For 1,2 and 3 : I actually agree with you on all counts 
But this is a real world.
Yes, but here I might add that I believe that most datasheet numbers are targeted for experienced professionals. it is clear for all, those component makers want to sell their stuff on big industrial volumes, not the mere DYI market. And many industries (inclusive audio) which buy these components actually agree with component makers what is relevant to be specified. They have (had) the know-how to correctly interpret what these numbers really mean.
TI or LT makes beautiful, even genial components. Somehow yet they are not the choice of audio professionals in big audio industry. This is for me an interesting puzzle asking for a proper answer
Cheers again,
Ionmw
But this is a real world.Yes, but here I might add that I believe that most datasheet numbers are targeted for experienced professionals. it is clear for all, those component makers want to sell their stuff on big industrial volumes, not the mere DYI market. And many industries (inclusive audio) which buy these components actually agree with component makers what is relevant to be specified. They have (had) the know-how to correctly interpret what these numbers really mean.
TI or LT makes beautiful, even genial components. Somehow yet they are not the choice of audio professionals in big audio industry. This is for me an interesting puzzle asking for a proper answer
Cheers again,
Ionmw
A good professional should know that single number THD is meaningless so will mostly ignore it.
If the part is interesting they will buy one and try it and test it themselves.
dave
Well, apparently Joe did not ignore it But also he did supply proper info.
All good professionals would have first evaluate the chip (and tested it for THD too) before their choice.
So we agree THD is on the list of measurements.
What next is needed... to identify a good replacement candidate for LM3875 on Joe's transamp? A model for THD, maybe?
Cheers
Ionmw
But also he did supply proper info.All good professionals would have first evaluate the chip (and tested it for THD too) before their choice.
So we agree THD is on the list of measurements.
What next is needed... to identify a good replacement candidate for LM3875 on Joe's transamp? A model for THD, maybe?
Cheers
Ionmw
Well, apparently Joe did not ignore it
So you can ignore the meaningless bit.
Forget THD. You can start by looking at the HD.
As i said early, the diskreet version of Daniel's TransAmp was better than the chip version (which was still very good). The diskreet version was originally built to drive the electron guns of a CRT and then modified for audio.
dave
Dave you know that I agree with you 50%. Physics say that the amplifier contribution is way exaggerated. Remember the 85's-90's, when reported "HiFi" values were kind of 0.000006% THD, insane! And yet they sound MORE awful than their THD1%'ers tube brothers.. or cousins or so. This I believe have been the reason Sansui went the way of Dodo, right after their genial feed-forward.
So, the CS is the secret. We all admit it, and some want to do a better job. I think is time to understand if LM3875 is really the best, and why, at this price point.
What do you think.
Ionmw
And yet they sound MORE awful than their THD1%'ers tube brothers.. or cousins or so. This I believe have been the reason Sansui went the way of Dodo, right after their genial feed-forward.So, the CS is the secret. We all admit it, and some want to do a better job. I think is time to understand if LM3875 is really the best, and why, at this price point.
What do you think.
Ionmw
Regards HD… it really needs to be expressed as a 3 dimensional manifold in 4-space (frequency of excitation, magnitude of output at all frequencies, magnitude of excitation). Due to display limitations this would likely be best displayed as a projection of a 2D surface in 3-space with a slider to vary the missing dimension and the ability to rotate the 2D projection.
dave
dave
Nice idea ! ... more suggestions
- add a 4th dimension for power.
- add a 5th dimension for variable Transimpedance vs. speakers and PS
- ... 6th, 7th etc.
CS amp = current source amp (aka. Transimpedance, aka. CFB current feed-back etc.)
I live to see a CFF (current feed forward) for DYIers.
Look, I am sorry. What I mean to say, I believe it is important for other DYIers friends not to follow my personal wishes, but to be able to solve their questions or problems, like this one:
- I have an old voltage amp. It has PS, it has buttons, connectors etc.
- some guys (like me) advise them to understand CS configuration, take Joe's values and notes and then simply disconnect the old output, make CS configuration, put proper PS for stereo operation (aka double PS).. and voila... magic.
- or... think this way: "oh no.. only LM3875 does the job... much better than the old components in my old amp... so I am doomed because I have to take out everything and build a new one... well yes with less features that I actually need or want... etc etc
- what to do... I am doomed etc
Or another kind of questions:
- If I want to simply reconfigure my old TA-F470 amp as CS... will sound better? Personally I am sure that the high quality Sanken transistors here will sound better than LM3875. This is not a criticism: I personally love TI and LT and many more companies.
