as i mentioned i'm keeping the drive level low enough to prevent SPIKE from doing anything. this is accomplished by a soft clipper in the linestage.
but i'm doing this for sound quality reasons not for protection.
i think mark is spot on about reliability.
i've worked quite a lot with the LM3886 and have had absolutely zero failures in genuine (sourced from Mouser) parts.
the only criticism i could imagine leveling at it is the usual delta between the + and - PSRRs.
but i'm doing this for sound quality reasons not for protection.
i think mark is spot on about reliability.
i've worked quite a lot with the LM3886 and have had absolutely zero failures in genuine (sourced from Mouser) parts.
the only criticism i could imagine leveling at it is the usual delta between the + and - PSRRs.
Last edited:
The TDA7293, which is the discussion of this thread, doesn't have the spike system.
However, it does have a current limiter that will almost transparently reduce/omit some bass notes, as needed. That system does not screech like the spike system because ST didn't put a hard clipper into the TDA7293.
The lesser bass with TDA7293's protector, occurs at somewhat past 45 watts; however, by paralleling 2 chips, we've raised the figure to 90W, so it doesn't activate when driving 8 ohm speakers to 70W.
However, it does have a current limiter that will almost transparently reduce/omit some bass notes, as needed. That system does not screech like the spike system because ST didn't put a hard clipper into the TDA7293.
The lesser bass with TDA7293's protector, occurs at somewhat past 45 watts; however, by paralleling 2 chips, we've raised the figure to 90W, so it doesn't activate when driving 8 ohm speakers to 70W.
Last edited:
Yeah, that's alright. Some of the things I've been saying is simply to indicate to new visitors that the project at post#1 is still valid and in use, and we're actually using the TDA7293 chip.
P.S.
I'm interested in volume level shifting with the clip indicator pin, an rc for ~15ms delay (may require transistor buffer for the clip indicator pin) and a jfet to accomplish the volume change. That is a compressor used to reduce the duration of clipping. Unlike a clipper, a slow compressor won't cause clipping sounds at the treble, but instead at the bass. That is a more palatable combination.
The "sound of":
Hard clipper and soft clipper do raspy (a bit trashed) treble and good bass at less output. -compare- Slow compressor does smooth intact treble and crunchy bass at proportional output.
Possible usage:
By using a clipper to limit peak output and a compressor to limit constant current output you can keep both specifications of output devices within safe limits while also maximizing them when needed.
P.S.
I'm interested in volume level shifting with the clip indicator pin, an rc for ~15ms delay (may require transistor buffer for the clip indicator pin) and a jfet to accomplish the volume change. That is a compressor used to reduce the duration of clipping. Unlike a clipper, a slow compressor won't cause clipping sounds at the treble, but instead at the bass. That is a more palatable combination.
The "sound of":
Hard clipper and soft clipper do raspy (a bit trashed) treble and good bass at less output. -compare- Slow compressor does smooth intact treble and crunchy bass at proportional output.
Possible usage:
By using a clipper to limit peak output and a compressor to limit constant current output you can keep both specifications of output devices within safe limits while also maximizing them when needed.
Last edited:
actually i've been thinking somewhat along the same lines but perhaps with LDR for attenuation.
i also want to explore how well a known authentic chip would perform with roughly CarlosFM-sized NFB parts such as no-cap 330R shuntR for 10k2 seriesR. (and using noninverting vs inverting+buffer).
BTW my authentic chips behave much more reasonably than my ebay-sourced chips.
i also want to explore how well a known authentic chip would perform with roughly CarlosFM-sized NFB parts such as no-cap 330R shuntR for 10k2 seriesR. (and using noninverting vs inverting+buffer).
BTW my authentic chips behave much more reasonably than my ebay-sourced chips.
Earlier in this thread I showed an outrigger off the side of the board with protector where you could stack up as many caps as necessary for decorous results (without an extreme of board resoldering labor), and I still do recommend that approach (especially concerning lowest loss and optimized capacitance value).
Either that or a dc tracker.
Either that or a dc tracker.
Just a heads-up to those interested in buying ebay boards. one of mine (sourced from Kuyaya -- see their link above) arrived with a 47uf Sam Young branded (yellow case) cap for the bootstrap. it bulged, popped, and bubbled electrolyte after less than one hour of run time. checking the below "bad cap" site, Sam Young brand is noted among those noted for problems.
see: Bad capacitors
i replaced mine with a 100uf Panasonic FC, which has worked just fine. the (authentic) 7293 chips survived the drama.
