First, a big thanks for NP for sharing so many great designs. I have recently built a balanced mono ACA’s and F5, and I am quite happy with them.
After reading through NP’s 2 great papers (more than papers, kind of design manifestation actually
);
1) Leaving Class A - http://www.firstwatt.com/pdf/art_leave_classa.pdf
2) Audio, Distortion and Feedback - http://www.firstwatt.com/pdf/art_dist_fdbk.pdf
And also seeing the distortion cancellation in my ACAs between single ended and balanced modes through a QA401; I have been thinking to build a balanced, no-feedback, very high Class A bias amps.
With its massive simplicity and universal positive feedback F4 looks just perfect.
As a starting point, recapping 6L6’s measurements for distortion (1kHz, 4 ohm, 3V, ~2Watts) for different Bias settings (part of his great build guide - A guide to building the Pass F4 amplifier.)
100 mV Bias - 0.065%
200 mV Bias - 0.016% (Default setting)
270 mV Bias - 0.008%
370 mV Bias - 0.0045%
Well, in the spirit of overkill, if one goes for say 470 mV Bias, some very rough extrapolation would give around 0.003% distortion
And, on top, given that this distortion is predominantly 2nd order, running two output stages in a balanced setup would yield a good chunk of distortion cancellation; if we assume 1/5 for this effect, then we might potentially end up having 0.0006% distortion!
Just to make it clear, I have no intentions to delve into sound quality vs. distortion measurement, yet as crazy as this number sounds for any amplifier including Class D’s etc, having a very simple, no feedback amp reaching this level is just insane.
I wonder whether these assumptions are sound, and/or anyone else built/measured similar very high biased balanced F4s?
After reading through NP’s 2 great papers (more than papers, kind of design manifestation actually
); 1) Leaving Class A - http://www.firstwatt.com/pdf/art_leave_classa.pdf
2) Audio, Distortion and Feedback - http://www.firstwatt.com/pdf/art_dist_fdbk.pdf
And also seeing the distortion cancellation in my ACAs between single ended and balanced modes through a QA401; I have been thinking to build a balanced, no-feedback, very high Class A bias amps.
With its massive simplicity and universal positive feedback F4 looks just perfect.
As a starting point, recapping 6L6’s measurements for distortion (1kHz, 4 ohm, 3V, ~2Watts) for different Bias settings (part of his great build guide - A guide to building the Pass F4 amplifier.)
100 mV Bias - 0.065%
200 mV Bias - 0.016% (Default setting)
270 mV Bias - 0.008%
370 mV Bias - 0.0045%
Well, in the spirit of overkill
, if one goes for say 470 mV Bias, some very rough extrapolation would give around 0.003% distortionAnd, on top, given that this distortion is predominantly 2nd order, running two output stages in a balanced setup would yield a good chunk of distortion cancellation; if we assume 1/5 for this effect, then we might potentially end up having 0.0006% distortion!
Just to make it clear, I have no intentions to delve into sound quality vs. distortion measurement, yet as crazy as this number sounds for any amplifier including Class D’s etc, having a very simple, no feedback amp reaching this level is just insane.
I wonder whether these assumptions are sound, and/or anyone else built/measured similar very high biased balanced F4s?
Last edited:
Feasibility
Btw, in addition to this theoretical performance, if one thinks about the feasibility;
1) Mosfet power dissipation
At 470 mV Bias with suggested 0.47 ohm source resistors, every output device will draw around 1A. And with suggested 24V rail voltage, each will dissipate 24W. So, it is still lower than 30W value that NP suggests, and should be OK!
2) Total power dissipation
With 6 output devices per side and 12 per channel for balanced mono, we have 12 x 24 = 288W dissipation per channel. It is on the higher side, but still feels quite doable with heavy heatsinks; probably Big 5U chassis with 40 cm deep per channel will be just fine.
3) Input JFETs current drive
Not sure whether they will be able to drive the circuit due to much higher class A bias, any ideas?
Btw, in addition to this theoretical performance, if one thinks about the feasibility;
1) Mosfet power dissipation
At 470 mV Bias with suggested 0.47 ohm source resistors, every output device will draw around 1A. And with suggested 24V rail voltage, each will dissipate 24W. So, it is still lower than 30W value that NP suggests, and should be OK!
2) Total power dissipation
With 6 output devices per side and 12 per channel for balanced mono, we have 12 x 24 = 288W dissipation per channel. It is on the higher side, but still feels quite doable with heavy heatsinks; probably Big 5U chassis with 40 cm deep per channel will be just fine.
3) Input JFETs current drive
Not sure whether they will be able to drive the circuit due to much higher class A bias, any ideas?
Where did you see the 30W recommendation? The 240/9240s are rated at a maximum of 150W so I'd think they'd have a good lifespan at 50W, and probably even a decent one at 75W. Given sufficient heat-sinking, of course.
When you say "mono" I assume you mean 1 5U chassis per channel (and not just mono PSUs)? That should do the trick.
My understanding (admittedly weak) is that the drive requirements from the front end have more to do with the output FET gate capacitances than their bias.
Cheers,
Jeff.
When you say "mono" I assume you mean 1 5U chassis per channel (and not just mono PSUs)? That should do the trick.
My understanding (admittedly weak) is that the drive requirements from the front end have more to do with the output FET gate capacitances than their bias.
