Member
Joined 2009
Paid Member
This might be too obvious a question - but I am still collecting parts for a dc-coupled SET and am wondering what is the primary reason why I'm doing a dc-coupled amp.
I thought it was to avoid the 'sound' of a coupling capacitor. But I am learning more and more that this is not something that can't be managed.
Is it to avoid 'blocking distortion' ? how big an issue is this ???
I thought it was to avoid the 'sound' of a coupling capacitor. But I am learning more and more that this is not something that can't be managed.
Is it to avoid 'blocking distortion' ? how big an issue is this ???
Blocking distortion occurs when the amplifier is overdriven and from this point it starts to flow a grid current into valve. Then capacity (capacitor in the signal path) starts to charge and this change a bias point.
Capacitors in the signal path also changes a phase of signal and then can occur a problem with negative feedback.
Well, you're probably doing a DC coupled amp because you're a DIY hobbyist, and it's fun to 'collect' different topologies. =)
DC coupled SET has the problem of mandatory cathode bias (or else PSU trickery, but most go with cathode bias). If cathode bias is RC, then you've only moved the coupling cap sound problem to another location, and since you need a much bigger cap there than a coupling cap would be, it might be fair to say you've made the problem actually worse, regarding the capacitor.
That said, I went thru a "DC only" phase myself. It did sound better to my ears than alternatives at the time, and clearly so. I actually ended up 'solving' the RC cathode bias problem with 3 parallel LED strings of 175 volts... That was pretty, and it did work, but very very stupid.
In a balanced amp there is no downside really, outside of wasted energy because of high B+, because the output stage has a CCS tail, which is completely transparent soundwise.
I'd venture a guess based on my experiences that most of DC coupling sonic gains can be attributed to better grid drive. When I started using MOSFET source followers to drive all output stage grids, my first listening impression was "ah, it's the DC sound again!"
DC coupled SET has the problem of mandatory cathode bias (or else PSU trickery, but most go with cathode bias). If cathode bias is RC, then you've only moved the coupling cap sound problem to another location, and since you need a much bigger cap there than a coupling cap would be, it might be fair to say you've made the problem actually worse, regarding the capacitor.
That said, I went thru a "DC only" phase myself. It did sound better to my ears than alternatives at the time, and clearly so. I actually ended up 'solving' the RC cathode bias problem with 3 parallel LED strings of 175 volts... That was pretty, and it did work, but very very stupid.
In a balanced amp there is no downside really, outside of wasted energy because of high B+, because the output stage has a CCS tail, which is completely transparent soundwise.
I'd venture a guess based on my experiences that most of DC coupling sonic gains can be attributed to better grid drive. When I started using MOSFET source followers to drive all output stage grids, my first listening impression was "ah, it's the DC sound again!"
Blocking distortion isn't an issue at all if you never overdrive the amp. But if you do, it really slows the overload recovery of the amp and can sound quite nasty or just kind of nasty depending on the design of the amp.
I did some experiments with a push-pull amp with a direct coupled mosfet follower driver once where I drove the amp to clipping with music playing and watched the waveform on a scope. I was quite amazed actually at how much clipping it took for it to get to the point where it sounded really bad. Instantaneous overload recovery is a very worthy design goal in my opinion.
Of course, in a single ended amp you won't get crossover distortion when the cap charges at overload like you would in a p-p amp, but it will still shift the bias and take some time to recover back to the bias point. You would have to do an experiment to see exactly how the overload recovery of the two methods subjectively sound to you and whether it is ultimately worth it to you.
But of course all of this is moot if you never drive the amp to clipping.
I did some experiments with a push-pull amp with a direct coupled mosfet follower driver once where I drove the amp to clipping with music playing and watched the waveform on a scope. I was quite amazed actually at how much clipping it took for it to get to the point where it sounded really bad. Instantaneous overload recovery is a very worthy design goal in my opinion.
Of course, in a single ended amp you won't get crossover distortion when the cap charges at overload like you would in a p-p amp, but it will still shift the bias and take some time to recover back to the bias point. You would have to do an experiment to see exactly how the overload recovery of the two methods subjectively sound to you and whether it is ultimately worth it to you.
But of course all of this is moot if you never drive the amp to clipping.
Agree with
but it's probably not only the (missing) cap in the grid driver arrangement, but likely also the lower drive impedance acheived using source or cathode (if you want to stay tube only) followers.
By using the ultrapath output stage arragement you can short the audio path directly back to B+ and thus bypass the PSU and it's reservoir cap. SET topologies with the ultrapath cap as the only cap in the whole audio circuit become possible. (PSU is out of the audio path in this arrangement)
but it's probably not only the (missing) cap in the grid driver arrangement, but likely also the lower drive impedance acheived using source or cathode (if you want to stay tube only) followers.