Last edited:
Thanks Steve! I've removed Kuyaya from the list. As far as I know, HappyShop has been shipping some good caps.
#1
100u bootstrap. That's one of three things to do for pretty bass.
#2
Right-sized in- coupling cap is the main thing for pretty bass (the cap shipped with the kit is far too small--perhaps not a bad cap but for sure a really bad capacitance value). A bigger cap is inconvenient to select and will need a little (electro or polyester) bypass cap for good imaging, but pretty bass is only doable if allowed by the right size cap. Will you allow amplification of pretty bass?
If that coupling cap is inconvenient for size, increase feedback resistor to 47K and increase its partner, the feedback-shunt-resistor by same proportion. When the latter is higher value, the needed coupling cap is smaller by same proportion. That combination was the "runner up" in the all values live competition, and probably valid. The amount of optimal possibilities were surprisingly few.
#3
Several options are addressed at post#30, but the little array in the photographs does have a bit more bass slam and power despite its greater voltage drop and clipping mitigation; because, also greater filtering yields a more favorable proportion (see also leonard audio).
P.S.
How did you like the sound of your new amp? If you had put in the two trimmers (those are, adjustable gain and adjustable input load), and the power filter, then I'll bet you could work the tone around any way you wanted, without having to add lots of tubes at the input, without needing a preamp. Those are, if the amplifier is modded as indicated (post1&30).
#1
100u bootstrap. That's one of three things to do for pretty bass.
#2
Right-sized in- coupling cap is the main thing for pretty bass (the cap shipped with the kit is far too small--perhaps not a bad cap but for sure a really bad capacitance value). A bigger cap is inconvenient to select and will need a little (electro or polyester) bypass cap for good imaging, but pretty bass is only doable if allowed by the right size cap. Will you allow amplification of pretty bass?
If that coupling cap is inconvenient for size, increase feedback resistor to 47K and increase its partner, the feedback-shunt-resistor by same proportion. When the latter is higher value, the needed coupling cap is smaller by same proportion. That combination was the "runner up" in the all values live competition, and probably valid. The amount of optimal possibilities were surprisingly few.
#3
Several options are addressed at post#30, but the little array in the photographs does have a bit more bass slam and power despite its greater voltage drop and clipping mitigation; because, also greater filtering yields a more favorable proportion (see also leonard audio).
P.S.
How did you like the sound of your new amp? If you had put in the two trimmers (those are, adjustable gain and adjustable input load), and the power filter, then I'll bet you could work the tone around any way you wanted, without having to add lots of tubes at the input, without needing a preamp. Those are, if the amplifier is modded as indicated (post1&30).
actually i've been happy overall with kuyaya and plan to continue to buy from them.
i don't really blame then for the Sam Young cap issue because at the board prices i don't really expect the vendor to do the amount of testing and homework necessary to have spotted this issue. indeed i did not spot the issue before the first actual failure. some of the other vendor use the same cap (yellow cap wrapper shows in the pictures so it is easy to spot).
the new amp is like a night and day -- completely different from my earlier inverting one and very much cleaner. as before i put the NFB loop right at the chip pins. but i'm not using a linestage with the new amp so i set up the new amp as non-inverting, the gain is set at about 33x using mil. spec. 1% carbon film resistors instead of the stock metal film stuff, and 100% direct coupling -- no input cap and no FB cap. output offset is under 10mv either hot or cold -- which is certainly good enough. also, since my last post i increased the bootstrap cap to 220uf low ESR Nichicon.
nobody seems to be able to pick out in a blind test which amp i'm playing among this new amp and either of my TI chipamps (one is directly coupled, and with very low FB R values, NI gainclone type and the other is a directly coupled TI BPA300-style design with OPA2227 for phase splitting (i use the low offset precision version not the looser tolerance "audio grade" chip.)
