You're making stuff up there pal. I could give you some advice about being more easy going but I'm not sure you'll be interested. If you are, PM me to keep it off the thread.
No dispute from me there.
And why do you think they don't stop with two-tone IMD?
I've not formed that interpretation but if you tell me what you're basing it on I'll revisit the text.
Peter Daniel of Audo Sector uses ZERO micro Farads in his PSU !
And recommends that to all his customers.
But as far as I can see he developed his design/build philosophy while his speakers were 96dB/W, 4 ohms.
Not what most of his customers will be using.
Thanks for steering the conversation back on topic.
I do have one question with this approach: How does the amplifier get to deliver any power to the load when the incoming voltage to the rectifier crosses 0 V?
Tom
Peter Daniel version has 0 uF capacitance in PSU (or 10uF X 2 for each channel, as he ships the kit that way), but has 1,500 uF x 2 for each channel attached right next to the output chip. Those work as smoothing + electrical energy reservoir, I think.
That does make sense. If you go that route, make sure that the connections from the transformer to the amplifier board are short. The charging pulses that flow from the rectifier to the supply caps are high amplitude, short duration, so high bandwidth pulses. With long wires, you'll be radiating all sorts of energy into nearby circuits.
With a large smoothing cap on a separate power supply board, you do get a bit more flexibility on the physical layout of the circuit.
When it comes to smoothing capacitance, more is better as this results in lower supply ripple (both mains ripple and load current induced ripple). It wouldn't surprise me if someone prefers the colouration you get with low supply capacitance. I'm just not one of them as I prefer my semiconductor amps to be transparent. Each to his own.
In my MOD86, I have 2x1000 uF on the board about 20-25 mm from the chip. Then a 22 uF and a 1 uF X7R ceramic right at the IC pin. I forget if I've linked to it already, but you can read about my approach to decoupling here: Taming the LM3886 - Supply Decoupling.
Tom
Last edited:
Hey you guys 
Hey let me weigh in here.
The OP has probably 3 of the top ten respected amplifier experts here....
Let's not argue huh?
I instinctively feel that a flat THD vs Freqs is important, in combination with a low enough THD level.
More importantly for OP, I also have the AS LM4780 board here.
The 10uF are local decoupling to the IC
The on board smoothing is the ~1500uF
I believe it is mentioned on the site that you can also add off board reservoir capacitance.
So then you have the fabled three tier PSU holy grail
a certain member will know I am repeating what they taught me!
To be clear, the AS rectifier boards are useless to me. The snubberised PSU option is interesting, but ultimately I was lanning to use the on board spots of Caps and build a simple unregulated PSU, 35A bridge and upwards of 0.033uF per rail, separate PSU per board.
I may rethink that and try to cobble together servo regulated one now...
Hey let me weigh in here.
The OP has probably 3 of the top ten respected amplifier experts here....
Let's not argue huh?
I instinctively feel that a flat THD vs Freqs is important, in combination with a low enough THD level.
More importantly for OP, I also have the AS LM4780 board here.
The 10uF are local decoupling to the IC
The on board smoothing is the ~1500uF
I believe it is mentioned on the site that you can also add off board reservoir capacitance.
So then you have the fabled three tier PSU holy grail
a certain member will know I am repeating what they taught me!
To be clear, the AS rectifier boards are useless to me. The snubberised PSU option is interesting, but ultimately I was lanning to use the on board spots of Caps and build a simple unregulated PSU, 35A bridge and upwards of 0.033uF per rail, separate PSU per board.
I may rethink that and try to cobble together servo regulated one now...
Last edited:
You mean, you actually want a reasonable signal-to-noise ratio in this thread? No! Say it ain't so. This is the Internet, after all...
I do admit to feeling a bit guilty for plugging up the thread. My apologies to the OP.
That jives with my subjective experience + objective measurements. The low THD part is also supported by Belcher (1978). See figure 8 of the paper linked to in Post #37.
According to Belcher (1978), the THD isn't as reliable an indicator of good sound quality as the multi-tone test, but it is an indicator nonetheless. See figures 7, 8 of the paper.
Interestingly, I've seen figures of -60 to -50 dB THD floating around DIY Audio for a while as the audible threshold of THD. That jives pretty well with Belcher's research. It would be interesting to dig a little deeper and see what the current state of the research is. I'll look into that.
Meanwhile, I'll continue to design world class amps. Hope you guys don't mind...
A Class AB amp is a dynamic, high-current load. The supply will be as complicated as the amplifier. Then again, as we've touched on several times in this thread, if you can't turn your amplifier project into a power supply project, you aren't doing it right.
Tom
Let me correct myself
The PSU board is a snubberised board I think (CarlsoFM?) and I wasn't planning to use mine.
The only thing I dislike is that the rail grounds remain separate until they reach the capacitors on the amp board.
I'd imagine if you keep the boards attached then the shirt run isn't a problem.
But if you build a bigger reservoir PSU it makes it easy to have a ground loop.
The PSU board is a snubberised board I think (CarlsoFM?) and I wasn't planning to use mine.
The only thing I dislike is that the rail grounds remain separate until they reach the capacitors on the amp board.
I'd imagine if you keep the boards attached then the shirt run isn't a problem.
But if you build a bigger reservoir PSU it makes it easy to have a ground loop.
Snubber or no snubber, the charging pulses are the same. The snubber just gets rid of the little ringing you get when the diodes turn off.
You could join the two at the transformer, effectively creating a centre tap.
You'd actually have the same ground loop in a build with just one board.
I think the "ground loop optimizers" are sometimes missing the big picture. In an effort to keep all ground traces separate, they end up dramatically increasing the ground impedance, which makes the induced error much higher. Instead, I (and many others) suggest throwing copper at the problem. Use a ground plane. Do maintain a quiet ground to use as a reference. In the ideal world, both the power ground with the nasty currents in it and the reference ground with, ideally, no current in it would be large, unbroken planes joined in exactly one spot: The GND connection of the speaker output connector.
Tom
It's fine. I am enjoying the discussion going on here, and learning a lot.
To me, it's more like steamed discussion is better than no interest.
+1
Having built a My_ref Rev C, I realize that I'm biased, but I don't understand why the My_ref family gets largely forgotten in discussions like this. It is one of the best designs available and was developed in this forum.
The MyRef is a Howland current pump. You can read about the inherent issues with that topology in Bob Pease's application note: http://www.ti.com/lit/an/snoa474a/snoa474a.pdf
Tom
Did you read Penasa's description of his circuit? (attached)
BTW the Howland current pump used is the 'improved' version and the project specify 0.1% resistors for the pump so those 'problems' are mostly 'fixed'.
In both the My_Evo and my variation the My_Ref circuit achieve distortion level around 0.00011% (THD), not so bad for a flawed circuit...
I don't know to debug the schematics, but what I see different about my ref schematic is that the lm3886 is used in inverted configuration vs regular configuration which is used in most of the other gainclones. And I am sure that I found somewhere on ti.com that inverted configuration will have less distortions vs non inverted one.
Not an universal truth but usually most opamps have less distortion when used as inverting amps.
Tom, who is also an opamp designer, could confirm, I guess.
Nevertheless the My_Ref is not an inverting amp, being a composite of two inverting amps.
What makes the My_Ref so particular is the use of Howland Current Pump, which masks almost completely LM3886's character and distortion and the use of different types of feedback/feedforward.
Also the LM318, which characterizes most the amp sound, is a very peculiar opamp with access to its inner circuit, exploited by Penasa with a second feedback on it.
The My_Ref is not as simple as it could seem at first glance...The designer worked on it several years before donating it to the community!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.