Thank you for your comments.
I'm still trying to figure out what Pacificblue was saying about how to change my power supply to make it better a few posts ago. Leave the 10KuF where they are on the power supply boards, get rid of the snubber resistor and cap, and add another 1KuF at Cs on the amplifier boards?
Any comment on my question from post #98?
I'm still trying to figure out what Pacificblue was saying about how to change my power supply to make it better a few posts ago. Leave the 10KuF where they are on the power supply boards, get rid of the snubber resistor and cap, and add another 1KuF at Cs on the amplifier boards?
Any comment on my question from post #98?
I'd rather have PacificBlue speak for himself, so instead, I'll quote some compatible designs:
Power board caps:
The 10,000uF caps are seen at the power board in the majority of audio amplifiers, in most practical examples. These are so very large that they're not really in range to have much of an effect on the audio band. They're for clean power, and they work for clarity.
Power board snubbers:
Chipamp.com's particular value of power board snubbers read as being specific to the case of an interconnect cable between power supply enclosure and amplifier enclosure, as seen on CarlosFM's personal amplifier in the year 2005. Is this compatible with your application or are these filters specific to somebody else's application? Perhaps PacificBlue and AndrewT can tell us.
Amplifier board power caps:
The 100uF caps at the amplifier boards are IN RANGE to affect the audio band. Adjust that value to larger! 100uF on LM3886 makes for some disproportionately LOUD UPPER MIDS that can out-shout your high treble.
*As shown on AN1192, the minimum value is 470uF, although most practical examples go larger.
**Amplifier board power caps need to be high quality--most practical examples for non-inverting Natsemi, use Panasonic caps.
Voicing a stable amplifier. . . or not?:
I have to ask if your amplifiers are still lacking Part "Cc" from the LM3886 datasheet? I have to ask, because for non-inverting LM3886, validity is quite literally installed with datasheet part Cc, the stabilizer cap. And, in the absence of measurement going without Cc will make it futile to voice the amp. Please add Cc, or else there *could be* great difficulty. Practical examples for your application include 100pF.
Temperature measurement as an indicator of quality:
Please measure average heatsink temperatures (in similar operating conditions and similar room temperatures) before and after any circuit change on average heatsink temperature, as an indicator of amplifier health.
Power board caps:
The 10,000uF caps are seen at the power board in the majority of audio amplifiers, in most practical examples. These are so very large that they're not really in range to have much of an effect on the audio band. They're for clean power, and they work for clarity.
Power board snubbers:
Chipamp.com's particular value of power board snubbers read as being specific to the case of an interconnect cable between power supply enclosure and amplifier enclosure, as seen on CarlosFM's personal amplifier in the year 2005. Is this compatible with your application or are these filters specific to somebody else's application? Perhaps PacificBlue and AndrewT can tell us.
Amplifier board power caps:
The 100uF caps at the amplifier boards are IN RANGE to affect the audio band. Adjust that value to larger! 100uF on LM3886 makes for some disproportionately LOUD UPPER MIDS that can out-shout your high treble.
*As shown on AN1192, the minimum value is 470uF, although most practical examples go larger.
**Amplifier board power caps need to be high quality--most practical examples for non-inverting Natsemi, use Panasonic caps.
Voicing a stable amplifier. . . or not?:
I have to ask if your amplifiers are still lacking Part "Cc" from the LM3886 datasheet? I have to ask, because for non-inverting LM3886, validity is quite literally installed with datasheet part Cc, the stabilizer cap. And, in the absence of measurement going without Cc will make it futile to voice the amp. Please add Cc, or else there *could be* great difficulty. Practical examples for your application include 100pF.
Temperature measurement as an indicator of quality:
Please measure average heatsink temperatures (in similar operating conditions and similar room temperatures) before and after any circuit change on average heatsink temperature, as an indicator of amplifier health.