Last edited:
PS: forgot to mention that all three are run at +/- 25v (8 ohm load) so these are relatively conservative designs. BTW i also have put the better 6922 linestage back in the setup which actually made only a little difference -- a bit quieter overall and a bit warmer from the 2nd harmonic distortion (roughly .07%) courtesy of the tubes/valves -- 2 parallel 6922s w/ DNMOSFET current source, output taken from mu follower point in the CCS, all fed w/shunt regulated supply. i don't have much impedance mismatch between source and the amps nad the interconnects are short, so the linestage isn't particularly important to have, except to satisfy my personal taste for a bit of "valve" flavoring. 
noticed this while re-reading the thread. i suspect that what you have going on here is that the diodes cut off the sinking of current from the board's caps that would occur (without the diodes) during the lower half of each ringing spike from all those big parallel caps in your PSU. normally such ringing alternately sources and sinks current with respect to the board's local reservoir/bypass caps but by cutting off the sinking half of it with the diodes you have eliminated a burden on those caps and increased overall power is therefore available for the chipamps. the diodes do not cut off the sourcing that occurs from the upper half of each spike and that contributes further to the available power.
Last edited:
The part that you just quoted is utilized to neutralize the amplifier's sonic signature at the power circuit. That is useful so that the time consuming chore of selecting capacitor size, favorable model and bypass caps for ac coupling is easier instead of overwhelming. The reward of such labor is lifelike music dynamics.
ac coupling provides lifelike music dynamics at the cost of amplifying an equal amount of the amplifier's sonic signature. An additional cost of ac coupling is the labor to optimize it.
dc coupling flattens the amplifier's sonic signature at the cost of flattening an equal amount of music signal. That is an ugly thing to do to an amplifier that doesn't require such heavy handed dampening. Some amplifiers do need it; but, this one does not. So, I've been having a hard time even making sense trying to reply to your posts, because of that one grotesque factor in an otherwise lovely job.
P.S.
Your lower rail voltage has allowed the lower gain to succeed, the net of which is a huge boost in imaging performance at the cost of lower output power, 22W. I think your quality over quantity approach is ABSOLUTELY FANTASTIC!!!
Well, except for the dc coupling (speechless and grossed out). Please do repair that part when you have a Lot of spare time for it.
Personally, I have no need for the "protection" features of full ac coupling because all of my speakers have ac coupled woofer and extra tweeter protection. However, all of my amplifiers are ac coupled in order to allow the amplification of full quality lifelike music dynamics.
ac coupling provides lifelike music dynamics at the cost of amplifying an equal amount of the amplifier's sonic signature. An additional cost of ac coupling is the labor to optimize it.
dc coupling flattens the amplifier's sonic signature at the cost of flattening an equal amount of music signal. That is an ugly thing to do to an amplifier that doesn't require such heavy handed dampening. Some amplifiers do need it; but, this one does not. So, I've been having a hard time even making sense trying to reply to your posts, because of that one grotesque factor in an otherwise lovely job.
P.S.
Your lower rail voltage has allowed the lower gain to succeed, the net of which is a huge boost in imaging performance at the cost of lower output power, 22W. I think your quality over quantity approach is ABSOLUTELY FANTASTIC!!!
Well, except for the dc coupling (speechless and grossed out). Please do repair that part when you have a Lot of spare time for it.
Personally, I have no need for the "protection" features of full ac coupling because all of my speakers have ac coupled woofer and extra tweeter protection. However, all of my amplifiers are ac coupled in order to allow the amplification of full quality lifelike music dynamics.
Last edited:
we'll have to accept respectful disagreement on coupling since we have not heard each others' amp.
i'm aiming high -- to rival a high-end class A amp. there are many ways to skin that cat.
i've tried it 4 ways so far:
1. your design but with smaller loop areas (helps a lot, BTW)
2. cap coupled, small loop areas, t-network feedback bypassed w/poly
3. pricey mu-metal transformer
4. correctly implemented DC instrumentation-style opa2227 driver
2, 3, and 4 made the grade. 1 was close behind but with noticeable high-order distortion products.
but you need not take my word for it -- #2 is very easy and cheap to do. #3 is not cheap to do and #4 is not easy to do correctly.