- How to make the best of my existent amp? Change wires or change everything?
- How not to destroy my expensive loudspeakers?
- How not to set the house on fire?
- How to ... check, debug and optimize my first results?
- most important one: when to deffer this job to a specialist ?
Right?
Ionmw
- add a 4th dimension for power.
- add a 5th dimension for variable Transimpedance vs. speakers and PS
- ... 6th, 7th etc.
CS amp = current source amp (aka. Transimpedance, aka. CFB current feed-back etc.)
I live to see a CFF (current feed forward) for DYIers.
Look, I am sorry. What I mean to say, I believe it is important for other DYIers friends not to follow my personal wishes, but to be able to solve their questions or problems, like this one:
- I have an old voltage amp. It has PS, it has buttons, connectors etc.
- some guys (like me) advise them to understand CS configuration, take Joe's values and notes and then simply disconnect the old output, make CS configuration, put proper PS for stereo operation (aka double PS).. and voila... magic.
- or... think this way: "oh no.. only LM3875 does the job... much better than the old components in my old amp... so I am doomed because I have to take out everything and build a new one... well yes with less features that I actually need or want... etc etc
- what to do... I am doomed etc
Or another kind of questions:
- If I want to simply reconfigure my old TA-F470 amp as CS... will sound better? Personally I am sure that the high quality Sanken transistors here will sound better than LM3875. This is not a criticism: I personally love TI and LT and many more companies.
- How to make the best of my existent amp? Change wires or change everything?
- How not to destroy my expensive loudspeakers?
- How not to set the house on fire?
- How to ... check, debug and optimize my first results?
- most important one: when to deffer this job to a specialist ?
Right?
Ionmw
I have an Loewe radio (transistors) in my office. No other radio sounds as good as that and I am a very picky guy on this matter.
I also drive my old Duals CL390 (700DM) or CL720 (1500DM) with a studio amp Graetz HSA-200. All real Klasse!
But my best sounding system was a tube, back at parents home. I want to meet that point again, but without tubes.
My first goal with Joe's transamp is to meet quality of Graetz... or the broken Sony F470. I will use as much of the broken amp. Mostly because of existence of proper Case and Functions (wife and kids standards).
Cheers again,
Ionmw
Not a question that can be answered independent of the speakers they are connected to. We have been living so long in a world where voltage amps are the norm that speaker designers pay little attention to flat impedance curves. I'd say it is worth a try, but in the end you may find it easier to start from scratch (i see no issue reusing your output devices).
BTW, a pentode amp with no feedback is effectively a current amp.
Everything is give & take, to get the advantages of CS amps, one has to pay attention that the speaker's impedance curve is flat -- or a complement of the FR. With the full range speakers we used with the transamp, the rise in impedance at the top & bottom would increase the bandwidth of the speakers as power input increases with the rise of the impedance at the extremes.
With voltage amps the system format is established and any amp tends to be able to drive any speaker… moving into CS amps one has to take the entire system into consideration on each individual case. Joe's work to create flat impedance speakers gives us more choices for more people -- me not so much, i am not a big fan of dome tweeters and XOs at frequencies higher than a ¼ wavelength of the driver centre-to-centre.
dave
Well yes. Since I do not want to destroy my old speakers, for this project I need first to build my very first set of speakers. For test but I do not mind if they will sound better than profesional ones. Question, do you mind if I will attempt to paint my 12"woofers copycat your style, just to limit their response up to 500Hz or 1KHz? I intend using innocent XOs (not too steep) to preserve the overall phase too, not just phase alignment at XO freq. I feel that I will need mechanical help for XOs to obtain linear phase + minimal overall phase change.
The 1/4 wavelength for acoustic design I do not understand yet. You mentioned it before too. Could you explain or send PM if it is sensitive? The wavelength of waves I do understand, it is my job.
I am fan of silk domes, I find them OK, just a bit grainy though (broken modes). Better than the pressure drivers, metal domes or cones, though. Else I did not have other experiences like electrostatics or acoustic transformers.
Ionmw
PS. speaker topic as beautiful it is does not eliminate the question of better controlling the Transimpedance design.
From F470 actually I will keep the infrastructure (safety sensors too), PS and all Input connection, selection etc. It is specifically the question: shall I repair the bias and B driver of Poweramp (which is good) and reroute output in CS configuration... or... to cut board before bias amp (remove bias+Bamp+Poweramp) and there to insert my test power modules SI/STK/LM3875 based transimpedance? All choices possible, not decided yet, but require progressively more PS regulation work.
