Further fodder for our consumption:
http://www.audiophonics.fr/images2/6189/SMPS300R.pdf
I don't know that much about switch mode design so I'll get some expert advice on this but I don't like the look of the 100nF cap across 10R connecting system ground to mains ground. In my conventional supplies I just have 100R in this position . . . . ahh further down the datasheet it says that there is a link the bottom of the board that makes this optional - this is encouraging
Thx for your comments meanman but I should say that I now regard some of my input in the original thread as uninformed hogwash !
In particular, having now tried O/P chokes I find that they play a very valuabe role with this design by increasing stability in the feedback loop and thereby improving the subjective sound - Thx Hugh
But about very low noise PSUs my views remain unchanged
cheers
mike
First I wanna say something regarding the schematic of the SMPS300R that isn't the schematic of the one I bought,I have the SMPS300RE not totaly the same.It has a CLC filter at the output.Also the chassis earth is different.
Last edited:
Hi guys
Thank you for your great responses. (Apology for my delay responding - over the weekend I had a special guest)
Hugh,
so higher currents generate an increase in 2nd, 3rd and 4th orders?
When you said
“If you run 1A through the output stage, you will dissipate 2 x 36 = 72W from each channel, if your rail voltage holds up”
. . are you suggesting that with the standard PS, in some way, currents above a certain level(?) would not work so well?
In particular - Mike/ Hugh – for 36 V rails @ 0.75 A - does the PS need any modifications?
- a modest increase in the value of the PS caps?
- the resistor in the diode string on the bias generator?
- would a 500 R trimmer and 47 R resistor be changes to values, or need to be jumpered in?
Cheers
Thank you for your great responses. (Apology for my delay responding - over the weekend I had a special guest)
Hugh,
so higher currents generate an increase in 2nd, 3rd and 4th orders?
When you said
“If you run 1A through the output stage, you will dissipate 2 x 36 = 72W from each channel, if your rail voltage holds up”
. . are you suggesting that with the standard PS, in some way, currents above a certain level(?) would not work so well?
In particular - Mike/ Hugh – for 36 V rails @ 0.75 A - does the PS need any modifications?
- a modest increase in the value of the PS caps?
- the resistor in the diode string on the bias generator?
- would a 500 R trimmer and 47 R resistor be changes to values, or need to be jumpered in?
Cheers
PS chokes
<mikelm>
In particular, having now tried O/P chokes I find that they play a very valuable role with this design by increasing stability in the feedback loop and thereby improving the subjective sound - Thx Hugh
Hi Mike
When you have time (If you haven't already) could you describe your PS changes/ implementation - and any other changes you've found worthwhile
Cheers
<mikelm>
In particular, having now tried O/P chokes I find that they play a very valuable role with this design by increasing stability in the feedback loop and thereby improving the subjective sound - Thx Hugh
Hi Mike
When you have time (If you haven't already) could you describe your PS changes/ implementation - and any other changes you've found worthwhile
Cheers
Hi Otto,
Increasing the current in the o/p stage does not increase H2 H3 H4, rather it significantly reduces H6 - H20 which if present can make the sound less clean and less transparent. With H6 - H20 reduced then the sound of the small amount of H2 - H5 becomes dominant and these distortions are relatively benign. Some people say the sound of an amp can be a bit more dynamic if biased around 200mA . . . so you can try both and see which you prefer. I prefer a higher bias. Most modest systems and reasonably efficient speakers only need about 10 watts to play very loud so with a bias of 0.5 - 1.0A you will probably be working in class A most of the time.
An unloaded supply will float up higher than it's rated voltage so loading the supply with a higher current is OK. If you use a 200VA or higher transformer per channel there should not be a problem.
Well, that's my half penny worth - over to you Hugh
Increasing the current in the o/p stage does not increase H2 H3 H4, rather it significantly reduces H6 - H20 which if present can make the sound less clean and less transparent. With H6 - H20 reduced then the sound of the small amount of H2 - H5 becomes dominant and these distortions are relatively benign. Some people say the sound of an amp can be a bit more dynamic if biased around 200mA . . . so you can try both and see which you prefer. I prefer a higher bias. Most modest systems and reasonably efficient speakers only need about 10 watts to play very loud so with a bias of 0.5 - 1.0A you will probably be working in class A most of the time.
An unloaded supply will float up higher than it's rated voltage so loading the supply with a higher current is OK. If you use a 200VA or higher transformer per channel there should not be a problem.
Well, that's my half penny worth - over to you Hugh
Last edited:
Thanks Mike
I know what you say about the first 5 - 10 watts; and see that my earlier thought of bias upping the lower order harmonics was 'the wrong end of the stick'
So only the transformers need to be upgraded. I'll look for comparison for the VA rating of the 'standard' trannies
It'd be great if it were feasible to 'jumper in' a switch with offboard the different component(s) to alter the bias . . Though to date I've not heard of this being done on any amp . . .
