I use 7 or 8 turns on 15mm diam for ~0.7uH
Using a bigger diam former will give more inductance.
Expect nearly double for a 28mm former.
So reducing to 5.5 Turns on 28mm is probably pretty close to 1uH.
Wheelers formula:
L= 7.87 * M² * N² / {3M + 9B + 10C} when fudged gives ~1000nH, Maybe a bit less, I'm getting 922nH.
M, B & C are mm N = Turns
M= mean diam, B = spool width, C = Turns height.
I think Purdic's formula applies to single layer coils.
Using a bigger diam former will give more inductance.
Expect nearly double for a 28mm former.
So reducing to 5.5 Turns on 28mm is probably pretty close to 1uH.
Wheelers formula:
L= 7.87 * M² * N² / {3M + 9B + 10C} when fudged gives ~1000nH, Maybe a bit less, I'm getting 922nH.
M, B & C are mm N = Turns
M= mean diam, B = spool width, C = Turns height.
I think Purdic's formula applies to single layer coils.
Last edited:
This calculator seems to work well:
Air Cored Inductor Calculator – M0UKD – Amateur Radio Blog
I've also used this calculator. I measured the resulting inductor on an impedance analyzer and it came in right on the money. I do prefer metric units, though.
Air Core Inductor Coil Inductance Calculator
The 12.2 mΩ is not important compared to the 965 Ω. At 20 kHz, the inductance can add quite a bit to the 965 Ω, though. Check out this post for an example: LM3886 P2P vs PCB (with data) Post #352. Do note that in my Modulus-86, I'm pushing the measurement limit of the APx525 audio analyzer, so the degradation in THD+N is most likely because the added inductance in the feedback network causes enough of a gain shift that the fundamental amplitude drops slightly, causing a worse THD+N measurement.
Tom
Air Cored Inductor Calculator – M0UKD – Amateur Radio Blog
I've also used this calculator. I measured the resulting inductor on an impedance analyzer and it came in right on the money. I do prefer metric units, though.
Air Core Inductor Coil Inductance Calculator
The 12.2 mΩ is not important compared to the 965 Ω. At 20 kHz, the inductance can add quite a bit to the 965 Ω, though. Check out this post for an example: LM3886 P2P vs PCB (with data) Post #352. Do note that in my Modulus-86, I'm pushing the measurement limit of the APx525 audio analyzer, so the degradation in THD+N is most likely because the added inductance in the feedback network causes enough of a gain shift that the fundamental amplitude drops slightly, causing a worse THD+N measurement.
Tom
I filled in all the trace resistances.
(for those who would gain from a groundplane)
and i made a calculation for the in+ resistor based on a 5 mm trace to the nearest GND plain. The R111 and L101 would then replace R1 and R11.
Is there anybody whose willing to make a calculation?
I'm eager to go on with the build itself and prepare for first measurements.
I don't have any tools except an 8Bit USB scope and a pretty decent soundcard: Xonar STXII. So i have to figure out if i can use the STXII or find a local distortion measurement device.
(for those who would gain from a groundplane)
An externally hosted image should be here but it was not working when we last tested it.
and i made a calculation for the in+ resistor based on a 5 mm trace to the nearest GND plain. The R111 and L101 would then replace R1 and R11.
Is there anybody whose willing to make a calculation?
I'm eager to go on with the build itself and prepare for first measurements.
I don't have any tools except an 8Bit USB scope and a pretty decent soundcard: Xonar STXII. So i have to figure out if i can use the STXII or find a local distortion measurement device.
I have a weird/odd idea for extra decoupling caps on the power pins.
What if i use SMD X7R chips and solder them strait-down onto the power pins.
and then solder the other side to a copper strip. (underneath the pcb)
Or use the Via's i placed at the powerlines to place a leaded multilayer.
I could even created a power plane 10mm underneath the PCB.
I think a saw a 2mm thick copper plate laying around at work.