The final amp I can test acoustically at my home or some neghbarhood friends (all professional musicians), electrical tests at lab in winter holiday, or so I hope.
The 1/4 wavelength for acoustic design I do not understand yet. You mentioned it before too. Could you explain or send PM if it is sensitive? The wavelength of waves I do understand, it is my job.
I am fan of silk domes, I find them OK, just a bit grainy though (broken modes). Better than the pressure drivers, metal domes or cones, though. Else I did not have other experiences like electrostatics or acoustic transformers.
Ionmw
PS. speaker topic as beautiful it is does not eliminate the question of better controlling the Transimpedance design.
From F470 actually I will keep the infrastructure (safety sensors too), PS and all Input connection, selection etc. It is specifically the question: shall I repair the bias and B driver of Poweramp (which is good) and reroute output in CS configuration... or... to cut board before bias amp (remove bias+Bamp+Poweramp) and there to insert my test power modules SI/STK/LM3875 based transimpedance? All choices possible, not decided yet, but require progressively more PS regulation work.
The final amp I can test acoustically at my home or some neghbarhood friends (all professional musicians), electrical tests at lab in winter holiday, or so I hope.
Last edited:
Very unlikely. A non-flat impedance curve might make them sound like crap, but all the same rules apply to keeping speakers alive.
Treating paper cones with mod-podge has been something i have been doing (and advocating) since the late '70s. The much greater effect of EnABL is something Bud gave to all diyers over half a decade ago.
If i do an XO it is almost invariably a 1st order XO. If you keep the XO at or below ¼ wavelength centre-to-centre with a 1st order XO then the end result is almost guaranteed to have really good phase performance and no loving.
This is a tidbit from Tom Danley.
If you do the math no current "tweeters" will let you keep a ¼ WL relationship with what is used lower in frequency. There are a very few small FR drivers that have every bit as good a top as tweeter and will let you push the XO down to below 400 Hz (the neighborhood where one can preserve that magic ¼ WL).
dave
I'm not an amp guy (i can build a kit or do a simplish tube amp from scratch, but this 3 legged "fuses" are something i have never really attempted to understand enuff to design an amp (or anything for that matter). And converting an old amp to something quite different is a redesign.
One thing i will note is that the PS of many of these Japanese amps is higher voltage and lower current than you would really like to start with. What are they in your amp (a quick look at the service manual does not reveal that, but at 80+80 rated the rails are probably too high for LM3875 which likes 25-40VDC)
dave
Thanks for acoustic 1/4 WL center-to-center references.
The test system here is planned to comprise an active 3-way approach: woofer-bass 2x12" piston up to 400, XO at 800, pure mids 400 up to 5K in a 2" paper dome, then XO at 10K with 1" silk dome. All domes are expensive Heco. The 1st set of 12" is Fostex (faster, soft membrane) and 2nd set of 12" is Yamaha (wider range, stiffer cone), both in parallel to damp each other their mechanical resonance; additional series notch in parallel to linearize the impedance. All drivers Z-characterized and will be a tri-amp with XO before CS finals, inclusive interstage buffers and overload, thermal, start-up shut-down sensors etc.
If any non-flatness remains I hope to compensate through the "active" XOs. I say active because they are placed before the PowerAmp cards but I hope they will be of innocent LC type, passive.
If I choose repair+rerouting of F470 in CS configuration I get the quality of expensive-parts but I cannot do above plans. Tough decision...
Need to shift this discussion for another day.
Cheers and regards,
Ionmw
P.S. I have seen the other day power cards for power deflection circuits; yes possibility to adapt them to audio is .. great say the least. I amhappy to learn somebody did it and reports predating results. Better off, the power components were... Sanken
The test system here is planned to comprise an active 3-way approach: woofer-bass 2x12" piston up to 400, XO at 800, pure mids 400 up to 5K in a 2" paper dome, then XO at 10K with 1" silk dome. All domes are expensive Heco. The 1st set of 12" is Fostex (faster, soft membrane) and 2nd set of 12" is Yamaha (wider range, stiffer cone), both in parallel to damp each other their mechanical resonance; additional series notch in parallel to linearize the impedance. All drivers Z-characterized and will be a tri-amp with XO before CS finals, inclusive interstage buffers and overload, thermal, start-up shut-down sensors etc.
If any non-flatness remains I hope to compensate through the "active" XOs. I say active because they are placed before the PowerAmp cards but I hope they will be of innocent LC type, passive.
If I choose repair+rerouting of F470 in CS configuration I get the quality of expensive-parts but I cannot do above plans. Tough decision...