Cheers
I know what you say about the first 5 - 10 watts; and see that my earlier thought of bias upping the lower order harmonics was 'the wrong end of the stick'
So only the transformers need to be upgraded. I'll look for comparison for the VA rating of the 'standard' trannies
It'd be great if it were feasible to 'jumper in' a switch with offboard the different component(s) to alter the bias . . Though to date I've not heard of this being done on any amp . . .
Cheers
Last edited:
<mikelm>
In particular, having now tried O/P chokes I find that they play a very valuable role with this design by increasing stability in the feedback loop and thereby improving the subjective sound - Thx Hugh
Hi Mike
When you have time (If you haven't already) could you describe your PS changes/ implementation - and any other changes you've found worthwhile
Cheers
In the quote above I am referring to 1.5uH output chokes ( with 5-10R in parallel ) rather than the chokes I use in my power supplies.
For power supplies I have C L C with chokes in both rail & earth legs. On Hugh's circuit it's like I replaced those little resistors between the two caps with chokes. The values I am using are something like 8800uF / 1.5mH / 15,400uF.
The caps are paralleled 2,200uF Panasonic FCs and the chokes are the secondaries of 9V + 9V 150VA EI transformers. The chokes should be 1 - 3mH & about 0.1R per winding. Don't use lower resistance than this or may get low frequency ringing.
PM me if you need more on this
mike
> what size transformers you have now?
None: Before i buy any parts, as time permits I'm going through the thread, to see what parts seem to warrant varying from the standard - or having the flexibly to try.
(I ask here, as others might also not be clear)
the chokes you described - what change in subjective sound or objective measurement does this bring?
was ths discussed before?
Thanks again
None: Before i buy any parts, as time permits I'm going through the thread, to see what parts seem to warrant varying from the standard - or having the flexibly to try.
(I ask here, as others might also not be clear)
the chokes you described - what change in subjective sound or objective measurement does this bring?
was ths discussed before?
Thanks again
Power is calculated from:
P = I^2 * R = V^2 / R = I * V
For an AC signal, rms voltages and rms currents must be used.
P = Irms^2 * R = Vrms^2 / R = Irms * Vrms
For a sinewave we can use peak values:
P = Ipk^2 * R / 2 = Vpk^2 / R / 2 = Ipk * Vpk / 2
Using 200mA of output device bias in a push pull output stage we have a peak ClassA current of <400mApk. Assuming exactly 400mApk of ClassA output current then the maximum ClassA power into 8r0 is 0.4^2 * 8 / 2 = 0.64W
500mA bias gives maximum ClassA power into 8r0 of 1^2 * 8 / 2 = 4W
750mA bias gives maximum ClassA power into 8r0 of 1.5^2 * 8 / 2 = 9W
1A bias gives maximum ClassA power of 2^2 * 8 / 2 = 16W
P = I^2 * R = V^2 / R = I * V
For an AC signal, rms voltages and rms currents must be used.
P = Irms^2 * R = Vrms^2 / R = Irms * Vrms
For a sinewave we can use peak values:
P = Ipk^2 * R / 2 = Vpk^2 / R / 2 = Ipk * Vpk / 2
Using 200mA of output device bias in a push pull output stage we have a peak ClassA current of <400mApk. Assuming exactly 400mApk of ClassA output current then the maximum ClassA power into 8r0 is 0.4^2 * 8 / 2 = 0.64W
500mA bias gives maximum ClassA power into 8r0 of 1^2 * 8 / 2 = 4W
750mA bias gives maximum ClassA power into 8r0 of 1.5^2 * 8 / 2 = 9W
1A bias gives maximum ClassA power of 2^2 * 8 / 2 = 16W
Andrew,
These figures are theoretical maths equations but are not the same with lateral mosfets which stay on for longer using constant bias generation figures.
When you use LTSpice you get bigger Class A figures on the FetZilla. I'm not au fait with the mosfet models, but I would draw your attention to the simulations which reveal different power for Class A. For example, at 1A quiescent the max output in Class A (that is, neither output device ever turns off) is 29 watts, 21.6Vpeak into 8R. Here is the LTSpice profile attached.
Cheers,
Hugh (AKSA)
These figures are theoretical maths equations but are not the same with lateral mosfets which stay on for longer using constant bias generation figures.
When you use LTSpice you get bigger Class A figures on the FetZilla. I'm not au fait with the mosfet models, but I would draw your attention to the simulations which reveal different power for Class A. For example, at 1A quiescent the max output in Class A (that is, neither output device ever turns off) is 29 watts, 21.6Vpeak into 8R. Here is the LTSpice profile attached.
