
VA calc errors--Please check transformer Amperage figure before purchase.
Most common error: Datasheet schematics have undersize NFB cap
Conditionally applicable errors:
Insufficient input load resistor (good range 5k to 12k)
Insufficient feedback resistor (good range 10k to 82k)
Too much feedback-shunt (good range 1k to 2.7k)
Most common error: K50 kit with inappropriate 180k feedback resistor sounds clinical, poor bass, and with shortchanged soundfield size.
Errors that are fixed by paralleling the chip amp:
Crossing/sticking noises at output devices
Overdamped drive shackles the woofer
Really strained miniature output devices
Not effortless sound
Good reference: John Linsley Hood, The Art of Linear Electronics When paralleling, it is unlikely that any two or three chips make the same error/noise at precisely the same moment in time--differences go into ballast resistors, not speakers. Yay!!! See also Jeff Rowland.
The sound of an expensive discrete amp? That is the sound of current headroom. Parallel chip amplifiers can do it too.
Quality control?
What would you do if the tiny and crappy BD911 was the only output device available? Sure, you'd parallel it. Now go look at a chip amp and see that there's TWO output devices crammed in there. Oh much worse! But, the fix is the same.
LM1875 doesn't have a fast attack limiter aboard, and therefore, the chip has no means to cause the error described. Therefore, I claim that the bass error isn't inside the chip. Let's search for the problem.
The lack of low bass, in my experience, can come from several locations:
Generally applicable errors:
Undersize NFB cap
Undersize Transformer (why use 1a transformer for 4a chip?)
VA calc errors--Please check transformer Amperage figure before purchase.
Most common error: Datasheet schematics have undersize NFB cap
Conditionally applicable errors:
Insufficient input load resistor (good range 5k to 12k)
Insufficient feedback resistor (good range 10k to 82k)
Too much feedback-shunt (good range 1k to 2.7k)
Most common error: K50 kit with inappropriate 180k feedback resistor sounds clinical, poor bass, and with shortchanged soundfield size.
Errors that are fixed by paralleling the chip amp:
Crossing/sticking noises at output devices
Overdamped drive shackles the woofer
Really strained miniature output devices
Not effortless sound
Good reference: John Linsley Hood, The Art of Linear Electronics When paralleling, it is unlikely that any two or three chips make the same error/noise at precisely the same moment in time--differences go into ballast resistors, not speakers. Yay!!! See also Jeff Rowland.
The sound of an expensive discrete amp? That is the sound of current headroom. Parallel chip amplifiers can do it too.
Quality control?
What would you do if the tiny and crappy BD911 was the only output device available? Sure, you'd parallel it. Now go look at a chip amp and see that there's TWO output devices crammed in there. Oh much worse! But, the fix is the same.
Why not simply dial it in?
The solo chip amp is highly sensitive to current, especially at feedback, so here's some dials for you.
Note: Variable resistors made from multi-turn trimpots--solder 2 out of 3 pins together before use, and then you have a 2-pin variable resistor--check with ohmmeter before use.
Setup:
(LM1875 non-inverting amp--see the datasheet)
♦ The input load resistor is a simple 10k.
♦ The input cap is 2.2u (like Nichicon ES 2.2u and 4.7n polyester bypass)
♦ The NFB cap is 470u (consider 1u or smaller electrolytic for bypass, or a 22n or smaller polyester dip cap for bypass, but don't use polypro).
♦ The 2 amp board power caps are also 470u, and they can be exactly the same model and same bypass efforts that you find attractive for NFB cap.
♦ The feedback resistor is 10k series to 250k and Parallel that 250k with a 100K multi turn trimmer (creating a 72k variable resistor)--total Range of variance is then 10k to 82k.
♦ The feedback-shunt resistor is a 1k series to 2.2k and Parallel that 2.2k with a 10k multi turn trimmer (creating a 1.8k variable resistor)--total Range of variance is then 1k to 2.8k.