Use the waterjet to cut and drill the holes.
What if i use SMD X7R chips and solder them strait-down onto the power pins.
and then solder the other side to a copper strip. (underneath the pcb)
Or use the Via's i placed at the powerlines to place a leaded multilayer.
I could even created a power plane 10mm underneath the PCB.
I think a saw a 2mm thick copper plate laying around at work.
Use the waterjet to cut and drill the holes.
I'm not sure which calculation you need. It looks like you have your simulation schematic drawn up. Just set up an AC simulation and click Run. Look at the output of the LM3886.
Tom
1 PCB is ready for testing.
Found some old parts for the heatsinks.
The heatsink for the rectifier i like the most.
Now waiting for a 6.3mm jack to be able to use the inputs of the soundcard.
An externally hosted image should be here but it was not working when we last tested it.
Found some old parts for the heatsinks.
The heatsink for the rectifier i like the most.
Now waiting for a 6.3mm jack to be able to use the inputs of the soundcard.
Save that CPU heatsink for a duty that requires it.
A ClassA bridge rectifier gets hot.
That cooler is twice as big as what two ClassA bridge rectifiers would need.
A ClassAB bridge rectifier, generally manages without a heatsink if the metal casing/backplate is exposed to air. Or just bolt it to the metal floor of the enclosure.
Get rid of the untwisted wiring.
Replace the three braided wires with a twisted triplet.
Replace the spaced out wires with twisted pairs.
A ClassA bridge rectifier gets hot.
That cooler is twice as big as what two ClassA bridge rectifiers would need.
A ClassAB bridge rectifier, generally manages without a heatsink if the metal casing/backplate is exposed to air. Or just bolt it to the metal floor of the enclosure.
Get rid of the untwisted wiring.
Replace the three braided wires with a twisted triplet.
Replace the spaced out wires with twisted pairs.
Last edited:
I am aware it's overkill, it was scrap metal laying around. I use it because i have no metal plate to mount it on. The bottom plate still needs to be cut.
So braiding is not as effective as twisting?
This is just the first lay-out i'm gonna try.
What about twisting VCC-gnd and Vee-Gnd separately? I have a a double GND connector. It will form a small loop but would that be measurable?
I'm also thinking about putting resistors between the rectifier and PSU caps. I will lose power but but maybe i will get a cleaner VCC and VEE. The PSU caps can deliver 14A each and i was thinking about dampening the ripple current from the rectifier.
I could even go a step further and create a R-C-R-C filter, by putting a resistor between the PSU and the PCB.
The amplifier is intended for personal use, not disco level, so power-output is the least important feature.
Thanks for all the input It is much appreciated.
Most power amplifiers draw non constant current from the supply rails when there is output current.
The pulses in the +ve supply rail flow out to the speaker when the speaker's AC waveform is in the upper half. They return to the pSU via the zero volts rail.
The pulses in the -ve supply rail flow back from the speaker when the speaker's AC waveform is in the lower half. They flow out via the zero volts rail.
These two sets of pulses add up to the whole AC waveform passing through the speaker.
By combining them you create an audio frequency sinewave composite.
If you now combine the two supply rail composite with the speaker return which becomes the zero volts rail, you end up with a Flow and Return composite that exactly cancel at all audio frequencies. This emits no emi, if the loop area is near zero.
Twist the three wires of the dual polarity supply rails to form the no emi composite.
The pulses in the +ve supply rail flow out to the speaker when the speaker's AC waveform is in the upper half. They return to the pSU via the zero volts rail.
The pulses in the -ve supply rail flow back from the speaker when the speaker's AC waveform is in the lower half. They flow out via the zero volts rail.
These two sets of pulses add up to the whole AC waveform passing through the speaker.
By combining them you create an audio frequency sinewave composite.
If you now combine the two supply rail composite with the speaker return which becomes the zero volts rail, you end up with a Flow and Return composite that exactly cancel at all audio frequencies. This emits no emi, if the loop area is near zero.
