Yes, I want only to squash the power supply's peaks so that the bass notes play in proportion.
Juergen,
with all do respect to Daniel sometimes he does "Sprecht Shiza" LOL
In any case, Power Supply "ripple rejection" / "stiffening" IS one area where TUBES and SS are very similar.
Daniel,
In all seriousness you have to think of ripple rejection/capacitance and Voltage lowering as three interrelated objectives.
In the tube world the PSRR of tubes is not as good as chips. So we rely upon "smoothing" the DC as much as possible. This is imperative in just about ALL tube designs regardless of application.
I often use a very old set of analogies that I learned a long time ago.
Think of electricity as WATER, the pressure is Voltage, Current is VOLUME.
Think of the output (sound) as a fixed size jet that can be turned off and on rapidly (like a fuel injector in a car) We want to be able to supply the WATER needed for ANY of the frequencies of the "jet". Low frequencies require MORE water but the PUMP (the Power Supply) takes TIME to produce a given volume/pressure. The HIGH frequencies are pretty easy since they require very short pulses at a given pressure and thus MUCH less Volume (current). BUT, low frequencies need the equivalent Pressure but MUCH more VOLUME.
So we have a pump (transformer) it has "pulses" at a specific frequency. (120Hz here in the US using a bridge) this travels down pipes or hoses to the different components. We also use the water to "activate" the jets valve. Any pulses on the water line are then transfered to the jet at that frequency regardless of the control circuit. Pulses above a certain threshold will result in water being spurted out of the jet. We need to limit the pulses so the jet functions as we want it to.
So we place "accumulators" (capacitors) in the line they fill up with a quantity of water to the pressure of the pump. While full if the jet calls for equal or less Volume than the pump delivers we get the supply from the pump not the accumulator. If it calls for more we get it from the accumulator until the accumulator is "exhausted" and we are then relying on the pump again.
Just use HUGE accumulators correct? Well that works to some extent because given infinite capacity we get perfect smooth endless supply of water. BUT, we cannot afford infinite capacitance because there is a limit to them imposed by its resistance (ESR) and cost, size etc.
The best solution is always a compromise. Since we have a pulse with a defined frequency (time) we can then "filter" it out by using a resistor and capacitor together to IMPEDE the pulse. (cannot eliminate it but can cut down its amplitude).
If the PUMP can supply a reasonable amount of Pressure and Volume then we simple place a filter in series that has a reactance suitable to impede the pulse.
This involves reducing the PRESSURE (voltage) thru the resistor in proportion to its resistance.
Since you have EXCESS voltage you can lower it by placing an RC filter in the line. You can use some of the calculators available online. Heres one to help.
You will have to do all the "playing around" with it but, most of the designs I have seen for chip amp supplies seem to have no rhyme or reason with which the are designed for ripple rejection (probably because of the high PSRR of the chips). Neverless I see no reason why we cannot "clean up" the supply further if you have excess voltage to spare.
A 1R resistor coupled with a 10,000uF cap to ground attenuates 120hz ripple by 18dB. You only lose 3V at 3AMPS RMS. Two of these in series should get rid of 6 volts and cut down on the ripple.
These need to be BIG resistors. BUT you could use larger caps and smaller R's to achieve similar results. Example: 0.5R and 30,000uf = loss of 1.5V at 3A and 21db of 120hz attentuation.
Now that we have that under our belt, how that relates to sound. Bigger caps equals TIGHTER bass, since we can supply more current more quickly.
Smaller caps = "muddier" bass or "SAG" sometimes (like in a guitar amp) "SAG" is desireable. In the "Audiophile" world it is much less desireable.
So the moral of the story:
Look at most tube amp power supplies, most will call for a Transformer with an output voltage at least 25% higher than the B+ requirement. Then the excess voltage is burned off in the RC or LC filters. Capacitance requirements are proportional to the current demand.
