I made some measurements on my circuit today. Unfortunately, either the circuit is shockingly bad, or I'm getting half wave currents dirtying my earth. I suspect the latter due to the lash-up nature of the breadboard and supply wiring. Also due to the fact that adjusting the bias had zero effect on the measurement. An epic spray of harmonics was measured!
Do you have a schematic for it? I wonder if there's been any recent changes that make it more susceptible to power noise?
Wild guess: In this range of output power, the transformer's voltage output is so low that small noise could account for a more impressive percentage of the total power supply voltage. Power noise pollution and bridge rectifier noise are weak current sources that can cause extraneous rail variance. And, the noise itself can cause extra heat in the amplifier. Perhaps you'd want to snub the rectifier and/or use a CRC?
It's probably down to the fact that everything was half-assed together with long clip leads without much regard to anything 
I'm going to order a 20VA transformer to power the thing. Do I have to put out 9W from both channels at the same time?
Edit: Nevermind, I uprated it to a 50VA.
I'm going to order a 20VA transformer to power the thing. Do I have to put out 9W from both channels at the same time?
Edit: Nevermind, I uprated it to a 50VA.
Last edited:
For decently low ripple performance, 18W*3=54VA. The 50va is right.It's probably down to the fact that everything was half-assed together with long clip leads without much regard to anything
I'm going to order a 20VA transformer to power the thing. Do I have to put out 9W from both channels at the same time?
About a 2a center tap (13.5+13.5)*2a=54VA
Capacitance could be 4400u (2 paralleled 2200u) on single rail or approximately 10,000u per each rail for split rail. Those are sort of minimums for low ripple performance, but it could use something better. The capacitance looks big because the transformer is quite small and so is the amplifier, but the speakers didn't get smaller. If less capacitance is desired then a higher amperage transformer is needed. Also, bigger transformer plus good size capacitance could sound just like a bigger amp that isn't turned up quite as loud, instead of sounding only like a little amp. I think I'd use regs.
Last edited:
For decently low ripple performance, 18W*3=54VA. The 50va is right.
About a 2a center tap (13.5+13.5)*2a=54VA
Capacitance could be 4400u (2 paralleled 2200u) on single rail or approximately 10,000u per each rail for split rail. Those are sort of minimums for low ripple performance, but it could use something better. The capacitance looks big because the transformer is quite small and so is the amplifier, but the speakers didn't get smaller. If less capacitance is desired then a higher amperage transformer is needed. Also, bigger transformer plus good size capacitance could sound just like a bigger amp that isn't turned up quite as loud, instead of sounding only like a little amp. I think I'd use regs.
I nearly got it right, thought I ordered a 4700uF cap for each rail. It'll do for now! I'll make space on the power supply PCB for another pair.
One thing to do with long wires is to pinch the wire so it's parallel with itself and twist it. This way the magnetic fields cancel and it's no longer an inductive anttenna. It can be untwisted for longer connections.
I gave up on breadboards and now just make solder sculpture circuits. After all, get some gunk in the breadboard and one random connection goes open circuit, this can blow up components through no fault of the experimenter.
I gave up on breadboards and now just make solder sculpture circuits. After all, get some gunk in the breadboard and one random connection goes open circuit, this can blow up components through no fault of the experimenter.
Just managed to get (in simulation) 0.004% distortion at 20Khz with TIPs on the output and 18mA Iq. Big performance gains were had with buffering the VAS from the output stage (which isn't really a huge surprise).
This is with CFP outputs too. If I can get reality to reflect this I will be very happy.
This is with CFP outputs too. If I can get reality to reflect this I will be very happy.
That's amazingly good performance. 0.004% thd with a 9W amp is sort of like 0.0004% with a 90W amp.
I was thinking about your LTP and your power supply. Naturally, these two circuits are enemies. The LTP will gladly throw the baby out with the bath water if exposed to power noise. Power noise rejection is an excellent feature we hope never to use. But, actually we hope that instead of a huge workout, the power noise rejection is only needed for spurious noise. Therefore, I think you'd like a CRC. I think you can do it by adding 2200u and a series resistor (per each rail) directly at the bridge rectifier. This CRC will have your 4700u||4700u power reservoir full up with reasonably clean DC instead of harmonic distortion. I think that it would be good to turn (dis)harmonic noise into heat inside of the power box instead of heat inside of the mint tin.
P.S.
You can also try snubbing the bridge rectifier for about 1v less noise (that's impressive if seen as a percentage) and a somewhat noticeable percentage less heat. Using a CRC decreases the reliance on snubbing the bridge rectifier and could allow using approximately 4.7n or smaller lossy/cheap polyester dip caps to snub the rectifier very easily (its either that or real RC's).