PS:
#3 and #4 transition easily to end-to-end balanced topology which is also a goal eventually. #3 is very expensive, however.
i'm aiming high -- to rival a high-end class A amp. there are many ways to skin that cat.
i've tried it 4 ways so far:
1. your design but with smaller loop areas (helps a lot, BTW)
2. cap coupled, small loop areas, t-network feedback bypassed w/poly
3. pricey mu-metal transformer
4. correctly implemented DC instrumentation-style opa2227 driver
2, 3, and 4 made the grade. 1 was close behind but with noticeable high-order distortion products.
but you need not take my word for it -- #2 is very easy and cheap to do. #3 is not cheap to do and #4 is not easy to do correctly.
PS:
#3 and #4 transition easily to end-to-end balanced topology which is also a goal eventually. #3 is very expensive, however.
Last edited:
Thanks Steve. That makes sense.
Also to mimic the sound of a class A amplifier, using ordinary class ab chip amplifiers for the job, you might want parallel /w ballast build to eat up some output device and crossover noise in the ballast resistors.
The no-ballast parallel amp on this thread is a low loss push for power type build with a lot of headroom for dynamic peaks but at the cost of greater output device noise, since there aren't any ballast resistors to eat that up along with some of the dynamics.
Anyway, I got a question:
Have you got a schema for option#2 above, the T-net?
Also to mimic the sound of a class A amplifier, using ordinary class ab chip amplifiers for the job, you might want parallel /w ballast build to eat up some output device and crossover noise in the ballast resistors.
The no-ballast parallel amp on this thread is a low loss push for power type build with a lot of headroom for dynamic peaks but at the cost of greater output device noise, since there aren't any ballast resistors to eat that up along with some of the dynamics.
Anyway, I got a question:
Have you got a schema for option#2 above, the T-net?
Last edited:
there are several schematics for it here at DIY.
let's assume for instance that you're working on a NIGC and you want low offset without the big electro in the feedback path.
then the approach is to have the usual voltage divider at the NIGC output but: 1) without the series cap and 2) instead of directly tying the junction of the two voltage divider Rs (point A) to the chip -IN (point B), you instead place a high value R (22k1 to 47k5) in between point A and point B. that high value R may be bypassed with a modest-value decent quality poly C (e.g.: Panasonic ECQ-E1105KF 1uf box cap).
you can tinker with the voltage divider to get the gain you want. and to minimize offset, you can tinker with the NIGC input network and the point-A-to-point-B R value. and to shape the bottom-end response you can tinker with the bypass C. all very tunable and no typical NIGC big electro in the feedback path.
in one of the specific circuits here on DIY, the poster had, in his LM3886 NIGC, a 20k resistor tied to the output on one one side and to 3 things on the other side: a 1k R to SigGnd, a 27k R to -IN, and a 1 uf cap to -IN. the 27k was determined experimentally (with a pot).
let's assume for instance that you're working on a NIGC and you want low offset without the big electro in the feedback path.
then the approach is to have the usual voltage divider at the NIGC output but: 1) without the series cap and 2) instead of directly tying the junction of the two voltage divider Rs (point A) to the chip -IN (point B), you instead place a high value R (22k1 to 47k5) in between point A and point B. that high value R may be bypassed with a modest-value decent quality poly C (e.g.: Panasonic ECQ-E1105KF 1uf box cap).
you can tinker with the voltage divider to get the gain you want. and to minimize offset, you can tinker with the NIGC input network and the point-A-to-point-B R value. and to shape the bottom-end response you can tinker with the bypass C. all very tunable and no typical NIGC big electro in the feedback path.
in one of the specific circuits here on DIY, the poster had, in his LM3886 NIGC, a 20k resistor tied to the output on one one side and to 3 things on the other side: a 1k R to SigGnd, a 27k R to -IN, and a 1 uf cap to -IN. the 27k was determined experimentally (with a pot).
i as well do not understand it
in any case, when i said "rival" i didn't mean "mimic".
it's going to sound different from class A. i just don't want it to sound too much worse.
Last edited:
Steve, you could build a parallel LM1875 and then you'll know. It is not easily done, but it is rather amazing.
AndrewT, perhaps I explained the positive and negative effects of output/ballast resistors in a way that was far too layman/colloquial; however, on this thread, our parallel amp doesn't have those resistors, so that the chosen set of compromises is different, which makes for an interesting comparison.
AndrewT, perhaps I explained the positive and negative effects of output/ballast resistors in a way that was far too layman/colloquial; however, on this thread, our parallel amp doesn't have those resistors, so that the chosen set of compromises is different, which makes for an interesting comparison.