By using the ultrapath output stage arragement you can short the audio path directly back to B+ and thus bypass the PSU and it's reservoir cap. SET topologies with the ultrapath cap as the only cap in the whole audio circuit become possible. (PSU is out of the audio path in this arrangement)
Just wanted to add that blocking distortion is even more of an issue when nfb is used.
I wrote carelessly. I meant the comparison in my "aha" moment to be between direct coupled and cap coupled to source follower direct coupled to grid. So the source follower one is no more DC, because there is a coupling cap in the signal path.
The follower - in my experience - makes the sound "DC" again, in fact "more DC than actual DC" if that makes any sense. These experiments have convinced me that the majority of DC benefits are not related to the capacitor as such, but more to current drive capability.
In short, it's not the cap that's "ruining the sound", it's the compromised grid drive. Fix the grid drive, and the cap can stay with no problems.
That's true. However you're still left with substituting a high quality low cost film cap with a big cap. Probably an electrolytic, or else a very expensive one. Ultrapath is very good, I've done it using a single FET capacitance multiplier in place of the capacitor - best way to do cathode bias in SE, in my experience.
In DC, subtract one good quality cap from signal path, add one capacitor of lesser quality, whichever place you choose to add it to.
DC coupling isn't always better,
Eg if you don't limit some frequency response you get Op Tx saturation.
Also if you play records you get cone flap with record warp..ie the cones of the speakers pulse with the record rotation.
The other problem can be dc offset under fault conditions.
Just a few thoughts.
Regards
M. Gregg
Eg if you don't limit some frequency response you get Op Tx saturation.
Also if you play records you get cone flap with record warp..ie the cones of the speakers pulse with the record rotation.
The other problem can be dc offset under fault conditions.
Just a few thoughts.
Regards
M. Gregg
I've often wondered if one could make a 'DC connected' amplifier that at the same time also uses a bit of active circuitry to 'down-bias' the grid of each stage by the +bias coming from the previous stage's anode & resistor current-to-voltage converter.
You know, say a +300 V power supply and a –300 V supply.
Let's have a 12×X7 passing 2 ma, fed to a 75 kΩ anode resistor. Voltage drop is 150 V, so the anode is comfortably at 150 V quiescent. The grid of the next stage might nominally have a 470 kΩ grid-to-ground resistor (nominally). Something somewhat bigger than the driving stage's effective impedance. (3× to 10× seems rational)
Well, if instead of the 470 kΩ grid-to-ground, it is instead two resistors of 470 kΩ in series, with the center tap going to the grid, and the “bottom one” going to –150 V supply (instead of ground), well … that grid is going to nominally be 0.0 V, isn't it?
YES: the signal is also divided in half (–6 dB of gain loss). Who cares! Now the signal can be used as direct-DC coupled, without a capacitor anywhere.
TRICK IS… how to regulate the –150 volt path?
I should think that this is where a bit of sand-state comes in handy. A very high impedance 'leak resistor' coming from Stage 1's anode to a high voltage emitter-follower to charge up an averaging capacitor (without getting the cap in the signal path), then a high-voltage ganged op-amp to invert the voltage relative to ground (sign reversal). now you have the negative supply, which only need supply microamps. A wee bit of “balancing the pair of series resistors” (at the top of this exercise) gets one darn close to 0.0 volts. Close enough. Just something other than +150 volts.
Heck … this sounds perfect as a drop-in module, black-box, tidy, doing its job. An averaging voltage inverter with an RC constant adjustable between sub-second to seconds.
You'd have your all-DC amplifier; you'd need maybe 1 more stage (to make up for the additive nature of –6 dB stage-gain loss), and away you'd go.
Just saying.
GoatGuy
You know, say a +300 V power supply and a –300 V supply.
Let's have a 12×X7 passing 2 ma, fed to a 75 kΩ anode resistor. Voltage drop is 150 V, so the anode is comfortably at 150 V quiescent. The grid of the next stage might nominally have a 470 kΩ grid-to-ground resistor (nominally). Something somewhat bigger than the driving stage's effective impedance. (3× to 10× seems rational)
Well, if instead of the 470 kΩ grid-to-ground, it is instead two resistors of 470 kΩ in series, with the center tap going to the grid, and the “bottom one” going to –150 V supply (instead of ground), well … that grid is going to nominally be 0.0 V, isn't it?
YES: the signal is also divided in half (–6 dB of gain loss). Who cares! Now the signal can be used as direct-DC coupled, without a capacitor anywhere.
TRICK IS… how to regulate the –150 volt path?