Need to shift this discussion for another day.
Cheers and regards,
Ionmw
P.S. I have seen the other day power cards for power deflection circuits; yes possibility to adapt them to audio is .. great say the least. I amhappy to learn somebody did it and reports predating results. Better off, the power components were... Sanken
I am OK with this plan.
I digged more and found Trafo is 500VA rated. It weighs so anyway. Good for transients.
Secondaries will put out 47V. I will keep so to preserve rest of F-470 functionality. So, just take the hit....... and need to dissipate 10V at 2A (aka 20W/stage) on each regulator (12x total).
New bridges: ordered 20x the Schottky types Joe recommended, just perfect.
PS complete rework is planned, whatever choice above. V-control LM317/337+TIP 142/147 (19A capable).
Optimizations: to modify the Z-out from 470ohm to 72ohm or even 50ohms: will condition the response on any speaker resonances or RL-behaviour (anything not handled by XOs design) midway towards voltage driving. The resonance effects on speakers anyway have to be damped. Right?
If you have any practical ideas please tell, so I can plan it now. I am open to suggestions in this phase.
I am more concerned how to paint 12" drivers iteratively (trial and error) since I want to apply BU overcoating too. The two wishes have a conflict.
Ionmw
I have brought this up before and there has only been a dismissive reaction... nothing to explain. Huh? There are some out there that simply treat much of what I have to say, I am not an approved sort worth listening to. There are people who know what I mean and I am choosing my words very carefully. But these people also have the ability to unlock the question raised, if they would take it seriously
Obviously the distortion profile is different - and yes, that measurement is for real and part of a series of measurements that was performed, both harmonic and IMD - and the pattern persists, the suppression of odd order relative to evens. Anybody out there can repeat the measurements, be my guest. Easy.
Dave is right, THD means nothing unless we see the spectral distribution of the harmonic distortions. This was established decades ago by Jean Hiraga and built upon ever since, like John Curl, that certain circuit blocks generate/suppress even/odd order distortions, so much that designers try to come up with the right mix-and-match in the end, balancing the circuit.
But here we have a mechanism that needs to be explained - why does the LM3875 have this profile when used in current mode when it does not have the same desirable traits when used in voltage mode. What mechanism is at play, how can it be explained? As I said, I have tried to bring this up a number of times and the 'talking heads' that could do something to decipher it, but since I am not in their good books, it gets dismissed as noise.
I don't like components that gets a cult following - but I can see why some find it reassuring as a psychological device. I have built about a hundred amplifiers based on LM3875. Will the 'current' version distortion profile that I have published also exist with Sanken devices when also used in current mode? I suspect yes, but it needs to be confirmed. After building, it's not difficult to measure and confirm.
The very ones who could maybe decipher what explains that distortion profile, I have enough faith in their ability to do that, but maybe they are not interested in current driving loudspeakers - their world-view is very orthodox and that means voltage is the way to go for them. So they are not interested when they should be intrigued?
What I do know is that the LM3875 sounds very different used this way with a compatible speaker and likely other chip and non-chip amplifiers can do the same. Take any kit amplifier from a local electronics store, even built entirely of discrete components, figure out a few things and end up putting the speaker inside the f/b loop with a current sensing resistor - it can be done with any conventional , IC or not, voltage amplifier. Many of them use similar circuits and chances are we will see that same distortion profile. I am only saying this as I don't think this topic is really about the LM3875. Maybe all f/b SS amps will behave this way?
Last edited:
Yes, actually this detail in your first page of this Thread have attracted my attention. I still hope to get closer to this answer with help from kind knowledgeable forum members.
I see possible that we ask too much of "why", so people who can really digress on LM3875 (or other IC) internals would prefer not to pick up this topic at all.
See AD app note below.
- Maybe I am wrong but I thought that even harmonics are normally smaller than odd ones in Opamp.
- Also, my take on IMD is that they are dependent of HD, so even having contributions below the limit of "direct" audible HD, the leakage of these terms into IMD may make a great difference. Maybe I am wrong to generalize, but this is my current understanding.
- I also note that mybelief until now was that even order harmonics are "always" (with normal good amps or OpAmps and when correct implemented) lesser than odd ones. The fact that our ears detect odd ones more efficiently is another story.
Still, I wonder if CS configuration would help clarity by internal amp mechanisms, e.g. not only by reaction of speakers. I will wait for answer.
Greetings,
Ionmw
P.S. Joe, to help us all make an easy comparison of configurations... could you post HD/IMD measurements for LM3875 as normal gainclone too? Are you setup to measure?