Cheers,
Hugh (AKSA)
Attachments
Hi Hugh
As you wrote earlier
– 1 A quiescent generates a lot of heat 
~ with the corollary of a lot of class A
How feasible would be my suggestion to 'jumper in' a switch with offboard the different component(s) to alter the bias.
Depending on:
With bias options of maybe
Cheers
As you wrote earlier
– 1 A quiescent generates a lot of heat ~ with the corollary of a lot of class A
How feasible would be my suggestion to 'jumper in' a switch with offboard the different component(s) to alter the bias.
Depending on:
- the music
- how loud you are listening
- the ambient temp
- the state of your electricity bill
With bias options of maybe
- Mike's preferred level - 0.8
- Normal 300 mA
- And maybe 500 mA
Cheers
Would it be possible to run the FetZilla into about 10 watts in Class A under 4 Ohms or more and the rest in Class B?
If yes, what would be the adjustments to make and can we still keep the same supply voltage?
I'd love to try high Bias on this amp since I have quite a large heatsink attached to it.
Thanks
Do
If yes, what would be the adjustments to make and can we still keep the same supply voltage?
I'd love to try high Bias on this amp since I have quite a large heatsink attached to it.
Thanks
Do
Hi Otto,
Yes you can put in a jumper on one of the diodes on the bias string - one would reduce the quiescent from 1A to 770mA, and a jumper across two diodes would reduce it to 560mA. In both cases, the 295R resistor beneath the diodes would be unchanged.
Supply would slightly rise, maybe from 34Vdc on each rail to 34.5V and maybe 35V. Effectively this would not much change the output power to Class A but this will change radically with the impedance of the speaker and the quiescent current.
Do,
My LTSpice indicates that at 1A quiescent and 4R load, Class A departs at 11.8Vp, which is 17.5W. Beyond this, power rises to 26.3Vp into 4R, which is 86.5 watts into 4R. This looks good, but it's pushing the outputs at 6.6A peak, and it will generate quite a bit of distortion. Here are the figures at flat strap:
Harmonic Frequency Normalized Normalized
Number [Hz] Component Phase [deg]
1 1.000e+03 1.000e+00 0.00°
2 2.000e+03 2.173e-03 -89.43°
3 3.000e+03 1.658e-03 13.08°
4 4.000e+03 1.389e-03
5 5.000e+03 1.189e-03 -158.67°
6 6.000e+03 9.351e-04 -86.60°
7 7.000e+03 8.313e-04 30.35°
8 8.000e+03 5.426e-04 96.86°
9 9.000e+03 4.693e-04 -144.81°
10 1.000e+04 2.432e-04 -91.88°
11 1.100e+04 1.615e-04 41.93°
12 1.200e+04 8.011e-05 101.06°
13 1.300e+04 5.839e-05 -178.31°
14 1.400e+04 1.377e-05 -64.73°
15 1.500e+04 5.024e-05 -87.01°
Total Harmonic Distortion: 0.360370%
H2 appears to be higher than -52dB, which is VERY high.
Yes you can put in a jumper on one of the diodes on the bias string - one would reduce the quiescent from 1A to 770mA, and a jumper across two diodes would reduce it to 560mA. In both cases, the 295R resistor beneath the diodes would be unchanged.
Supply would slightly rise, maybe from 34Vdc on each rail to 34.5V and maybe 35V. Effectively this would not much change the output power to Class A but this will change radically with the impedance of the speaker and the quiescent current.
Do,
My LTSpice indicates that at 1A quiescent and 4R load, Class A departs at 11.8Vp, which is 17.5W. Beyond this, power rises to 26.3Vp into 4R, which is 86.5 watts into 4R. This looks good, but it's pushing the outputs at 6.6A peak, and it will generate quite a bit of distortion. Here are the figures at flat strap:
Harmonic Frequency Normalized Normalized
Number [Hz] Component Phase [deg]
1 1.000e+03 1.000e+00 0.00°
2 2.000e+03 2.173e-03 -89.43°
3 3.000e+03 1.658e-03 13.08°
4 4.000e+03 1.389e-03
5 5.000e+03 1.189e-03 -158.67°
6 6.000e+03 9.351e-04 -86.60°
7 7.000e+03 8.313e-04 30.35°
8 8.000e+03 5.426e-04 96.86°
9 9.000e+03 4.693e-04 -144.81°
10 1.000e+04 2.432e-04 -91.88°
11 1.100e+04 1.615e-04 41.93°
12 1.200e+04 8.011e-05 101.06°
13 1.300e+04 5.839e-05 -178.31°
14 1.400e+04 1.377e-05 -64.73°
15 1.500e+04 5.024e-05 -87.01°
Total Harmonic Distortion: 0.360370%
H2 appears to be higher than -52dB, which is VERY high.