Cheers,
Jeff.
oh yes , you did
unless you have significant experience with mosfet dissipation , don't be brave on someone else's expense
I'm using IRFP150 (practically 2 x IRFP240 in one case) on 50W heat , but gooood heatsinking and keratherm is a must
IRFP240 .......... sane is 35W , 40W is still sane , while more than that is strictly your own responsibility
unless you have significant experience with mosfet dissipation , don't be brave on someone else's expense
I'm using IRFP150 (practically 2 x IRFP240 in one case) on 50W heat , but gooood heatsinking and keratherm is a must
IRFP240 .......... sane is 35W , 40W is still sane , while more than that is strictly your own responsibility
Hi Jeff,
On the 30W recommendation, please see BA2 article:
"...If you have enough heat sinking, you can consider as high as 30 watts per device, allowing the reliable use of fewer devices.
If you decide to higher, remember that Mosfets are better at higher current until they catch fire and fulfill the promise of the amplifier's name."
http://www.firstwatt.com/pdf/art_ba2.pdf
Cheers,
Dennis
On the 30W recommendation, please see BA2 article:
"...If you have enough heat sinking, you can consider as high as 30 watts per device, allowing the reliable use of fewer devices.
If you decide to higher, remember that Mosfets are better at higher current until they catch fire and fulfill the promise of the amplifier's name."
http://www.firstwatt.com/pdf/art_ba2.pdf
Cheers,
Dennis
Thanks Zen Mod!
I did not know that there is a much powerful variant of irfp240. This sparked a new thought!
What happens if instead of 3xirfp240 per side per channel, I use only 1 irfp150?
My understanding is;
Downside;
= I will loose the 1/3 of current drive (as essentially moving from 3 devices to 2)
Upside;
= Great work around for the joy of the MOSFET matching, with just one device p or n device per side, no need to worry about it at all (p vs n matching is dealt with trimming pot)
= I can decrease the source resistors as NP mentiones. Actually, with only 1 device, do I even need a source resistor?
= It will also lower the total heat dissipation by 1/3, from 288W per channel to ~190W, much easier to accommodate
Btw, fully agree on the need for the proper heatsinking. I checked my F5 based on 4U case 300mm, while the case is not warm, looking with a thermal camera output devices are not efficiently using the full case capacity. I was thinking about having copper plates between Mosfets and aliminium heatsinks to a) better dissipate heat from MOSFETS and b) more uniformly distribution temperature across the heatsink
Thoughts?
I'm using IRFP150 (practically 2 x IRFP240 in one case) on 50W heat , but gooood heatsinking and keratherm is a must
IRFP240 .......... sane is 35W , 40W is still sane , while more than that is strictly your own responsibility
I did not know that there is a much powerful variant of irfp240. This sparked a new thought!
What happens if instead of 3xirfp240 per side per channel, I use only 1 irfp150?
My understanding is;
Downside;
= I will loose the 1/3 of current drive (as essentially moving from 3 devices to 2)
Upside;
= Great work around for the joy of the MOSFET matching, with just one device p or n device per side, no need to worry about it at all (p vs n matching is dealt with trimming pot)
= I can decrease the source resistors as NP mentiones. Actually, with only 1 device, do I even need a source resistor?
= It will also lower the total heat dissipation by 1/3, from 288W per channel to ~190W, much easier to accommodate
Btw, fully agree on the need for the proper heatsinking. I checked my F5 based on 4U case 300mm, while the case is not warm, looking with a thermal camera output devices are not efficiently using the full case capacity. I was thinking about having copper plates between Mosfets and aliminium heatsinks to a) better dissipate heat from MOSFETS and b) more uniformly distribution temperature across the heatsink
Thoughts?
Last edited:
They are 6 ohm nominal, probably going down to 4ohm min.
Though, I would love these amps to play well with potentially more demanding loads just in case I change my speakers in the future.
More important; I remember reading BA3 and F4 being subjectively more authoritative (3 mosfet pair output stages) compared to F5 with only 1 pair, so I don’t want to loose that aspect
Though, I would love these amps to play well with potentially more demanding loads just in case I change my speakers in the future.
More important; I remember reading BA3 and F4 being subjectively more authoritative (3 mosfet pair output stages) compared to F5 with only 1 pair, so I don’t want to loose that aspect
With 23V rails you can only make use of about 5.5A into 4ohm. And even if you do run out of current drive this output stage will just drop into class AB rather than clip like a single-ended output would.
Beyond that I think you're just burning bias to generate heat to lower distortion. How fewer devices running at the same temperature will affect that I haven't a clue.
Beyond that I think you're just burning bias to generate heat to lower distortion. How fewer devices running at the same temperature will affect that I haven't a clue.
good luck finding a mate for IRFP150
good for one sex mosfets in output ( see Babelfish J thread) , but problematic for complementary ...... because there is no complementary , easily available
at least to my knowledge
That is very true. I am very tempted with my theoretical 0.0006% distortion calculation in the very first post of the thread.
Only a few gifted people can discern distortion lower than 0.01%.
I will see whether I am gifted enough
In all fairness, this is more of technically amazing case for the engineer inside me that I would love to figure out more.
No intentions to go for a distortion vs sound quality debate..
This is DIY. Because it's technically interesting, or aesthetically pleasing, or a challenge, or oddball, or anything else that floats your boat are valid reasons to build. There's no requirement that everything be about sound.
(Not to refute Indra1's post, as the info is interesting. Just sayin'.)
(Not to refute Indra1's post, as the info is interesting. Just sayin'.)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.