Typical TUBE PS might be something like this
1K5 --> 40uF --> 100R --> 150uF -->load @ 100mA we lose 160V but attenuate 120hz by like 70db
For SS current demands we need to factor the Resistors DOWN and the capacitors UP.
so for similar ripple rejection @ 3 amps and loss of 1.5V we could do this
0R5 --> 30,000uF --> 0R5 --> 30,000uF --> 0R5 --> 30,000uF --> load
By using "trios" of paralleled 10,000uF caps there is some advantage in that the ESR's act as filters also.
So you see that IMHO there is no substitute for capacitance in the PS and within reason you can never have enough.
BTW, you need to watch the cutoff frequency of the filters since it will also lower your bass response if it gets to high (above 10hz or so) since the output is AC you do not want to limit the ability to supply the current.
Sorry for the long winded post but it actually helps me to think this all thru from the beginning.
with all do respect to Daniel sometimes he does "Sprecht Shiza" LOL
In any case, Power Supply "ripple rejection" / "stiffening" IS one area where TUBES and SS are very similar.
Daniel,
In all seriousness you have to think of ripple rejection/capacitance and Voltage lowering as three interrelated objectives.
In the tube world the PSRR of tubes is not as good as chips. So we rely upon "smoothing" the DC as much as possible. This is imperative in just about ALL tube designs regardless of application.
I often use a very old set of analogies that I learned a long time ago.
Think of electricity as WATER, the pressure is Voltage, Current is VOLUME.
Think of the output (sound) as a fixed size jet that can be turned off and on rapidly (like a fuel injector in a car) We want to be able to supply the WATER needed for ANY of the frequencies of the "jet". Low frequencies require MORE water but the PUMP (the Power Supply) takes TIME to produce a given volume/pressure. The HIGH frequencies are pretty easy since they require very short pulses at a given pressure and thus MUCH less Volume (current). BUT, low frequencies need the equivalent Pressure but MUCH more VOLUME.
So we have a pump (transformer) it has "pulses" at a specific frequency. (120Hz here in the US using a bridge) this travels down pipes or hoses to the different components. We also use the water to "activate" the jets valve. Any pulses on the water line are then transfered to the jet at that frequency regardless of the control circuit. Pulses above a certain threshold will result in water being spurted out of the jet. We need to limit the pulses so the jet functions as we want it to.
So we place "accumulators" (capacitors) in the line they fill up with a quantity of water to the pressure of the pump. While full if the jet calls for equal or less Volume than the pump delivers we get the supply from the pump not the accumulator. If it calls for more we get it from the accumulator until the accumulator is "exhausted" and we are then relying on the pump again.
Just use HUGE accumulators correct? Well that works to some extent because given infinite capacity we get perfect smooth endless supply of water. BUT, we cannot afford infinite capacitance because there is a limit to them imposed by its resistance (ESR) and cost, size etc.
The best solution is always a compromise. Since we have a pulse with a defined frequency (time) we can then "filter" it out by using a resistor and capacitor together to IMPEDE the pulse. (cannot eliminate it but can cut down its amplitude).
If the PUMP can supply a reasonable amount of Pressure and Volume then we simple place a filter in series that has a reactance suitable to impede the pulse.
This involves reducing the PRESSURE (voltage) thru the resistor in proportion to its resistance.
Since you have EXCESS voltage you can lower it by placing an RC filter in the line. You can use some of the calculators available online. Heres one to help.
You will have to do all the "playing around" with it but, most of the designs I have seen for chip amp supplies seem to have no rhyme or reason with which the are designed for ripple rejection (probably because of the high PSRR of the chips). Neverless I see no reason why we cannot "clean up" the supply further if you have excess voltage to spare.
A 1R resistor coupled with a 10,000uF cap to ground attenuates 120hz ripple by 18dB. You only lose 3V at 3AMPS RMS. Two of these in series should get rid of 6 volts and cut down on the ripple.
These need to be BIG resistors. BUT you could use larger caps and smaller R's to achieve similar results. Example: 0.5R and 30,000uf = loss of 1.5V at 3A and 21db of 120hz attentuation.
