I have been exploring various headphone amp circuits and I had an idea that since I havent run across it anywhere probably wont work, but what the heck, here goes..
In a tube gain stage - FET source follower output stage, assuming a 1 ohm Zout, could one use a 1 to 1 transformer in place of a big electrolytic output cap? My thought is that the 1 to 1 transformer will reflect back the headphones Zin, and regardless of what it is, it should be a good match for the follower. This assumes enough voltage gain and quiescent current to run a wide range of headphones.
In a tube gain stage - FET source follower output stage, assuming a 1 ohm Zout, could one use a 1 to 1 transformer in place of a big electrolytic output cap? My thought is that the 1 to 1 transformer will reflect back the headphones Zin, and regardless of what it is, it should be a good match for the follower. This assumes enough voltage gain and quiescent current to run a wide range of headphones.
Attachments
Well it would allow you to use a presumably small and better sounding off the shelf transformer for a universal headphone amplifier, without having to have an expensive custom one wound with multiple secondary taps. Some people prefer the sound of OTs over electrolytics. I think it would be fun to try all three ways of coupling: big cap, OT, and DC servo.
I havent done the research into transformer inductance and OT frequency response yet, but given how low the power output needs of most headphones are, you might be able to use a small line transformer. Which actually leads me to another question. If you look at the frequency response plots of headphones here:
Headphone Measurements | InnerFidelity
You will see that the low frequency response of many, Im tempted say most, well regarded headphones starts rolling off significantly below 100 hz. Maybe you dont need a transformer that has enough inductance to give you flat response to 20hz and instead can optimize for mid and high frequencies.
I havent done the research into transformer inductance and OT frequency response yet, but given how low the power output needs of most headphones are, you might be able to use a small line transformer. Which actually leads me to another question. If you look at the frequency response plots of headphones here:
Headphone Measurements | InnerFidelity
You will see that the low frequency response of many, Im tempted say most, well regarded headphones starts rolling off significantly below 100 hz. Maybe you dont need a transformer that has enough inductance to give you flat response to 20hz and instead can optimize for mid and high frequencies.
> in place of a big electrolytic output cap?
You show a cap driving the transformer. So no difference.
10K:10K will be a poor fit for 32-600r phones. DCR will be 500r-1,000R per winding, absorbing the majority of your power. 600:600 would be a better bet.
The step-down is useful. Headphones need a few V at many mA. Your tube wants many V (24V-300V) but if you run many mA at high voltage you have a heater as well as an amplifier. Note that 600r:600"split" is a standard part that can be wired 600:150, which is a good improvement.
> assuming a 1 ohm Zout
That's surely the Small Signal Zout. In a "Power Amp" (even if milli-Power) you want to know the Large Signal Zout. Since the Source Follower is a Heavy NFB stage, small Zout is much lower than the limit for large signals. Since the MOSFET will probably go to few-Ohm, the limit is that Source resistor. If it is a typical 470 Ohms, pencil "470" as your large-signal Zout.
Resistance-coupled amplifier is at-most 6% efficient. If you find a transformer which will carry DC happy you can do 50% efficient (means less heat or less THD at the same power in/out). However I don't know of any. A 5W 5K:16 OT would work with high high power supply (300V).
You show a cap driving the transformer. So no difference.
10K:10K will be a poor fit for 32-600r phones. DCR will be 500r-1,000R per winding, absorbing the majority of your power. 600:600 would be a better bet.
The step-down is useful. Headphones need a few V at many mA. Your tube wants many V (24V-300V) but if you run many mA at high voltage you have a heater as well as an amplifier. Note that 600r:600"split" is a standard part that can be wired 600:150, which is a good improvement.
> assuming a 1 ohm Zout
That's surely the Small Signal Zout. In a "Power Amp" (even if milli-Power) you want to know the Large Signal Zout. Since the Source Follower is a Heavy NFB stage, small Zout is much lower than the limit for large signals. Since the MOSFET will probably go to few-Ohm, the limit is that Source resistor. If it is a typical 470 Ohms, pencil "470" as your large-signal Zout.