Adjust:
Although the trimmers have been restricted to decent values, please be careful to avoid setting the gain outside the range of 10x to 40x. You'll need to use your ohmmeter between settings. Divide the feedback and feedback-shunt and then add 1 to find the gain setting. These two dials vary both gain and feedback current, and thus a vast variety of differences is possible, and the choice is yours.
Reason:
Turning a dial could be easier than hours of soldering different resistor values. After dialing in optimal resistor values, we could then replace trimmer arrangement with nearest value ordinary resistor instead.
Alternatives:
Dialing in feedback settings is not the only way to fine tune LM1875. External compensations can be added for that job. That is doable, but I couldn't possibly explain it. In this case textbooks can be really helpful. Much textbook material for small signal op-amps can be applied to LM1875 since it is a simple op-amp.
The solo chip amp is highly sensitive to current, especially at feedback, so here's some dials for you.
Note: Variable resistors made from multi-turn trimpots--solder 2 out of 3 pins together before use, and then you have a 2-pin variable resistor--check with ohmmeter before use.
Setup:
(LM1875 non-inverting amp--see the datasheet)
♦ The input load resistor is a simple 10k.
♦ The input cap is 2.2u (like Nichicon ES 2.2u and 4.7n polyester bypass)
♦ The NFB cap is 470u (consider 1u or smaller electrolytic for bypass, or a 22n or smaller polyester dip cap for bypass, but don't use polypro).
♦ The 2 amp board power caps are also 470u, and they can be exactly the same model and same bypass efforts that you find attractive for NFB cap.
♦ The feedback resistor is 10k series to 250k and Parallel that 250k with a 100K multi turn trimmer (creating a 72k variable resistor)--total Range of variance is then 10k to 82k.
♦ The feedback-shunt resistor is a 1k series to 2.2k and Parallel that 2.2k with a 10k multi turn trimmer (creating a 1.8k variable resistor)--total Range of variance is then 1k to 2.8k.
Adjust:
Although the trimmers have been restricted to decent values, please be careful to avoid setting the gain outside the range of 10x to 40x. You'll need to use your ohmmeter between settings. Divide the feedback and feedback-shunt and then add 1 to find the gain setting. These two dials vary both gain and feedback current, and thus a vast variety of differences is possible, and the choice is yours.
Reason:
Turning a dial could be easier than hours of soldering different resistor values. After dialing in optimal resistor values, we could then replace trimmer arrangement with nearest value ordinary resistor instead.
Alternatives:
Dialing in feedback settings is not the only way to fine tune LM1875. External compensations can be added for that job. That is doable, but I couldn't possibly explain it. In this case textbooks can be really helpful. Much textbook material for small signal op-amps can be applied to LM1875 since it is a simple op-amp.
I've seen dip switches on paralleled resistors to achieve the same. 16 available values on a 4 line dip switch/4 resistor. Very easy to then replace switch with appropriate jumpers when good values are found.
Thanks! That is really great! Got an example? Maybe a schematic?
That was badly phrased. My apologies! I meant to say that the "power caps of the amplifier board" could be the same 470u value and same high quality as the cap you've selected for the NFB cap.
All of the caps that are located on the amplifier board are important.
If there's no other means to test the quality of caps, you can install a sample as your NFB cap and literally eavesdrop on the quality.
Last edited:
Everything is just relative. Good or bad, that's relative. I have been searching for the best amp for so many years. Isn't it very silly if everything is actually available in a cheap LM1875? I must be deaf or having a hallucination...
LM1875 is not a new toy for me. I don't expect it to sound like a high end amp. It is not even the chip I have achieved the best system with.
But the real problem with my current implementation is probably the input opamp. I have achieved several outstanding chip amp systems, and all are with pure class-A front ends. I choose opamp here so it can become a "benchmark reference". May be I should use B1, or DOZ preamp?
My current opamp input stage uses dual-opamp socket (2xopamps in one package). I don't have enough inventory for this. Last night I tried several opamps form 0.3 to 15V/us, and I have to accept that (in that circuit) the best sound was given by an Analog Device opamp with 1.8V/us.