Now that we have that under our belt, how that relates to sound. Bigger caps equals TIGHTER bass, since we can supply more current more quickly.
Smaller caps = "muddier" bass or "SAG" sometimes (like in a guitar amp) "SAG" is desireable. In the "Audiophile" world it is much less desireable.
So the moral of the story:
Look at most tube amp power supplies, most will call for a Transformer with an output voltage at least 25% higher than the B+ requirement. Then the excess voltage is burned off in the RC or LC filters. Capacitance requirements are proportional to the current demand.
Typical TUBE PS might be something like this
1K5 --> 40uF --> 100R --> 150uF -->load @ 100mA we lose 160V but attenuate 120hz by like 70db
For SS current demands we need to factor the Resistors DOWN and the capacitors UP.
so for similar ripple rejection @ 3 amps and loss of 1.5V we could do this
0R5 --> 30,000uF --> 0R5 --> 30,000uF --> 0R5 --> 30,000uF --> load
By using "trios" of paralleled 10,000uF caps there is some advantage in that the ESR's act as filters also.
So you see that IMHO there is no substitute for capacitance in the PS and within reason you can never have enough.
BTW, you need to watch the cutoff frequency of the filters since it will also lower your bass response if it gets to high (above 10hz or so) since the output is AC you do not want to limit the ability to supply the current.
Sorry for the long winded post but it actually helps me to think this all thru from the beginning.
DC to Daylight stupid problem--never does this happen when I try it on purpose.
This calculator, from your link above, confuses me a bit, because. . .
It seems to combine the desirable, a 19db drop in noise, with the undesirable, the more current = the larger the voltage drop. My current draw should vary between idle and max on some music sources. Problem!
Question: Does the calculator represent an increase (worsening) of voltage fluctuations, or does it represent voltage fluctuation that would be present anyway, whether or not you used an RC? Sinking the voltage, only to have it sink yet more under load, is pointless when it is already fairly clean DC voltage.
Problem: Right now, the power supply seems able to convert signals from well below the 60 cycles, now up to the ghz range, including flipping the DC meter on open key transmitter, vacuum cleaner running from same outlet, etc. . . get a signal and you get yet more DC. Not cool! Noises are all converted to DC--really clean DC, along with a whopper of a caveat--about 10% of its total output is "weak" DC (top 10% of voltage has insufficient amperage). This is wrong. Its really unseemly for a bass amplifier. How to make only strong DC?
This calculator, from your link above, confuses me a bit, because. . .
It seems to combine the desirable, a 19db drop in noise, with the undesirable, the more current = the larger the voltage drop. My current draw should vary between idle and max on some music sources. Problem!
Question: Does the calculator represent an increase (worsening) of voltage fluctuations, or does it represent voltage fluctuation that would be present anyway, whether or not you used an RC? Sinking the voltage, only to have it sink yet more under load, is pointless when it is already fairly clean DC voltage.
Problem: Right now, the power supply seems able to convert signals from well below the 60 cycles, now up to the ghz range, including flipping the DC meter on open key transmitter, vacuum cleaner running from same outlet, etc. . . get a signal and you get yet more DC. Not cool! Noises are all converted to DC--really clean DC, along with a whopper of a caveat--about 10% of its total output is "weak" DC (top 10% of voltage has insufficient amperage). This is wrong. Its really unseemly for a bass amplifier. How to make only strong DC?
Yes.
It seems plausible that this would make an ample amount of very pretty bass along with BPA200 and 30+30vdc rails. At least it seems to be worth a try. Heatsinking may be a small challenge, but its one that I can do.
Maybe someone can relate this comment:
Not using any weak power, may make the sound of strength.
Last edited:
Hey guys, are we trying to make something like this: Excellent 2 X LM3886+6N11Tube Audio Amplifier Board - eBay (item 190349126840 end time Nov-22-09 20:26:20 PST)
is this somekind of business proposal?
Anyway...this pcb is not DIY-friendly, because of the difficulties of heatsink mounting AND a man should show his tube, instead of hiding it in an amplifiers case.