I should think that this is where a bit of sand-state comes in handy. A very high impedance 'leak resistor' coming from Stage 1's anode to a high voltage emitter-follower to charge up an averaging capacitor (without getting the cap in the signal path), then a high-voltage ganged op-amp to invert the voltage relative to ground (sign reversal). now you have the negative supply, which only need supply microamps. A wee bit of “balancing the pair of series resistors” (at the top of this exercise) gets one darn close to 0.0 volts. Close enough. Just something other than +150 volts.
Heck … this sounds perfect as a drop-in module, black-box, tidy, doing its job. An averaging voltage inverter with an RC constant adjustable between sub-second to seconds.
You'd have your all-DC amplifier; you'd need maybe 1 more stage (to make up for the additive nature of –6 dB stage-gain loss), and away you'd go.
Just saying.
GoatGuy
PS: since the LP network on the 'bias inverter' is auto-adaptive to whatever the quiescent state is of the driving stage, even as it ages, or as alternate tubes are substitutes, it'd continue to establish complimentary DC blocking. Just as nicely as “cathode resistors” do that job (they, with bypass C, make a self-adapting 'bucking' power supply of sorts). … GoatGuy …
You can do something similar with a stacked supply topology, where the ground of the following stage is referenced to the B+ of the input stage.
Member
Joined 2009
Paid Member
If the dc sound is all about grid drive then does IT drive sound the same as dc too (when the IT is done well)?
I've read many a testament that 12ax7 can drive a 2a3 when done dc-wise this isn't a strong driver yet reports say it sounds great. Is there more going on?
I've read many a testament that 12ax7 can drive a 2a3 when done dc-wise this isn't a strong driver yet reports say it sounds great. Is there more going on?
No question IT drive can sound good if done well. Good ITs are expensive however, and require a low impedance drive to perform best as well.
12ax7 has no problem dc driving a 2a3 grid. The 12ax7 output impedance maybe high, but with no grid leak resistor required, the 2a3 grid's input impedance is essentially infinite. Just make sure the 12ax7 is not current starved, to prevent HF rolloff.
12ax7 has no problem dc driving a 2a3 grid. The 12ax7 output impedance maybe high, but with no grid leak resistor required, the 2a3 grid's input impedance is essentially infinite. Just make sure the 12ax7 is not current starved, to prevent HF rolloff.
Member
Joined 2009
Paid Member
Maybe a question I can only know the answer by experimenting - how would 12ax7 compare with or without MOSFET source follower for driving a 2A3 grid?
Assuming that your 12AX7 is a voltage amplifier, you won't be able to drive the grid of the 2A3 positive very well since it will presumably be the 12AX7's load that is pulling up the grid. So if you want to do A2 with the 2A3, the follower will do it much better. If you don't want to do A2, then you can probably do it with the 12AX7 (I am assuming, I haven't run any numbers).
Will there be a big audible difference between the two when they are not driving the grid positive? I don't know.
Take a look at the direct-coupled amplifier chapter 11 in this classic book. See p. 478 and p. 486 for example.
https://www.jlab.org/ir/MITSeries/V18.PDF
Last edited:
MIT Press... even back in the ~tube era~, was making downright impressive technical tomes. Thanks!
The MIT Radiation Lab series came out of WWII research, and was put into print in the years after the war.
There were 28 volumes covering a wide range of areas, including an index.
https://www.jlab.org/ir/MITSeries.html
Somewhere in a box, I have an electrical engineering textbook from MIT printed in the late 20's IIRC. I found it at an antique shop 15 years or so ago.
At that time I was pursuing my masters in engineering degree, and the stuff in the book was over my head, so I bought the book. Some of my professors were impressed with the degree of science and math that MIT threw at "such primitive technology."
It isn't such a big deal with the 2A3 but there are 4 reasons I use mosfet followers to drive power output tubes.
Overload recovery, Miller capacitance, driver linearity, and removing the need for AC current to pass "through" a coupling capacitor.
The cap is moved to the gate of a mosfet where the gate bias resistor can be larger than a megohm. The gate to source capacitance of the fet is bootstrapped, so it isn't an issue, and high voltage fets with a gate to drain capacitance in the 5pF range are common, so the driver tube and the coupling cap are working into a very light load.
At that time I was pursuing my masters in engineering degree, and the stuff in the book was over my head, so I bought the book. Some of my professors were impressed with the degree of science and math that MIT threw at "such primitive technology."
It isn't such a big deal with the 2A3 but there are 4 reasons I use mosfet followers to drive power output tubes.
Overload recovery, Miller capacitance, driver linearity, and removing the need for AC current to pass "through" a coupling capacitor.
The cap is moved to the gate of a mosfet where the gate bias resistor can be larger than a megohm. The gate to source capacitance of the fet is bootstrapped, so it isn't an issue, and high voltage fets with a gate to drain capacitance in the 5pF range are common, so the driver tube and the coupling cap are working into a very light load.
- Status
- Not open for further replies.