Resistance-coupled amplifier is at-most 6% efficient. If you find a transformer which will carry DC happy you can do 50% efficient (means less heat or less THD at the same power in/out). However I don't know of any. A 5W 5K:16 OT would work with high high power supply (300V).
A transformer is worth using when you need what a transformer can do which other components cannot do: wideband impedance transformation, common-mode isolation.
Using 1:1 transformer instead of (well, actually, as well as!) an output capacitor means swapping a cheap low distortion very wideband component for an expensive higher distortion narrower band component. This does not seem to be wise engineering, except for those who prefer the distortions which transformers introduce.
Using 1:1 transformer instead of (well, actually, as well as!) an output capacitor means swapping a cheap low distortion very wideband component for an expensive higher distortion narrower band component. This does not seem to be wise engineering, except for those who prefer the distortions which transformers introduce.
Thanks for all your replies. Before addressing your various objections I just want to make sure that my original question ie that using a 1 to 1 OT after a fet follower with a 1 ohm output impedance will work from an impedance matching perspective at least, with any headphones from 12 - 600 ohms as I describe, ie the FET will see whatever the Z in of the headphones are, within reason.
@6A3sUMMER The diode clamps the voltage to 12 volts. This is a standard circuit without the OT, you can read a good description of it here at ESP or in Merlin Blencowe's great book "Designing High-Fidelity Tube Preamps"
@PRR >You show a cap driving the transformer. So no difference.
Im showing a cap because I was thinking about the possibility of using a line transformer, in which case it would a 10uf or less polypropylene. However thinking about it maybe Edcor would provide a 600:600 XSE or GXSE SE OT either of which can handle the 10 - 50ma of quiescent current which I think is all you would need in this situation. I will find out today.
>That's surely the Small Signal Zout. In a "Power Amp" (even if milli-Power) you want to know the Large Signal Zout. Since the Source Follower is a Heavy NFB stage, small Zout is much lower than the limit for large signals. Since the MOSFET will probably go to few-Ohm, the limit is that Source resistor. If it is a typical 470 Ohms, pencil "470" as your large-signal Zout.
Ah, I didn't know this and maybe this kills the idea. Is there some way to keep the Zout down across the range of likely signal strengths? Across a broad range of headphone impedances/sensitivity combinations, the power needed for 110db peaks is less than 25mW. What if use a CCS instead of a resistor?
@DF96 Probably not wise engineering, neither is using tubes, ; ). Fun to explore unusual possibilities.
@6A3sUMMER The diode clamps the voltage to 12 volts. This is a standard circuit without the OT, you can read a good description of it here at ESP or in Merlin Blencowe's great book "Designing High-Fidelity Tube Preamps"
@PRR >You show a cap driving the transformer. So no difference.
Im showing a cap because I was thinking about the possibility of using a line transformer, in which case it would a 10uf or less polypropylene. However thinking about it maybe Edcor would provide a 600:600 XSE or GXSE SE OT either of which can handle the 10 - 50ma of quiescent current which I think is all you would need in this situation. I will find out today.
>That's surely the Small Signal Zout. In a "Power Amp" (even if milli-Power) you want to know the Large Signal Zout. Since the Source Follower is a Heavy NFB stage, small Zout is much lower than the limit for large signals. Since the MOSFET will probably go to few-Ohm, the limit is that Source resistor. If it is a typical 470 Ohms, pencil "470" as your large-signal Zout.
Ah, I didn't know this and maybe this kills the idea. Is there some way to keep the Zout down across the range of likely signal strengths? Across a broad range of headphone impedances/sensitivity combinations, the power needed for 110db peaks is less than 25mW. What if use a CCS instead of a resistor?
@DF96 Probably not wise engineering, neither is using tubes, ; ). Fun to explore unusual possibilities.
> handle the 10 - 50ma of quiescent current
Ah, your image in post #1 shows no DC in the OT. So not sure what your plans are.
> @DF96 Probably not wise engineering, neither is using tubes, ; ).