Let the tube punch through the top of the case like an...you know!
Regards
First test was this Building a Gainclone chip amp with T-network feedback. by Franz G, Joe Rasmussen and Nick Whetstone.
I just adapted it for LM3886 and connected it to the power supply posted earlier. A 4.7uF 250v polyester (normally seen on tweeters) was added to the output of that power supply at V+ to V- to short noise that would have otherwise increased with frequency. Caps at the chip pins are 470uF//100nF (perhaps they should be 330uF and/or perhaps the input dc block cap should be slightly smaller).
Curious results:
1). First time that I've ever heard an LM3886 not screaming, shouting, or otherwise misbehaving like a poorly compensated op amp. Yes, it sounds perfectly acceptable. Very decent large soundfield too. And the pretty bass sound of LM3886 is present (otherwise, I think that nobody would spend so much time fooling with this little chip).
2). Far too much gain!!! Not even the weakest MP3 player needs so much gain. I'd be the last person to complain about too much gain, but this really is too much.
3). Practically no power--it didn't play a bit louder than LM1875, until. . .
4). Since there's no NFB cap, I tried an output cap, and ended up with partial success up until the output cap trial and error became just like a copy of one rail of the power supply. . . then (only then) there was appropriate output power. Apparently, 0mv dc offset works much better than 10mv dc offset. Weird.
5). Maybe there is too much bass. You certainly wouldn't need a bass booster with this one.
The T-network design had too much gain and a few other oddities. It sounds surprisingly nice.
I just adapted it for LM3886 and connected it to the power supply posted earlier. A 4.7uF 250v polyester (normally seen on tweeters) was added to the output of that power supply at V+ to V- to short noise that would have otherwise increased with frequency. Caps at the chip pins are 470uF//100nF (perhaps they should be 330uF and/or perhaps the input dc block cap should be slightly smaller).
Curious results:
1). First time that I've ever heard an LM3886 not screaming, shouting, or otherwise misbehaving like a poorly compensated op amp. Yes, it sounds perfectly acceptable. Very decent large soundfield too. And the pretty bass sound of LM3886 is present (otherwise, I think that nobody would spend so much time fooling with this little chip).
2). Far too much gain!!! Not even the weakest MP3 player needs so much gain. I'd be the last person to complain about too much gain, but this really is too much.
3). Practically no power--it didn't play a bit louder than LM1875, until. . .
4). Since there's no NFB cap, I tried an output cap, and ended up with partial success up until the output cap trial and error became just like a copy of one rail of the power supply. . . then (only then) there was appropriate output power. Apparently, 0mv dc offset works much better than 10mv dc offset. Weird.
5). Maybe there is too much bass. You certainly wouldn't need a bass booster with this one.
The T-network design had too much gain and a few other oddities. It sounds surprisingly nice.
Last edited:
is this somekind of business proposal?
Anyway...this pcb is not DIY-friendly, because of the difficulties of heatsink mounting AND a man should show his tube, instead of hiding it in an amplifiers case.
Let the tube punch through the top of the case like an...you know!
Regards
For that application, the 6N11 somehow looks insufficient. Perhaps this: 6L6 may be more interesting?
An externally hosted image should be here but it was not working when we last tested it.
Daniel,
We all seem to have run off the original path intended for this experiment. Let's try to re-define the GOALS and have a run at building the ACTUAL project rather than spend time chasing "Phantom Noise" and "Pretty Bass".
My opion is this,
We design for a "Full Range" amplifier using a Tube as either a cathode follower buffer or Gain stage + Follower Buffer.
The topology of the chip portion needs to be as simple as possible. The tube section is a "piece of cake" I can pump out a schematic for just about any Dual Triode out there in an hour or two.
By making the GOAL a Full Range Amp we can just feed it with whatever signal we chose. In my case I feed it with the low pass output of my crossover.
My OPINION is that the chip needs to "disappear" in the amp. Leaving whatever distortion and colorings that the tube presents to be audible.