OR, with faster didoes (such as the extremely quiet MR diodes), you can snub "before and after" the bridge rectifier with RC's across the secondaries and across the rails. I actually forgot how to do this one, but there's a couple of posts by PacificBlue that looked good. If I remember correctly, it was an equal-and-opposite sort of arrangement that makes DC, instead of connecting the noise to the 0v. I do remember that the parts count was quite low and that, if using fast/soft diodes its more about snubbing the transformer, not so much about snubbing diodes.
And, For mains side filtering, it should be fairly easy to extract the MOV and mains side filters from a discarded computer power supply, for free.
I was thinking about your LTP and your power supply. Naturally, these two circuits are enemies. The LTP will gladly throw the baby out with the bath water if exposed to power noise. Power noise rejection is an excellent feature we hope never to use. But, actually we hope that instead of a huge workout, the power noise rejection is only needed for spurious noise. Therefore, I think you'd like a CRC. I think you can do it by adding 2200u and a series resistor (per each rail) directly at the bridge rectifier. This CRC will have your 4700u||4700u power reservoir full up with reasonably clean DC instead of harmonic distortion. I think that it would be good to turn (dis)harmonic noise into heat inside of the power box instead of heat inside of the mint tin.
P.S.
You can also try snubbing the bridge rectifier for about 1v less noise (that's impressive if seen as a percentage) and a somewhat noticeable percentage less heat. Using a CRC decreases the reliance on snubbing the bridge rectifier and could allow using approximately 4.7n or smaller lossy/cheap polyester dip caps to snub the rectifier very easily (its either that or real RC's).
OR, with faster didoes (such as the extremely quiet MR diodes), you can snub "before and after" the bridge rectifier with RC's across the secondaries and across the rails. I actually forgot how to do this one, but there's a couple of posts by PacificBlue that looked good. If I remember correctly, it was an equal-and-opposite sort of arrangement that makes DC, instead of connecting the noise to the 0v. I do remember that the parts count was quite low and that, if using fast/soft diodes its more about snubbing the transformer, not so much about snubbing diodes.
And, For mains side filtering, it should be fairly easy to extract the MOV and mains side filters from a discarded computer power supply, for free.
Almost got the board together... just got to wangle my 2 input caps into a spot and i'll be ready to plug it in and test it.... Oh yes and I gotta find a heatsink that will fit, do the job and leave space for input, output and a volume pot!
Not sure if it will fit in a tin, but i'll be happy if it works.. Maybe if i just tap the caps down a bit with a hammer and twist the chip with a screwdriver it just squeeze in.
Not sure if it will fit in a tin, but i'll be happy if it works.. Maybe if i just tap the caps down a bit with a hammer and twist the chip with a screwdriver it just squeeze in.
Good going
I'm going the way over the weekend, so I wont be able to make my boards until early next week.My transformer arrived this morning, as well as a bridge rectifier and capacitors. I've always needed a +/- supply so this was a good excuse to build one. The transformer will happily fit in a PC power supply case. I have plenty of old ones!
I added a small value capacitor across the input of the amplifier - this improves its behaviour dramatically, now it's not having to dear with unreasonably fast rise times on the square waves from my sig gen. My earths are definitely getting dirtied up by the lash up wiring, but this will go away when I build it and earth it like a proper amplifier.
Well i'm surprised... It worked first time... I Have it in a watch tin at the moment... bit bigger than a mint/bakki tin but wanted to see it in A tin...
Its not a great sound, but that could be a number of things... crappy test speakers or it could be my power supply... 240 - 12v torroid at 3.5 Amp... got it running through a makeshift rectifier made up with 4 1n4001's and a 100uf cap. Giving me about 18v DC.
I have put RCA inputs and just got wires hanging for power and speaker output, but I might change to small jackplugs if it goes in a bakki tin.
Then I can start working on the making it pretty bit!
Its not a great sound, but that could be a number of things... crappy test speakers or it could be my power supply... 240 - 12v torroid at 3.5 Amp... got it running through a makeshift rectifier made up with 4 1n4001's and a 100uf cap. Giving me about 18v DC.
I have put RCA inputs and just got wires hanging for power and speaker output, but I might change to small jackplugs if it goes in a bakki tin.
Then I can start working on the making it pretty bit!
Partners to this cap could include picofareds capacitor parallel with the feedback resistor (for linearity), Vas compensations of different types (for pretty square wave), and output RC (for fine tuning) as well.
If you post some photos, perhaps someone will offer a helpful response on how to fit the amp into the enclosure. When posting photos, click the button marked "go advanced" and then scroll down to the button marked "Manage Attachments" and then you can post your photos.littlerick said:
A heat spreader bar in the tin can be used in thermal management. This can be made with aluminum bar stock from the hardware store (or aluminum scrap) and white thermal paste from the computer store.