Thanks for all your input. I actually have not made any mods for a while. I am not near any decent electronics supply, and every time I order from Mouser I get the shipping fee. So I've been holding off until my plan is clear. Right now I am running the chipamp boards with the snubberized PS. I have a tiny cap across the RCAs, and a 1uF cap after the pot. I removed R1, which is the 1K series resistor. For a while I had a 330 pF cap across R2, but I removed that with my last mod. In any case, I am soaking up all the advice you all are kindly offering, and eventually I'll order some stuff and try it out!
Nearly equivalent. The differences depend on the length and width of the traces, but are probably more theoretical than effectively audible or measurable. As Andrew suggested, you will have to listen to the differences yourself to find out whether you can hear any at all and how much they are worth to you, if you can hear some.
No, it improves the amp's ability to deliver very deep and strong bass and to maintain a given power level for longer without sagging. The improved bass may mask some of the mids and highs which may give the impression that there are less of them, but it will be more similar to the real thing.
To give an idea of the effect, set it in relation to speaker cable. 3 m of 2,5 mm² cable will have a total resistance flow and return of ~42 mOhms. 1.000 µF of capacitance will have the same impedance at ~3,8 kHz, if you don't count the ESR. 10.000 µF of capacitance at ~380 Hz and 100.000 µF at ~38 Hz. Their impedances rise below those frequencies and fall above them. And they are effectively as much in series to the speaker as the cable, even if NFB may attenuate the power supply impedance's effect somewhat.
Take a look at the reference design in the LM4780's datasheet. You can see how the capacitors are becoming smaller on the PCB as they are placed nearer to the LM's power supply pins. The biggest capacitance is off board in the power supply.
I was just looking at the datasheet you mentioned, and I noticed figure 5 on page 7, labeled Special Audio Amplifier Application Circuit. This seems to be exactly the circuit that you have been recommended I put together! I don't know why they didn't include such a diagram in the datasheet for the LM3886.
I'm thinking I'll get rid of the 100uF and 0.1uF on top of the board, collectively Cs now, and put a 1000uF one in their place.Then I'll get 10uF electrolytic caps, and use my existing 0.1uF polypropylene film cap, then attach both of the small ones under the board to the leads of the larger one. Seems this should work if my diagrams of caps in parallel are sonically equivalent!
What kind of inductor is the 0.7uH in the Thiele network? What wattage rating should the parallel resistor have there? Seems like 1/4 watt should be OK. I was thinking I'd attach those two things to my output jacks.
I'm thinking I'll get rid of the 100uF and 0.1uF on top of the board, collectively Cs now, and put a 1000uF one in their place.Then I'll get 10uF electrolytic caps, and use my existing 0.1uF polypropylene film cap, then attach both of the small ones under the board to the leads of the larger one. Seems this should work if my diagrams of caps in parallel are sonically equivalent!
What kind of inductor is the 0.7uH in the Thiele network? What wattage rating should the parallel resistor have there? Seems like 1/4 watt should be OK. I was thinking I'd attach those two things to my output jacks.
Hi,
no, don't attach a low esr film or ceramic caps to the leads of a low ESR electrolytic cap. This may result in PSU oscillation.
The very smallest (lowest ESL) cap is fitted direct to the power pins of the chip amp. This is on top of the pins, but there is not much room there. Some builders do this.
The next closest is the underside of the PCB attached to the same pads as the power pins.
The third best alternative is on top of the PCB as physically close to the power pins AND using the shortest possible trace lengths to joins the +ve cap and the -ve cap and the +ve power pin and the -ve power pin together in one short string.
The fourth best is two caps alongside the chipamp and find any route that is shortish to connect to the power pins. With a very short cap to cap trace.
The fifth best is two caps near the chip amp with shortish traces to the power pins but a long trace between the caps.
The sixth best is a waste of small caps.
A low ESR electrolytic is probably going to be located as the fourth best described above.