SO:
Here are the things for us to decide.
Changing the chip is an option.
Increasing the complexity is also an option (ie; Parallel or Bridged chips)
I would like to avoid the need for a PC Board, P to P wiring is much easier for Tube people to "swallow" in considering a project such as this.
I STILL do not understand why the gain of the chip portion can't be reduced to something on the order of 5 or 6? It seems that the chip is more likely to amplify noise from the case, powersupply and wiring, reducing the gain reduces this noise also.
Here us my opinion on the "Optimum" configuration that I would like to attemp. I think there is enough "expert" help available here to make a more complex project a "first time build" for me.
The choice of tube is irrelevant for now, but I would like to do a gain stage direct coupled to a "Split Load" inverter. (this will require a relatively HIGH B+ voltage for the Tube section as we need to run the gain stage at 1/3rd B+ in order to have a "balanced" load on the inverter.)
The Inverted and Non Inverted portions of the signal would then each have a "Parallel Pair" of chips. (this could be two LM 3886 or equivalent) or maybe just one LM4780?
Those "pairs" would then be "bridged" on the output only to the speaker load thru 0.1R resistors. No inductor, no Zobel just attach them and see what happens.
The pairs can be inverting or non-inverting as long as they are the SAME. The signal is already inverted for the lower pair.
Since the wiring might be a bit complex for P to P I can live with a piece of perfboard for the chips with copper and solder traces for the major connections.
OK, so there is what I want. Comments or suggestions? Please do not tell me why I SHOULDN'T do this, just what PROBLEMS I need to prepare for.
This choice is made to have this thing "LOOK" like a "Williamson" Push Pull Tube amplifier. Since one of the amps in my system is very much like this design it "blends" in well.
We all seem to have run off the original path intended for this experiment. Let's try to re-define the GOALS and have a run at building the ACTUAL project rather than spend time chasing "Phantom Noise" and "Pretty Bass".
My opion is this,
We design for a "Full Range" amplifier using a Tube as either a cathode follower buffer or Gain stage + Follower Buffer.
The topology of the chip portion needs to be as simple as possible. The tube section is a "piece of cake" I can pump out a schematic for just about any Dual Triode out there in an hour or two.
By making the GOAL a Full Range Amp we can just feed it with whatever signal we chose. In my case I feed it with the low pass output of my crossover.
My OPINION is that the chip needs to "disappear" in the amp. Leaving whatever distortion and colorings that the tube presents to be audible.
SO:
Here are the things for us to decide.
Changing the chip is an option.
Increasing the complexity is also an option (ie; Parallel or Bridged chips)
I would like to avoid the need for a PC Board, P to P wiring is much easier for Tube people to "swallow" in considering a project such as this.
I STILL do not understand why the gain of the chip portion can't be reduced to something on the order of 5 or 6? It seems that the chip is more likely to amplify noise from the case, powersupply and wiring, reducing the gain reduces this noise also.
Here us my opinion on the "Optimum" configuration that I would like to attemp. I think there is enough "expert" help available here to make a more complex project a "first time build" for me.
The choice of tube is irrelevant for now, but I would like to do a gain stage direct coupled to a "Split Load" inverter. (this will require a relatively HIGH B+ voltage for the Tube section as we need to run the gain stage at 1/3rd B+ in order to have a "balanced" load on the inverter.)
The Inverted and Non Inverted portions of the signal would then each have a "Parallel Pair" of chips. (this could be two LM 3886 or equivalent) or maybe just one LM4780?
Those "pairs" would then be "bridged" on the output only to the speaker load thru 0.1R resistors. No inductor, no Zobel just attach them and see what happens.
The pairs can be inverting or non-inverting as long as they are the SAME. The signal is already inverted for the lower pair.
Since the wiring might be a bit complex for P to P I can live with a piece of perfboard for the chips with copper and solder traces for the major connections.
OK, so there is what I want. Comments or suggestions? Please do not tell me why I SHOULDN'T do this, just what PROBLEMS I need to prepare for.
