Yes, I'm clear about that. However, I don't see them as big advantages.
We can easily get gain in the power stage with the addition of a pair of drivers. Even though I prefer to drive mosfets directly with the VAS, in this instance, I would rather have drivers because driving the power mosfets directly from the VAS requires quite a bit of "juice". If the VAS don't have enough current, the amp will slew limit early.
As for direct mounting, I have no issues with mica and thermal compound. I do agree that mounting is faster without insulators and there's slightly better heat transfer but I don't view it as a quantum gain.
Putting aside the PS, can you help me out on where the Positive feedback loop is? I simply can't seem to locate it.
Cheers
Mike
We can easily get gain in the power stage with the addition of a pair of drivers. Even though I prefer to drive mosfets directly with the VAS, in this instance, I would rather have drivers because driving the power mosfets directly from the VAS requires quite a bit of "juice". If the VAS don't have enough current, the amp will slew limit early.
As for direct mounting, I have no issues with mica and thermal compound. I do agree that mounting is faster without insulators and there's slightly better heat transfer but I don't view it as a quantum gain.
Putting aside the PS, can you help me out on where the Positive feedback loop is? I simply can't seem to locate it.
Cheers
Mike
Hi Zen Mod
Yes, I was wrong. I can see they are +HV and -HV. Somehow, I saw them as +/-35V.
The rails are indeed much higher. As Paul pointed out, the P9505 is about +/- 84V. I took another look at the QSC1400 and their rails are at +/-75V. So the rails are at the same voltages as a conventional PS where 0V in the center of the PS.
I understand that in order to have Gain in the current stage, the mosfets are arranged as common source. Output will therefore be at Drain.
What I am saying is if we revert back to a conventional PS by shifting the 0V back to the center of the PS and adding a pair of drivers after the VAS, we can still get Gain in the output stage.
I don't know how much it will affect the voltage nodal point in Strickland's patent by doing this. What is clear in his patent is 3 feedback mechanism are employed. 2 are negative feedback and 1 is positive feedback. I still cannot find the positive feedback loop.
There is one other point that I am puzzled. In the P9505, there are only 4 pairs of lateral mosfets. I assume they are 2SK1058 and 2SJ162 running at +/-84V rails. Is this not a bit too demanding on the mosfets?
Cheers
Mike
Yes, I was wrong. I can see they are +HV and -HV. Somehow, I saw them as +/-35V.
The rails are indeed much higher. As Paul pointed out, the P9505 is about +/- 84V. I took another look at the QSC1400 and their rails are at +/-75V. So the rails are at the same voltages as a conventional PS where 0V in the center of the PS.
I understand that in order to have Gain in the current stage, the mosfets are arranged as common source. Output will therefore be at Drain.
What I am saying is if we revert back to a conventional PS by shifting the 0V back to the center of the PS and adding a pair of drivers after the VAS, we can still get Gain in the output stage.
I don't know how much it will affect the voltage nodal point in Strickland's patent by doing this. What is clear in his patent is 3 feedback mechanism are employed. 2 are negative feedback and 1 is positive feedback. I still cannot find the positive feedback loop.
There is one other point that I am puzzled. In the P9505, there are only 4 pairs of lateral mosfets. I assume they are 2SK1058 and 2SJ162 running at +/-84V rails. Is this not a bit too demanding on the mosfets?
Cheers
Mike
Michael Chua said:
as I see it , there are at least 3 important things :
1 . all mentioned amps ( Haflers) have that TransNova output config , but I can't see in any of them error correction implemented in Patent
2.you can - if you want - re-arrange almost any "conventional" amp to this sort of output/PSU , same as you can re-arrange Hafler to conventional output/PSU config.
3. is it Hafler output common drain , or common source - that's completely another issue , without special connection to TransNova output/PSU config ........ or I'm misiing something ?
maybe , because I really need to spend some more time on patent reading ....
you can see different output configurations ( common drain, common source ) in Papa's amps .... each of them have own virtues ...
anyway - regarding number of output mosfets .... Hafler was always even conservative in power ratings of their amps
even if these mosfets are ratyed for Uds of 160V , I suspect that they can see full throttle PSU voltage ever
Michael Chua said:
The power supply cap's aren't any more in series with the load than for the standard connection, see attached picture - all three are equivalent. The local feedback through the parallelled resistor+cap essentially makes the circuit a current controlled source follower.