Aksa,
you said similar a while back, maybe in this Thread.
Your definition of ClassA is different from mine.
I can't accept "not turning off" or "turning off more slowly/softly at the low Vgs end" as ClassA.
If we look at the plot you kindly took the time to compile for us, I can see the gross 2nd harmonic distortion in the waveform.
That, at least for me, shows that the nearly turning off device is not actively controlling the ClassA output current.
That sentence indicates where my definition of ClassA comes from.
Both devices in the push pull stage must actively control the ClassA output current.
And from that: if one of the devices is starting to "turn off", then it has lost the active part of the control. It's the quite precise "balance" of the upper and lower devices that gives the lack of "crossover" distortion.
In my view, if a device goes into a current passing (not quite turned off) phase of operation but not actively controlling the ClassA current, then it is not in push pull ClassA operation.
It's that type of operation where some designers deliberately design for "not turning off" and claim ClassA operation with AA and AAA nomenclature and others, to differentiate from true ClassA.
I suspect we will not agree on this.
If you have a counter argument, then the Members will probably want to hear, so each can decide.
Sorry that does not sound right.
If you want to take the time to present a counter argument, etc.
you said similar a while back, maybe in this Thread.
Your definition of ClassA is different from mine.
I can't accept "not turning off" or "turning off more slowly/softly at the low Vgs end" as ClassA.
If we look at the plot you kindly took the time to compile for us, I can see the gross 2nd harmonic distortion in the waveform.
That, at least for me, shows that the nearly turning off device is not actively controlling the ClassA output current.
That sentence indicates where my definition of ClassA comes from.
Both devices in the push pull stage must actively control the ClassA output current.
And from that: if one of the devices is starting to "turn off", then it has lost the active part of the control. It's the quite precise "balance" of the upper and lower devices that gives the lack of "crossover" distortion.
In my view, if a device goes into a current passing (not quite turned off) phase of operation but not actively controlling the ClassA current, then it is not in push pull ClassA operation.
It's that type of operation where some designers deliberately design for "not turning off" and claim ClassA operation with AA and AAA nomenclature and others, to differentiate from true ClassA.
I suspect we will not agree on this.
If you have a counter argument, then the Members will probably want to hear, so each can decide.
Sorry that does not sound right.
If you want to take the time to present a counter argument, etc.
Last edited:
No problem Andrew. We agree to disagree.
Either way this is an ideology difference, with no effect on reality.
What is important is whether the FetZilla sounds better or worse with more quiescent, and indeed this is a subjective assessment.
I concede that as current through an active reduces, it loses control over the signal, as shown by increasing H2 on the source emitters, but then, the active device is dominating procedure anyway. The H2 created at the output signal is very small; all you see is the current through the source resistors.
Nonetheless, any difference between Class A at 16 watts or 29 watts is probably immaterial for the FetZilla, since the first two or three watts are definitely the clincher on the sound quality.
Cheers,
Hugh
Either way this is an ideology difference, with no effect on reality.
What is important is whether the FetZilla sounds better or worse with more quiescent, and indeed this is a subjective assessment.
I concede that as current through an active reduces, it loses control over the signal, as shown by increasing H2 on the source emitters, but then, the active device is dominating procedure anyway. The H2 created at the output signal is very small; all you see is the current through the source resistors.
Nonetheless, any difference between Class A at 16 watts or 29 watts is probably immaterial for the FetZilla, since the first two or three watts are definitely the clincher on the sound quality.
Cheers,
Hugh
Last edited:
Andrew,
I'm embarrassed to answer this because I've not tried it. When you work with an amp productionising and filling the documentation, you don't examine all the options. I don't believe Greg, Lineout and Mikelm did it either, but you'd have to ask them. Significantly you are asking me now, more than two years later!
I always had one comment about the FetZilla - that is lacks depth of image. This is a major issue for me, I strive to throw up a virtual depth in the sound field. I think it was related to the choice of the voltage amplifier. But, as with so many aspects of this art, I do not have a proof for this. But I can say that the depth of image is magnificent with the NAKSA.
Cheers,
Hugh
I'm embarrassed to answer this because I've not tried it. When you work with an amp productionising and filling the documentation, you don't examine all the options. I don't believe Greg, Lineout and Mikelm did it either, but you'd have to ask them. Significantly you are asking me now, more than two years later!
I always had one comment about the FetZilla - that is lacks depth of image. This is a major issue for me, I strive to throw up a virtual depth in the sound field. I think it was related to the choice of the voltage amplifier. But, as with so many aspects of this art, I do not have a proof for this. But I can say that the depth of image is magnificent with the NAKSA.
Cheers,
Hugh
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.