Twist the three wires of the dual polarity supply rails to form the no emi composite.
I don't know/remember what size of amp you are planning, nor what impedance speakers you use.
A 100W into 8ohms rated amplifier will deliver transient current pulses exceeding 10A very regularly and may exceed 15Apk occasionally.
Higher power and/or lower impedance can easily demand transient currents exceeding 30Apk.
That's why we see 2pair and 3pair and 4pair output stages, to meet the transient demands.
I use at least a 2pair output stage for a 100W into 8ohms amplifier to meet that 15Apk demand.
Those peak currents are sourced from the capacitors. Not from the transformer.
Last edited:
That makes sence.
But the same appears true for the VDD-GND and VEE-GND setup.
The thing is, i have speaker wire in a twinax setup. What about keeping the twinax setup true from Capacitor poles to PCB.
those 2 pairs could be twisted together as well OR at least keep them parallel.
But the same appears true for the VDD-GND and VEE-GND setup.
The thing is, i have speaker wire in a twinax setup. What about keeping the twinax setup true from Capacitor poles to PCB.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
those 2 pairs could be twisted together as well OR at least keep them parallel.
The amp will drive 2x Mission 771 (8ohm) and 2x 760I (6ohm)
( i hope to find a second set of 771's)
It is all in a near field setup. So i don't need high power output!
Every speaker will get it's own LM4780 chip with paralleled outputs.
The 14A is the ripple max at 105° at 100KHz. ESR 12mOhm @100Hz 10mOhm at 100KHz. Kemet ALS30A223KE063
On the PCB are 2x 4700 50V caps (link)
2 47Uf 50V polymer caps (link)
and there also come 2 4u7F multilayer chip 50V X7R. (link)
If necessary there is room reserved in the cabinet for a second set of 22000uF PSU caps.
Finally the plates are cut. It was scrap metal so heavily damaged.
But a sanding machine should be able to fix that.
Does someone have a good way to clean up scratched aluminium?
I can start building
PSU, softstarter and DC blocker are tested and ok.
I'm now drilling holes into the heatsinks. Simon please do'nt f**k up
But a sanding machine should be able to fix that.
Does someone have a good way to clean up scratched aluminium?
I can start building
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
PSU, softstarter and DC blocker are tested and ok.
I'm now drilling holes into the heatsinks. Simon please do'nt f**k up
Try laying it on a sheet of 180 or 220 grit "wet or dry" paper. This should remove the deeper scratches.
Then progressively go through 400 grit via 600grit to 1000grit
This can give a very nice "brushed finish" look, if you can keep the sanding movement going in a straight line as you pass over the wet or dry sheet.
Then progressively go through 400 grit via 600grit to 1000grit
This can give a very nice "brushed finish" look, if you can keep the sanding movement going in a straight line as you pass over the wet or dry sheet.
Thanx andrew
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
I'm revisiting my PCB again and came up with some questions.
Since i use both amps in the LM4780 parallel;
Can i use only 1 100nF-2.7ohm filter on the output and 1 50pF-28K filter from output to the input?
if 1 of the amps starts behaving strange there will be a huge difference in the outputs which will stress the normal operating amp. But my main concern is that the 50pF cap which should lower the gain for high frequencies will never act the same as the other channel.
So what about trying to compensate both paralleled amps with 2 shared filters/compensation networks instead of 2 filters/networks each?
Since i use both amps in the LM4780 parallel;
Can i use only 1 100nF-2.7ohm filter on the output and 1 50pF-28K filter from output to the input?
if 1 of the amps starts behaving strange there will be a huge difference in the outputs which will stress the normal operating amp. But my main concern is that the 50pF cap which should lower the gain for high frequencies will never act the same as the other channel.
So what about trying to compensate both paralleled amps with 2 shared filters/compensation networks instead of 2 filters/networks each?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.