In headphone driving it is not absurd to use a dump-truck to fill a flower-pot. So yeah: use the lowest efficiency topology and don't quantify device and load V and I, you can probably be happy without breaking the bank.
Ah, your image in post #1 shows no DC in the OT. So not sure what your plans are.
> @DF96 Probably not wise engineering, neither is using tubes, ; ).
In headphone driving it is not absurd to use a dump-truck to fill a flower-pot. So yeah: use the lowest efficiency topology and don't quantify device and load V and I, you can probably be happy without breaking the bank.
>In headphone driving it is not absurd to use a dump-truck to fill a flower-pot. So yeah: use the lowest efficiency topology and don't quantify device and load V and I, you can probably be happy without breaking the bank.
Oh come on. I'm not showing a more complete circuit because i'm looking for help on whether or not the basic idea will work at all and whether or not one uses a gapped OT or cap coupled one is irrelevant to the basic question. I'm simply trying to learn more, experiment, and have some fun. Questions of efficiency at the level of headphone amplifier power are absurd on a forum where people help each other build high wattage Class A OTL tube power amplifiers and 1000 watt plus volts SE transmitting tube amps.
One of the things that make headphone amplifiers interesting to play with is that the power levels are so low you can explore all kinds of crazy stuff for the Sheer Hell of It, that would be completely impractical at loudspeaker levels. If this kind of thing interests you great, if not that's fine too.
As it stands based on the responses, it looks like what I want to do will work and the remaining issues I have to explore are:
What are my options for setting up a source follower output stage so that Zout stays close to 1 ohm at signal peaks?
What is the optimal impedance of the OT primary and secondary, given a range of 12 to 600 ohm headphones. PRR has suggested 600 ohms as a good place to start because these are readily available. Id like to know what the theoretical optimum might be in this application and how to calculate it.
What minimum inductance is needed in the OT for 3db down low frequency response to 20hz and given real world headphone response maybe closer to 100hz and how to calculate this? How low a response can I expect from off the shelf line transformers? Are any potentially usable?
For a voltage stage I now feel inclined to do something equally wasteful and eccentric. How about the Salas 6V6 linestage. As I recall it has a gain of around 12db which should be on the plenty side of reasonable and a wonderfully inappropriate use of a device.
If anyone has any positive suggestions or show stoppers (especially as in "Stop!, this ain't going to work"), related to all this I would appreciate them, otherwise I will post more after further research.
I think I will name this the Totally Inappropriate Least Efficient Topology Eccentric Headphone Amplifier. TILETEHP for short.
Oh come on. I'm not showing a more complete circuit because i'm looking for help on whether or not the basic idea will work at all and whether or not one uses a gapped OT or cap coupled one is irrelevant to the basic question. I'm simply trying to learn more, experiment, and have some fun. Questions of efficiency at the level of headphone amplifier power are absurd on a forum where people help each other build high wattage Class A OTL tube power amplifiers and 1000 watt plus volts SE transmitting tube amps.
One of the things that make headphone amplifiers interesting to play with is that the power levels are so low you can explore all kinds of crazy stuff for the Sheer Hell of It, that would be completely impractical at loudspeaker levels. If this kind of thing interests you great, if not that's fine too.
As it stands based on the responses, it looks like what I want to do will work and the remaining issues I have to explore are:
What are my options for setting up a source follower output stage so that Zout stays close to 1 ohm at signal peaks?
What is the optimal impedance of the OT primary and secondary, given a range of 12 to 600 ohm headphones. PRR has suggested 600 ohms as a good place to start because these are readily available. Id like to know what the theoretical optimum might be in this application and how to calculate it.
What minimum inductance is needed in the OT for 3db down low frequency response to 20hz and given real world headphone response maybe closer to 100hz and how to calculate this? How low a response can I expect from off the shelf line transformers? Are any potentially usable?
For a voltage stage I now feel inclined to do something equally wasteful and eccentric. How about the Salas 6V6 linestage. As I recall it has a gain of around 12db which should be on the plenty side of reasonable and a wonderfully inappropriate use of a device.
