IF you are a tube purist, this post is not for you. If you are interested in the many ways we can reach higher levels of audio nirvana, please come on in.
The O2 headphone amplifier I built got me thinking about how I may have not given opamps a fair shake. The amp is one of the most transparent pieces I have ever used. This got me thinking about using an opamp in the front end of a tube amp. My research led me to this interesting little guy. MSK-130. Looking at this opamp, I think I could make an interesting amplifier. It looks like it has enough gain to make an amplifier with just output tubes.
With the output voltage swing of +/-95V it looks to me like it just falls short of driving a 300B, depending on what voltage you are at. Am I right about this? Is it ok to fall short here?
Are there any other interesting things we could use this for? Maybe an interstage instead of a front end (similar to powerdrive)?
I am still a newbie when it comes to designing amplifier instead of building them. So if I missed some fundamental reason this couldn't be done, please explain it to me so I don't make the same mistake twice.
The O2 headphone amplifier I built got me thinking about how I may have not given opamps a fair shake. The amp is one of the most transparent pieces I have ever used. This got me thinking about using an opamp in the front end of a tube amp. My research led me to this interesting little guy. MSK-130. Looking at this opamp, I think I could make an interesting amplifier. It looks like it has enough gain to make an amplifier with just output tubes.
With the output voltage swing of +/-95V it looks to me like it just falls short of driving a 300B, depending on what voltage you are at. Am I right about this? Is it ok to fall short here?
Are there any other interesting things we could use this for? Maybe an interstage instead of a front end (similar to powerdrive)?
I am still a newbie when it comes to designing amplifier instead of building them. So if I missed some fundamental reason this couldn't be done, please explain it to me so I don't make the same mistake twice.
It's a great idea and definitely worth trying, though I would never say that opamps sound transparent.
All tube amps have op-amp front ends if you consider the recording studio a "front end". Most of the music we hear has gone thought many TL072 (or the like) op amps. Not saying this is good, just that it is the way it is.
I've never understood why n opamp inside an audio power amp does not work. I've tried simulations and the results are poor. You'd think that with such a huge amount of gain the NFB would work well to correct out any problems but I can never gt the total distortion numbers to be reasonable if I include an op-amp in the loop.
As line level devices they are nearly perfect, just can't get them to work in a power amp. Could be that I'm simply not smart enough, but I'm not alone in that.
Yes, it looks similar to the LME49810 driver I am using on my tracer. It works quite well. I am planning to use a pair to drive 6336 in SE.
I am not saying that opamps are transparent per se. I am saying that the implementation of the opamps in the O2 are transparent. If you have not built one and tried it yourself, I highly recommend it.
Somewhere on the Tubecad site several years ago, Broskie had a circuit of a standard op-amp driving a PP EL-84. Simple enough. I've thought about doing it, but then you need a +/- 15v power supply as well, and I'm not sure if there would be any benefit.
)
I bet that chip costs near $1000. The big high voltage opamps from Apex used to cost $150 or so...
I have found the chip for ~ $85 from one source online so price is possibly not a problem. I know that there have been opamp front end tube designs before but I have not seen one that operated at the same voltage as the output stage. I also have not seen one that could swing such large V at such large currents. What do you think would be the largest tube(s) this could push in a 2 stage amp where the opamp is the first stage?
There is no reason why this should be necessary. Usually, tube output stages have gain.
You don't need much current either to drive a tube in usual configuration.
Depends on type and amount of feedback. But as you see, EL34 is not a problem and since you can parallel them, there are no limits.
"There is no reason why this should be necessary. Usually, tube output stages have gain."
KISS - Why build an amp with 2 power supplies when one will suffice?
"You don't need much current either to drive a tube in usual configuration."
True, but I have a hunch that some front ends do not supply enough current for certain musical events. I think that a front end that can provide a current much larger then needed at all times will have a positive affect. This is only a hunch however. I am not sure tube purists care about this as the only thing they accept to drive a tube is another tube.
KISS - Why build an amp with 2 power supplies when one will suffice?
"You don't need much current either to drive a tube in usual configuration."
True, but I have a hunch that some front ends do not supply enough current for certain musical events. I think that a front end that can provide a current much larger then needed at all times will have a positive affect. This is only a hunch however. I am not sure tube purists care about this as the only thing they accept to drive a tube is another tube.
Let's see.
Not that it'll matter for a power amp, but with all the FETs, the MSK-130 is going to be rather noisy, especially at low frequencies. The noise graph shows 4-10 nV/rtHz, with clear indication of 1/f noise (typical of FETs). A good low noise bipolar amp is under 2 nV/rtHz, fairly flat over frequency. (Since this is input-referred noise, it doesn't matter that the output is on a higher voltage scale than the average amp.)
PSRR isn't very good. It falls steadily from low frequencies and probably hits ~0dB at about 1MHz. You'll want to use ferrite beads, extra decoupling and some shielding if possible, when using this in a noisy environment.
Open loop gain is quite good, and the user compensation option allows some flexibility. However, given the high voltage, the gain could easily stand to be 30 or 40dB higher still; proportionally speaking, the distortion and bandwidth aren't impressive. The common-mode input range is what you'd expect, so you have to use this device for some gain if you want the output to go reasonably far between rails. That is, you can't wire it as a unity-gain follower and expect useful results.
Since gain is fairly low, bandwidth is similarly on the low side. Even nearly uncompensated (only 3.3pF, compared to the 68pF + 100 ohm unity-stable network), gain crosses 40dB at 1MHz, which isn't bad (100MHz GBW product!), but if you have to use it at 40dB gain to begin with, you get a paltry 1MHz bandwidth (and no distortion-reducing feedback at that frequency).