This choice is made to have this thing "LOOK" like a "Williamson" Push Pull Tube amplifier. Since one of the amps in my system is very much like this design it "blends" in well.
Yes, post 269 is an interview of one possible design. Its 50k input load, inverted LM3886. Its promising, has too much gain, and needs more work.
Yes, I do want a full range amp because I will use passive components for my own woofers.
Inverted LM3886 looks promising so far. Problems are congested and too much gain, but with very little coloration. Working on it.
The setup has 4 of LM3886 chips, and my plans are to interview 4 different designs, keep the best one or two, recycle the rest, until there are 4 good designs to present for review.
Possible solution with TDA7294 or a paralleled LM3886, as both of these options CAN have less coloration. I don't know yet. But these options are "more responsible design" than a solo LM3886 into a 4 ohm load.
With 4 ohm speakers, the options are:
Solo
Paralleled
Paralleled + Bridged
Probably shouldn't p2p TDA7294. I've done it, but its more like artwork. lol!
There's usually 2 sorts of op amps, "unity stable" and "uncompensated" and these big power op amps are the latter. They need some gain.
If you can think up a very simple discrete that also uses a unity stable op amp as a driver, then you can have a low gain p2p amplifier.
Parallel pairs (or trios) of LM3886 with a tube driver? I, personally, think that this is something that should be done. No complaints. Its commonly done with DRV134 op amp. So we're replacing that with a tube? Sure.
It won't be boring.
EDIT: At double-power (like you described), the gain problem is reduced somewhat.
I like this.
And, thank you for the adventure.
Last edited:
Hi,
The chip may start to oscillate if the gain is lower than 10. I have built an LM3886 amp for speaker testing and set the gain to 10 exactly without any problem.
It might be worth it to give a lower gain setting a try, just P2P an LM3886 for a quick test - attaching a Zobel will give a fast response on whether it is oscillating - the resistor of the Zobel will get hot. I may give this a try later.
Thanks man!!!
While you're at it, can you try inverted LM3886? There is a noticeable and potentially useful difference in frequency response.
Both Nick Whetstone (Decibel Dungeon) and I do seem to prefer inverting mode. This is more pleasant and may avoid the baxandall.
No time like the present: I changed the gain in one of my LM3886 amp boards to nearly 5 first with no signs of instability and then to 4 with the same result: no sign of instability.
I used the recommended schematic from National and there is a Zobel and output inductor in place.
Hope this helps.
I used the recommended schematic from National and there is a Zobel and output inductor in place.
Hope this helps.
Hey,
Here some info on the 4780,
mira
I think it is the Best Option. We can "Parallel" the two sections. Then we can use those sections either way. The desing of the "paralleled" 4870 is either used 2x for Stereo or 4x for stereo Push Pull (Bridged) in Tubes we call it Push Pull because one is Pushing and the other Pulling (due to the inverted signal) the nice part about using the tube to invert the signal is the TWO Chip sections can be IDENTICAL avoiding Offset problems and distortion.
Daniel,
Which tubes do you have TWO of?
Here some info on the 4780,
mira
I think it is the Best Option. We can "Parallel" the two sections. Then we can use those sections either way. The desing of the "paralleled" 4870 is either used 2x for Stereo or 4x for stereo Push Pull (Bridged) in Tubes we call it Push Pull because one is Pushing and the other Pulling (due to the inverted signal) the nice part about using the tube to invert the signal is the TWO Chip sections can be IDENTICAL avoiding Offset problems and distortion.
Daniel,
Which tubes do you have TWO of?
YOU ARE THE MAN!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Let's go with FOUR!
Since the 4780 is just dual 3886's we can use that as a starting point.
Thank you.
It's important to keep in mind that this worked for me and may not for another implementation (such as P2P or perf board) . My board layout is very compact and I used my regulated lab power supply to run it. Better to test beforehand before committing.
FWIW the LM4780 has smaller pins and they are closer together, making P2P much harder to do.
My layout:
I changed R8 to make the gain lower.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.