The warmed air that goes up and out of the output vents at near the top, creates a miniscule vacuum force when it leaves and this is why cool air will rush into the bottom intake vents if the tin has some rubber feet giving clearance for the intake vents to work. It maybe be nice to have air intake vents near capacitors, to cool them, as long as a finger couldn't get onto the pins. Decorous ventilation is appreciated.
The input load on the schematic can be either potentiometer or resistors. Volume control is appreciated, but not required.
In the "simple elegance" approach popular with chip amplifiers, you can size your output caps to suit your speakers. For 8 ohm speakers the range is approximately 800u for miniature compact speakers with 4" woofers, to 4400u for full size speakers with 8" woofers, and your schematic shows the average size between those two examples. Those are all just estimates. A real speaker can react noticeably well to having chosen the right size cap for it.
Be sure to use non-shorting connection for speaker jacks. What works well is RCA's, binding posts, and retail looking plastic spring clip speaker jacks, because they don't short when you plug in the speaker.
However, because of violent explosion, 1/8" and 1/4" jacks won't do for either speaker or power connectors. I found this out already.
Kudos for getting it up and running so fast!littlerick said:
Before judging the sound, you might want at least a minimal power supply with something like 1n5405 or larger size diode bridge rectifier followed by a parallel pair of 2200u caps (4400u or more works fine for single rail), capacitance at the power supply.
Last edited:
My car chip happened to need a volume pot for lower distortion. I'll probably give that pot a slight automation in the form of a low parts count "overdo reducer," so now awaiting the mailbox for some germanium parts to arrive for a fun experiment. Unfortunately, the car chip has a locked gain setting, so the prospect is hit or miss because I can't fine tune the gain. There may be barely enough for 9 watts. I don't know yet. Personally, because I'm host, I'm not intending to win. But, I would like a fun and useful project.
Last edited:
Yes, we're still on. Some people are waiting for parts, some are busy with Christmas activities.
I need some input on an appropriate closing date, since I don't have an idea on that. Just like phase linear, we shouldn't go either too slow or too fast.
At this time, I do have news of one Blameless style discrete amp, one truck chip, and one car chip, all of which are looking good so far.
Another possible way to get low bias is to play a game of limbo (how low can you go) by under-volting a more powerful chip amp, and setting the gain lower to both match up with the smaller scale amp as well as reduce some of the "naf" consequences of lower bias.
And, of course New Class A equipped with auto bias and Class G power supply (makes AaG) is a possible way to get a fairly high power Class A into a mint tin. However, the newest high-efficiency linear tech is currently pushing phone antennas instead of speakers, efficient fet outputs have been hard to find and the amp designers are busy with Christmas activities.
I need some input on an appropriate closing date, since I don't have an idea on that. Just like phase linear, we shouldn't go either too slow or too fast.
At this time, I do have news of one Blameless style discrete amp, one truck chip, and one car chip, all of which are looking good so far.
Another possible way to get low bias is to play a game of limbo (how low can you go) by under-volting a more powerful chip amp, and setting the gain lower to both match up with the smaller scale amp as well as reduce some of the "naf" consequences of lower bias.
And, of course New Class A equipped with auto bias and Class G power supply (makes AaG) is a possible way to get a fairly high power Class A into a mint tin. However, the newest high-efficiency linear tech is currently pushing phone antennas instead of speakers, efficient fet outputs have been hard to find and the amp designers are busy with Christmas activities.
DMOSfets and Lateral Fets found!
Power MOSFET - Wikipedia, the free encyclopedia
Searching for lateral fets results in mostly obsolete devices because both the language and the packaging are obsolete. Today, these are called power mosfet and for an additional set of inconveniences, the model numbers are all changed, the tabs are cut off, and sometimes the center pin is also cut off. Nevertheless, they do still exist, they're plentiful, they're efficient and they're inexpensive. At low voltages, such as this thread, the power mosfet is extremely efficient. With Class AB, we could perhaps reach 20 watts per channel (max?), or even more with Class AaG.
Power MOSFET - Wikipedia, the free encyclopedia
Searching for lateral fets results in mostly obsolete devices because both the language and the packaging are obsolete. Today, these are called power mosfet and for an additional set of inconveniences, the model numbers are all changed, the tabs are cut off, and sometimes the center pin is also cut off. Nevertheless, they do still exist, they're plentiful, they're efficient and they're inexpensive. At low voltages, such as this thread, the power mosfet is extremely efficient. With Class AB, we could perhaps reach 20 watts per channel (max?), or even more with Class AaG.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.