All said, it'll work well for the specific applications where you'd need it -- for audio, you don't have any problems (20kHz gets a good 34dB of NFB, keeping distortion down). If you had something like, say, piezoelectric actuators (that don't need to run particularly fast!), they'll be stable and precise. At a gain of 40dB, you're looking at ~0.5uV/rtHz midband noise, or 50uV over a 10kHz bandwidth. Out of say 100Vrms output, that's a good 126dB dynamic range, getting worse at LF (<500Hz or so) when 1/f noise gets worse, at which point an outer positional servo might take over.
By far and large, however, the electrical specs are completely unimportant to this device. Why? Take a closer look: yes it's in a somewhat familiar tabbed SIP package, but this is no HEPTAWATT amplifier. This is also no monolithic chip. This is a thick-film hybrid, built on an alumina substrate. Thermal matching will be poor between transistors, so the offset, drift and noise specs are under steady conditions; these will vary widely under load and under temperature gradients. Ceramic is EXPENSIVE. I mean, not really, chip resistors are made on the same stuff (one-side-metallized alumina), but hybrids always have been, and always will be, a high cost product. You don't want to torque down those screws unevenly and crack your only device, and that's if you can even find one. Besides the price, availability of these must be somewhere between zero and nonexistent. If you look at the manufacturer's website, they mainly make aerospace components. And they probably make these to order. Stock? What's that...
Tim
Not that it'll matter for a power amp, but with all the FETs, the MSK-130 is going to be rather noisy, especially at low frequencies. The noise graph shows 4-10 nV/rtHz, with clear indication of 1/f noise (typical of FETs). A good low noise bipolar amp is under 2 nV/rtHz, fairly flat over frequency. (Since this is input-referred noise, it doesn't matter that the output is on a higher voltage scale than the average amp.)
PSRR isn't very good. It falls steadily from low frequencies and probably hits ~0dB at about 1MHz. You'll want to use ferrite beads, extra decoupling and some shielding if possible, when using this in a noisy environment.
Open loop gain is quite good, and the user compensation option allows some flexibility. However, given the high voltage, the gain could easily stand to be 30 or 40dB higher still; proportionally speaking, the distortion and bandwidth aren't impressive. The common-mode input range is what you'd expect, so you have to use this device for some gain if you want the output to go reasonably far between rails. That is, you can't wire it as a unity-gain follower and expect useful results.
Since gain is fairly low, bandwidth is similarly on the low side. Even nearly uncompensated (only 3.3pF, compared to the 68pF + 100 ohm unity-stable network), gain crosses 40dB at 1MHz, which isn't bad (100MHz GBW product!), but if you have to use it at 40dB gain to begin with, you get a paltry 1MHz bandwidth (and no distortion-reducing feedback at that frequency).
All said, it'll work well for the specific applications where you'd need it -- for audio, you don't have any problems (20kHz gets a good 34dB of NFB, keeping distortion down). If you had something like, say, piezoelectric actuators (that don't need to run particularly fast!), they'll be stable and precise. At a gain of 40dB, you're looking at ~0.5uV/rtHz midband noise, or 50uV over a 10kHz bandwidth. Out of say 100Vrms output, that's a good 126dB dynamic range, getting worse at LF (<500Hz or so) when 1/f noise gets worse, at which point an outer positional servo might take over.
By far and large, however, the electrical specs are completely unimportant to this device. Why? Take a closer look: yes it's in a somewhat familiar tabbed SIP package, but this is no HEPTAWATT amplifier. This is also no monolithic chip. This is a thick-film hybrid, built on an alumina substrate. Thermal matching will be poor between transistors, so the offset, drift and noise specs are under steady conditions; these will vary widely under load and under temperature gradients. Ceramic is EXPENSIVE. I mean, not really, chip resistors are made on the same stuff (one-side-metallized alumina), but hybrids always have been, and always will be, a high cost product. You don't want to torque down those screws unevenly and crack your only device, and that's if you can even find one. Besides the price, availability of these must be somewhere between zero and nonexistent. If you look at the manufacturer's website, they mainly make aerospace components. And they probably make these to order. Stock? What's that...
Tim
Last edited:
Slightly off topic...
I've had some success using TI OPA452 / OPA453 (~$5.00) driving grids into class A2... but the 452 is limited to +/-36V or 72V total supply. There's also the OPA454 that does 100V but is in a surface-mount package only.
Arguably not the highest fidelity amplifier, though it measures surprisingly good.
Class-A2 and a power opamp driver
Pete
I've had some success using TI OPA452 / OPA453 (~$5.00) driving grids into class A2... but the 452 is limited to +/-36V or 72V total supply. There's also the OPA454 that does 100V but is in a surface-mount package only.
Arguably not the highest fidelity amplifier, though it measures surprisingly good.
Class-A2 and a power opamp driver
Pete
pmillett I actually was just poking around your website and saw the work you did with opamps and decided to check out my old posting. Have you played around at all with a push pull opamp configuration yet?
any chance you assembled a pair and gave them a real listening to? It seems like you could design a very affordable tube amp with a setup like that.
any chance you assembled a pair and gave them a real listening to? It seems like you could design a very affordable tube amp with a setup like that.
Nope, I never did go any further with that one.
Right now I'm playing with class AB2 using MOSFET source followers...
Pete
Right now I'm playing with class AB2 using MOSFET source followers...
Pete
Another one has seen the light.
One little red driver board + 2 big Fuji mosfets + two unsuspecting output tubes = big power and low distortion.
- Status
- Not open for further replies.