While Sanyo Oscon on the feedback path gave cleaner sound, the bass sounds "colder" and I don't like.
Just almost. LM1875 does have a failing, but the workaround isn't difficult. The soundfield is as deep as a hi-fi should do, but curiously not wide. The fix is to build monoblocs for their greater stereo separation. Also, if you desire a more natural tone or even a dark voice tone, the parallel LM1875 can do that.
Well, they're quite good and indeed the quick trip to hi-fi; however. . .
Actually, my favorite amplifiers are the wild dynamics Circlophone and also my old TDA7294 (with rather excessive fine tuning).
You've already got an op-amp. Why stack another for quality reduction? The LM1875's gain capacity goes high enough to clip itself even if driven by an MP3 player. Sufficient gain capacity is one of the conveniences of the 25 watt amplifier, and there's no need of a preamp to increase the clipping.
However, if you've got the unavoidable urge to stack another op-amp in front of the power op-amp, try nesting/composite topology to negate insertion loss. Instead of reducing quality, nesting/composite topology can make improvements. ah! That also includes more control over the sound of your power op-amp.
P.S.
Other answer: DOZ is close. I like the very similar MooseFet triode simulator preamp since it is very clear and so easily adaptable to suit a vast variety of personal preferences.
Yeah, a cheap trick. My resistor is even more expensive than the chips
There has been many blind tests around since Gainclone popularity some years ago. The winners have been always the overture series. I wish I could make a TDA7294 to sound good. Because I like the mosfet sound of it.
I just did that again last night, but I never perceived/noticed special changes to the voice tone. It is just more dynamic.
I thought chip amp guys might like tone control, no?
Sufficient gain capacity is one of the conveniences of the 25 watt amplifier, and there's no need of a preamp to increase the clipping.
No I don't think a chip by itself, especially small ones, can sound good enough for me without a driving preamp (low gain or buffer). Especially considering the use of potentiometer for volume control.
I have an (useless) idea to stack at least 10 very low noise opamps (OPA111) in front of a very high slew rate opamp (at least 1000V/us) for a preamp. May be I can put them in front of another power opamp
I have experienced twice a good system using chip amps. It is of course very dependent on other components such as the speaker.
First one, is a TEA chip, no more than 2W, driven by a preamp similar to the MooseFet (but better of course). At that time, I have tube amp and several Alephs (So you know that the benchmark is not that "low").
Second one, TDA2009 chip, no more than 3W for 0.1% THD, driven by a preamp similar to DOZ.
Another one was not really a "success", LM3875 with triode preamp, with power supply full of chokes. Inspired by JLTi I guess.
But chip amps are just chip amps. I do this just for fun because I have the parts scattered all around the house.
the chipamps do not need a buffer.
The non-inverting topology allows virtually any Zin value to be adopted and there is absolutely no value in increasing this further.
It's the source that may need a buffer.
The source must be capable of driving the cable and the Rin that follows. That Rin can and should include RF filtering and should also include a DC blocking capacitor.
If you have a vol pot driving a cable expect there to be problems. vol pots cannot properly drive cables.
Sort your source and remove the buffers from the front end of the chipamp.
There is one exception. Inverting topology Chipamps. These can have a lowish Rin. Few sources can drive these directly. Here a "converter" is a real performance benefit. See Cordell's jFET LTP "converter".
Well, do it.
TDA7294 non-inverting amp:
Pair 220u amplifier board power caps and a single 2u polyester rail to rail.
47u (or 68u) for bootstrap
10k for input load resistor
10k for mute (no cap)
22k with 10u timer for standby
60k (120k||120k) for feedback resistor
2.7k for feedback-shunt resistor
220u (or more) for NFB cap
1u (or less!) for input
Output zobel is 10n with ~8R
Add RF filtering to input cap
Use a stiff power supply (20,000u per rail, 3a per chip)
Use less voltage than max (under-volt)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.