If anyone has any positive suggestions or show stoppers (especially as in "Stop!, this ain't going to work"), related to all this I would appreciate them, otherwise I will post more after further research.
I think I will name this the Totally Inappropriate Least Efficient Topology Eccentric Headphone Amplifier. TILETEHP for short.
How about using a pair of these matching transformers:
EDCOR - PCW10K-7K/300-32
The circuit is called a Para-feed output, or Parallel Feed output.
The topology is sound and there are numerous implementations if you search for them.
EDCOR - PCW10K-7K/300-32
The circuit is called a Para-feed output, or Parallel Feed output.
The topology is sound and there are numerous implementations if you search for them.
Last edited:
Interesting! I wasn't aware Edcor made a headphone transformer like this, thanks for pointing it out. Im still going to work on the 600:600 idea, but I will probably try this as well at some point.
Different idea than mine but the use of the gyrator is interesting. Im not sure I understand the advantage of it as a plate load over a straight CCS. Its true that you need close to twice the plate voltage for B+ than you would, but with a headphone amp, that could be 200 - 300 volts depending on the tube you use and you get the added PSRR.
Last edited:
The advantage is, stable anode voltage independent on the tube you inserted, filament voltage, tube's age, etc... Think of the gyrator load as of a CCS with servo.
Me thinks your proposal is triple flawed
1. If your using a FET or any kind power transistor in the output stage, your Zout will be so low that you can (and should!) drive your phones directly. Adding any other cruft like tubes, caps or xfmrs will only get in the way in addition to adding cost and complexity. (Of course then you won't be listening thru a toob amp anymore, so if that's your point...)
2. As others have pointed out, you are not replacing the output cap with a transformer here, your are adding a transformer to your topology. This makes absolutely no sense at all.
Or more generally, using 1-to-1 transformers in headphone amps makes no sense at all. If you're absolutely sold to the idea of using a OPT, at least use a step down type to benefit from impedance matching capabilities of transformers.
3. Headphone tube amps are predestined to be driven by OTL topologies. Try a 6N6P in WCF topology as a starting point.
>Or more generally, using 1-to-1 transformers in headphone amps makes no sense at all.
17 posts in and someone addresses the question that is the reason for the thread! Though I have enjoyed the other posts, especially the gyrator stuff : )
Why? Why doesn't it make any sense at all? I assume my understanding of what would be going on is wrong. Which is why I posted the question. Which is: Assuming you can operate the source follower so that it has 1 ohm Zout or thereabouts, then a 1:1 OT will reflect back whatever the Zin of the headphones is, creating a good impedance match and thus allowing you to drive any pair of headphones from an impedance matching perspective, with a higher quality transformer due to the low windings ratio. Forget everything else for the moment and please just tell me if this is right or wrong.
17 posts in and someone addresses the question that is the reason for the thread! Though I have enjoyed the other posts, especially the gyrator stuff : )
Why? Why doesn't it make any sense at all? I assume my understanding of what would be going on is wrong. Which is why I posted the question. Which is: Assuming you can operate the source follower so that it has 1 ohm Zout or thereabouts, then a 1:1 OT will reflect back whatever the Zin of the headphones is, creating a good impedance match and thus allowing you to drive any pair of headphones from an impedance matching perspective, with a higher quality transformer due to the low windings ratio. Forget everything else for the moment and please just tell me if this is right or wrong.
A 1-to-1 transformer does not provide any impedance matching. Your tube output stage still has to provide all the high current drive with low Zout. NFB helps with the Zout, and lots of it is preferred anyways to tackle the wide range of headphone impedances wich can occur. Under these operating conditions, it is preferred to use an output cap as a cheaper and better performing alternative.
Cant respond in detail at the moment but want to point out that my output stage in the example is a fet source follower not a tube. Hence the possibility of 1 ohm Zout (in simple source follower or if not some other SS buffer), hence the idea to begin with.
Another side question is: given a tube gain stage, would there be any advantage to using an emitter follower rather than a source follower in a similar circuit for this application?
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.