The main power caps, those exceeding 4700uF, are optional. I never miss them out. Some designers and Builders claim these big remote caps ruin the sound quality of the amplifier. You can be the only judge of that.
no, don't attach a low esr film or ceramic caps to the leads of a low ESR electrolytic cap. This may result in PSU oscillation.
The very smallest (lowest ESL) cap is fitted direct to the power pins of the chip amp. This is on top of the pins, but there is not much room there. Some builders do this.
The next closest is the underside of the PCB attached to the same pads as the power pins.
The third best alternative is on top of the PCB as physically close to the power pins AND using the shortest possible trace lengths to joins the +ve cap and the -ve cap and the +ve power pin and the -ve power pin together in one short string.
The fourth best is two caps alongside the chipamp and find any route that is shortish to connect to the power pins. With a very short cap to cap trace.
The fifth best is two caps near the chip amp with shortish traces to the power pins but a long trace between the caps.
The sixth best is a waste of small caps.
A low ESR electrolytic is probably going to be located as the fourth best described above.
The main power caps, those exceeding 4700uF, are optional. I never miss them out. Some designers and Builders claim these big remote caps ruin the sound quality of the amplifier. You can be the only judge of that.
Concur! This is application specific.
Caps:
Some practical data covers it somewhat. These are all Non-Inverting examples with split rail power supplies. A separate power supply board is assumed for the following real examples, but not otherwise mentioned. All of the capacitors listed below are upon the amplifier board of their respective amplifiers.
STK465 (and similar), 100uF
TDA7294, 220uF
Sure Electronics 100uF//100uF//100uF
LM1875, 330uF
LM3886, 470uF
LM3875, 1000uF
The amplifiers listed are from dullest (stk) to brightest (lm3875), and that little table, above, as shown, is compensating. Capacitor size mentioned is approximate minimum practical value from real life examples.
From several similar sounding results, its usually good to choose the one that causes the coolest temps because that choice will be clearer, as is the case with the Sure Electronics paralleled smaller caps example. However, that is likewise application specific, so make use of the temperature probe.
-
Grounding effects:
If the speaker return point is closer to the potentiometer then that's brighter dynamics with less level frequency response (typical of bus ground) or of the speaker return point is closer to the power board, then that's duller dynamics with a more level frequency response (typical of star ground).
One might think that the popular kits, with their speaker return on amplifier boards are having trouble with "sound of" NFB cap Because of grounding arrangements. If using an NFB cap, it might be good to locate power star ground with speaker return, both at the power supply board. Likewise, it might be good to use either a groundlift resistor or an instrumentation amplifier's input topology to create a signal star ground.
Your inductor//resistor question:
The inductor for the Thiele network can be determined by an on-line inductor calculator (or 10 turns around a pen lid), and the inductor is sized to drop the load on frequencies far higher than the audio band. The Resistor needs to be at least as sturdy as that which is used in your zobel. Typically, a 2w or 3w model is used. Honestly, you're right that the resistor doesn't bear much load at all; however, sturdy resistors are used for longevity.
Frequency response by ear:
Check frequency response by ear via employing Rod Elliot's Headphone Adapter: Headphone Adapter for Power Amplifiers along with Grado's SR80 or an economical substitute from Koss, like this bit of junk with its remarkably level frequency response and absence of pandering: Koss KTXPRO1. <---Link These are inexpensive and flat, but they don't otherwise do any favors. That's great for testing, albeit not luxurious. If you think that the absence of luxury might distract you then, spend for the Grado SR80. Check for DC offset before connecting headphones.
Frequency response by computer:
Check frequency response with computer by enclosing the amplifier within the loop of free RMAA software and then observe the frequency response chart. I export my chart to photo software and re-scale it disproportionately 10 times taller. RightMark Audio Analyzer. Products. Audio Rightmark
You can compare the loopback cable results with and without the amplifier patched in. Check for input caps on computer and/or DC offset on amplifier before connecting computer.