Michael Chua said:
You could reverse the question: What are the advantages of doing it the "normal" way? The advantages and disadvantages of each are pretty small.
Michael Chua said:
One of the advantages of the circuit is that the driver stage is powered from lower voltage. Say we want 100mA of VAS standing current to be able to drive the gates nicely:
* In the conventional circuit VAS power supply voltage would need to be +-(88 + 12) = +-100V for the same rail efficiency. Being able to provide 10V to the gates even when near the rail. VAS dissipation: 0.1A * 200V = 20W
* In the grounded source circuit: supply voltage can be as low as +-15V for the vas. That would give 3W of VAS losses at the same current.
Almost a 10 time improvement in VAS power dissipation for the same standing current. The VAS can use cheaper and faster transistors thanks to the lower voltage and power dissipation.
Michael Chua said:
It's not really applicable to this output stage as it is a transresistance amplifier - current in, voltage out. Though, it can be seen if the source is transformed into it's equivalent voltage source, it's more like a common drain amplifier (follower) than common source.
Michael Chua said:
As already pointed out earlier, that's because the positive feedback wasn't used in this amp. Not in the 9505 either. I'd guess they had problems stabilizing it and abandoned it. Alternatively, it didn't provide enough (or any) improvement to be worthwile.
Michael Chua said:
That seems to be common practice with lateral fets. Hafler used 3 pairs on +-90V in one of their models for instance. Also, the improved thermal transfer without mica will help.
But the biggest reason it's possible is that music has low average power. I think it was Eva that successfully used 2 pairs of Sanken 2SC3264 on +-90V driving 4 ohms - which might seem to be way too little. And it is if you want to be able to get sinewawe out at the worst case output level for power dissipation. But music is pretty different.
In this case it's even less demanding than that though. The laterals say 400W 10ms, only 33% less although power rating is half. And there's two times as many. 1600W 10ms pulse SOA instead of 1200W for the 2 2SC3264.
And then there's the negative current temperature coefficient and the (parasitic?) internal overtemperature protection at ~200 degrees C which makes the laterals almost impossible to destroy as long as gate current is limited.
Attachments
Hello, I need help with my Hafler 9505. I was only able to find 23,000 uF 85 volt 20% PS Capacitors and am wondering if these will be ok? I have installed them and the amplifier plays fine with bookshelf speakers is what Ionly dare to use until I find out if these caps will work. The voltage across the caps was 84.5 VDC with my DMM if this helps and with the 20%= 85+17 volts that is 102 volts max rating of the capacitors. will this work, please help!!! the caps have to be 3 5/8" tall... Thanks Bill
The 20% is the tolerance on the capacitance value, not a tolerance on maximum voltage (which you should derate 15 - 20%) which in this case is 85V. I would not operate these much beyond 70V tbh. These capacitors are not safe for operation at 84V unless I have misunderstood something. (I'm assuming the voltage was measured across the cap.)
Hello,
Not sure how the posting goes on this site as this is my second try at it. I was only able to find caps that are 23,000 uF @85vdc 20% that fit because of height restrictions and I didnt want to build a satelite capacitor bank. I wanted to know it these will work. I measured 84.5 vdc when warmed up.
Please help me out. Power Supply Question.
Bill
Hello, I am unable to find replacement caps for this amplifier, can anyone help?
I need 3 5/8 height 2 1/2" dia and screw spacing 1.13" 28.6mm scres not snap in. The schematics read 100vdc. What is going to happen if I leave the 85vdc caps in, please explain... Thanks, Bill
I just learning this site so bare with me
The caps will work, just with a limited lifetime. I'd worry if your line voltage tends to rise at times, though.
Stressed electrolytic caps can explode, or more likely in this case, vent and release nasty chemicals.
You could use these while searching for more exact replacements, possibly NOS/surplus.
Are you certain that you actually need new caps?
Last edited:
Hi Guys
The circuit in the first post is suppose to have the advantage that only the output devices see the full supply voltage, and indeed this is true. The front end operates at +/-24V which could just as easily be half this amount for the same performance. The input stage only has to generate a signal equivalent to the gate-to-gate swing required to fully enhance the mosfets, which is about 14Vpp.