Paralleling/choosing caps:
When paralleling caps, be prepared to or try a couple of different models of cap. When trying different models of cap, try not more than 5 models, else either wrong cap value or wrong filter value is indicated. A generally useful smaller cap is the 470nF (0.47uF) Nichicon ES, which can assist 220uF and nearby sizes. I use that for the Via Tremor sound card output, to gently increase the upper treble and the resolution.
Speeding design with MonoBlock:
For fine tuning, MonoBlocks may be more convenient, since you can A-B compare and cyclically improve the loser in that competition until it beats the winner. Personally, I do have a small collection of those, because when there's an excellent development, I save it, unaltered, for a point of reference. My favorite is the TDA7294 dual mono (2 monoblocs in one enclosure, with one potentiometer), has lasted these many years, thanks to AndrewT.
Filter calculator:
Calculators give you the figure that has already taken effect, which in some cases isn't exactly what you want. General calculators for filters give you a -3db figure (or greater); however, if you're after a flat response, you'll need to adjust that filter by at least 200% so that a shunt filter (like zobel) doesn't cut into desired bandwidth or a pass filter (like NFB cap) doesn't omit desired bandwidth. Textbook approaches can be wonderful starting places, but just don't get yourself on the wrong side of a long taper filter.
The simple RMAA test can spot some audio filter errors easily.
The inductor for the Thiele network can be determined by an on-line inductor calculator (or 10 turns around a pen lid), and the inductor is sized to drop the load on frequencies far higher than the audio band. The Resistor needs to be at least as sturdy as that which is used in your zobel. Typically, a 2w or 3w model is used. Honestly, you're right that the resistor doesn't bear much load at all; however, sturdy resistors are used for longevity.
Frequency response by ear:
Check frequency response by ear via employing Rod Elliot's Headphone Adapter: Headphone Adapter for Power Amplifiers along with Grado's SR80 or an economical substitute from Koss, like this bit of junk with its remarkably level frequency response and absence of pandering: Koss KTXPRO1. <---Link These are inexpensive and flat, but they don't otherwise do any favors. That's great for testing, albeit not luxurious. If you think that the absence of luxury might distract you then, spend for the Grado SR80. Check for DC offset before connecting headphones.
Frequency response by computer:
Check frequency response with computer by enclosing the amplifier within the loop of free RMAA software and then observe the frequency response chart. I export my chart to photo software and re-scale it disproportionately 10 times taller. RightMark Audio Analyzer. Products. Audio Rightmark
You can compare the loopback cable results with and without the amplifier patched in. Check for input caps on computer and/or DC offset on amplifier before connecting computer.
Paralleling/choosing caps:
When paralleling caps, be prepared to or try a couple of different models of cap. When trying different models of cap, try not more than 5 models, else either wrong cap value or wrong filter value is indicated. A generally useful smaller cap is the 470nF (0.47uF) Nichicon ES, which can assist 220uF and nearby sizes. I use that for the Via Tremor sound card output, to gently increase the upper treble and the resolution.
Speeding design with MonoBlock:
For fine tuning, MonoBlocks may be more convenient, since you can A-B compare and cyclically improve the loser in that competition until it beats the winner. Personally, I do have a small collection of those, because when there's an excellent development, I save it, unaltered, for a point of reference. My favorite is the TDA7294 dual mono (2 monoblocs in one enclosure, with one potentiometer), has lasted these many years, thanks to AndrewT.
Filter calculator:
Calculators give you the figure that has already taken effect, which in some cases isn't exactly what you want. General calculators for filters give you a -3db figure (or greater); however, if you're after a flat response, you'll need to adjust that filter by at least 200% so that a shunt filter (like zobel) doesn't cut into desired bandwidth or a pass filter (like NFB cap) doesn't omit desired bandwidth. Textbook approaches can be wonderful starting places, but just don't get yourself on the wrong side of a long taper filter.
The simple RMAA test can spot some audio filter errors easily.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.