The output stage operates common-source, which in the advertising of the times was described as a "drain follower". At idle, the devices see the supply as +/-50V. As one device turns on and the other turns off, the supply is seen to swing from the extreme of +0,-100 and +100,-0 - not really to zero but you see the trend.
Feedback is taken from the floating supply centre-tap. The supply is built as any split-rail supply but not grounded directly. Instead, the mosfets pull the supply up and down and the speaker is driven from the supply CT. Feedback is taken from this point which puts a "crimp" in the power of the feedback loop. The feedback node is not directly regulated by the output devices, so performance of amplifiers like these is never as good as the conventional approach. The Trans Nova amps were essentially the same.
Some designers object to the fact that the power transformer interwinding capacitances have signal currents through them and signal voltages across them. This can cause outbound noise onto the mains in addition to introducing other supply modulations that can lead to instabilities of the amplifier. Clever designs tend to hold less promise than their novelty suggests.
The quality of the filter caps does not have to be any different than that for a conventional amp. of course, we are assuming long-life low-leakage caps are used, not voodoo "audio" caps nor low-grade junk.
Note: I'll admit to not having read the whole thread so the above should already have been covered...
Have fun
The circuit in the first post is suppose to have the advantage that only the output devices see the full supply voltage, and indeed this is true. The front end operates at +/-24V which could just as easily be half this amount for the same performance. The input stage only has to generate a signal equivalent to the gate-to-gate swing required to fully enhance the mosfets, which is about 14Vpp.
The output stage operates common-source, which in the advertising of the times was described as a "drain follower". At idle, the devices see the supply as +/-50V. As one device turns on and the other turns off, the supply is seen to swing from the extreme of +0,-100 and +100,-0 - not really to zero but you see the trend.
Feedback is taken from the floating supply centre-tap. The supply is built as any split-rail supply but not grounded directly. Instead, the mosfets pull the supply up and down and the speaker is driven from the supply CT. Feedback is taken from this point which puts a "crimp" in the power of the feedback loop. The feedback node is not directly regulated by the output devices, so performance of amplifiers like these is never as good as the conventional approach. The Trans Nova amps were essentially the same.
Some designers object to the fact that the power transformer interwinding capacitances have signal currents through them and signal voltages across them. This can cause outbound noise onto the mains in addition to introducing other supply modulations that can lead to instabilities of the amplifier. Clever designs tend to hold less promise than their novelty suggests.
The quality of the filter caps does not have to be any different than that for a conventional amp. of course, we are assuming long-life low-leakage caps are used, not voodoo "audio" caps nor low-grade junk.
Note: I'll admit to not having read the whole thread so the above should already have been covered...
Have fun
Last edited:
Hi Struth,
I'd like to elaborate this point a bit more if you don't mind.
Let's look at the push-pull OPS from "fundamental" point of view. It fact, output devices control the way, the current from PSU flows through the load.
All the rest of the amplifier circuit provides the required voltage gain and the means of driving the output devices.
The closer PSU to ideal one (constant V, unlimited "I", zero impedance) - the better. No doubt.
Now - please see the picture. In fact, the only difference between (1) "conventional" OPS arrangement and (2) the one, utilised in QSC, Hafler 915 and some other amps, is the grounding point and the point where we take NFB from (opposite sides of the load).
What I'd like to say - your comments with regards to PSU issues (transformers, filtering caps) are true for both cases.
Some people also claim the 2-nd option never performs as good as the 1-st one. I disagree with that because of the considerations mentioned above. No principal difference.
What do you think?
Cheers,
Valery
Attachments
Last edited:
Hi Forr,
Good example
However, the word "audiophile" in the article looks like a certain exaggeration
Seems too "class B" to me.Here is the other one from Linear Audio with properly biased lateral FETs:
https://linearaudio.net/sites/linearaudio.net/files/May 28-Figure 9 - Full Amp.pdf
Cheers,
Valery
I tried similar but much simpler solution, the link to pdf with scheme: http://www.diyaudio.com/forums/solid-state/57892-tcj-evolve-simply-good-amps-idea-2.html#post654105
Well, combination of these two - the one mentioned by Forr and Pawel's one - would form a very decent amplifier. I would also add a bias spreader - it's more or less ok to run as is at +/-15V rails, but at +/-25...50V, for example, thermal feedback is a must, if we are talking about IRFPs (laterals - a different story).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.